{ "schema_version": 1, "checked_through": "2026-10-21", "source_files": [ "README.md", "data/literature-2026-10.json" ], "date_note": "Baseline dates retain README month precision. Update dates and availability were checked through the stated cutoff. Building the website does not refresh those facts.", "focus_note": "Editorial browsing labels. Supporting environment preserves historical world-generation and simulation work as context; it does not assert demonstrated improvements to a game player.", "categories": [ "Dataset", "Model", "Harness", "Benchmark", "Survey" ], "counts": { "works": 210, "source_rows": 257, "update_works": 32, "by_category": { "Dataset": 31, "Model": 75, "Harness": 46, "Benchmark": 98, "Survey": 7 }, "by_focus": { "supporting environment": 24, "player": 180, "survey": 6 }, "baseline_works": 178 }, "entries": [ { "id": "arxiv-2610.27999", "name": "Matrix-Game 4.0", "title": "Matrix-Game 4.0: Harnessing Video Models for Real-Time Playable Game Worlds", "date": "2026-10-15", "date_precision": "day", "date_basis": "Official repository release news is October 15; repository news reports technical report October 16; arXiv v1 submitted October 20.", "venue": "arXiv", "categories": [ "Model", "Harness" ], "primary_category": "Model", "keywords": [ "Engine-Grounded WM", "Causal Neural Rendering", "Real-Time Gameplay", "Executable Game Engine", "Coding-Agent Orchestration", "Neural Renderer" ], "keywords_by_category": { "Model": "Engine-Grounded WM, Causal Neural Rendering, Real-Time Gameplay", "Harness": "Executable Game Engine, Coding-Agent Orchestration, Neural Renderer" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2610.27999" }, { "label": "Project", "url": "https://matrix-game-v4-0.github.io/page/" }, { "label": "Code", "url": "https://github.com/SkyworkAI/Matrix-Game" }, { "label": "Models", "url": "https://huggingface.co/Skywork/Matrix-Game-4.0" } ], "summary": "Combines an executable game engine, coding-agent orchestration, and a trained causal video renderer for low-latency playable shooter and driving worlds.", "release_status": "Inference code, browser demo, game source, model weights, and assets for three scenes are available.", "notes": "The engine owns persistent state and game rules; the video model renders low-poly proxies. Related world-generation infrastructure, not an autonomous native-action game-playing policy. A secondary Harness tag reflects the executable engine and coding agent; do not call it a fully learned WAM. Released assets are not a training dataset release.", "source": "update", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 92, "date": "2026-10", "title": "Matrix-Game 4.0: Harnessing Video Models for Real-Time Playable Game Worlds", "venue": "arXiv", "keywords": "Engine-Grounded WM, Causal Neural Rendering, Real-Time Gameplay", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2610.27999" }, { "label": "Project", "url": "https://matrix-game-v4-0.github.io/page/" }, { "label": "Code", "url": "https://github.com/SkyworkAI/Matrix-Game" }, { "label": "Models", "url": "https://huggingface.co/Skywork/Matrix-Game-4.0" } ] }, { "section": "Harness", "line": 175, "date": "2026-10", "title": "Matrix-Game 4.0: Harnessing Video Models for Real-Time Playable Game Worlds", "venue": "arXiv", "keywords": "Executable Game Engine, Coding-Agent Orchestration, Neural Renderer", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2610.27999" }, { "label": "Project", "url": "https://matrix-game-v4-0.github.io/page/" }, { "label": "Code", "url": "https://github.com/SkyworkAI/Matrix-Game" } ] } ], "sources": [ "https://arxiv.org/abs/2610.27999", "https://github.com/SkyworkAI/Matrix-Game", "https://matrix-game-v4-0.github.io/page/", "https://huggingface.co/Skywork/Matrix-Game-4.0" ], "paper_submission_date": "2026-10-20", "search_text": "matrix-game 4.0: harnessing video models for real-time playable game worlds matrix-game 4.0 2026-10-15 arxiv model harness engine-grounded wm causal neural rendering real-time gameplay executable game engine coding-agent orchestration neural renderer combines an executable game engine, coding-agent orchestration, and a trained causal video renderer for low-latency playable shooter and driving worlds. the engine owns persistent state and game rules; the video model renders low-poly proxies. related world-generation infrastructure, not an autonomous native-action game-playing policy. a secondary harness tag reflects the executable engine and coding agent; do not call it a fully learned wam. released assets are not a training dataset release. supporting environment" }, { "id": "work-09f299a049d3", "name": "SpireClimb", "title": "SpireClimb", "date": "2026-10-14", "date_precision": "day", "date_basis": "Official launch-report publication date; no arXiv paper identified", "venue": "Benchmark / research report", "categories": [ "Benchmark" ], "primary_category": "Benchmark", "keywords": [ "Slay the Spire 1/2", "Visual Computer Use", "Maximum Difficulty", "Fixed Seeds" ], "keywords_by_category": { "Benchmark": "Slay the Spire 1/2, Visual Computer Use, Maximum Difficulty, Fixed Seeds" }, "links": [ { "label": "Report", "url": "https://samaritan-research.org/publications/ai-systems-are-improving-rapidly-at-slay-the-spire-and-its-sequel" }, { "label": "Project", "url": "https://samaritan-research.org/benchmarks/spireclimb" } ], "summary": "Evaluates screenshot-based mouse/keyboard agents on both Slay the Spire games at maximum difficulty. Each model plays 60 StS1 and 75 StS2 runs on fixed seeds; the benchmark averages the two game win rates and freezes the StS2 version.", "release_status": "Official benchmark methodology and launch report public; no dedicated author-linked code repository or standalone trajectory corpus verified.", "notes": "Uses Inspect ReAct with screenshots, persistent notes across context compaction, and wiki-restricted search. The earlier epoch.ai URL refers to this same benchmark.", "source": "update", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 270, "date": "2026-10", "title": "SpireClimb", "venue": "Benchmark / research report", "keywords": "Slay the Spire 1/2, Visual Computer Use, Maximum Difficulty, Fixed Seeds", "links": [ { "label": "Report", "url": "https://samaritan-research.org/publications/ai-systems-are-improving-rapidly-at-slay-the-spire-and-its-sequel" }, { "label": "Project", "url": "https://samaritan-research.org/benchmarks/spireclimb" } ] } ], "sources": [ "https://samaritan-research.org/benchmarks/spireclimb", "https://samaritan-research.org/publications/ai-systems-are-improving-rapidly-at-slay-the-spire-and-its-sequel", "https://epoch.ai/benchmarks/spireclimb" ], "search_text": "spireclimb spireclimb 2026-10-14 benchmark / research report benchmark slay the spire 1/2 visual computer use maximum difficulty fixed seeds evaluates screenshot-based mouse/keyboard agents on both slay the spire games at maximum difficulty. each model plays 60 sts1 and 75 sts2 runs on fixed seeds; the benchmark averages the two game win rates and freezes the sts2 version. uses inspect react with screenshots, persistent notes across context compaction, and wiki-restricted search. the earlier epoch.ai url refers to this same benchmark. player" }, { "id": "arxiv-2610.20584", "name": "Player-1", "title": "Player-1: Learning to Play from Synthetic Experience", "date": "2026-10-14", "date_precision": "day", "date_basis": "arXiv v1 and official repository news", "venue": "arXiv", "categories": [ "Dataset", "Model" ], "primary_category": "Model", "keywords": [ "Player-1-Data", "6", "044 Trajectories", "172 Games; Release Pending", "VLA", "Synthetic-Game SFT + PPO", "Keyboard/Mouse; Code & Weights Pending" ], "keywords_by_category": { "Dataset": "Player-1-Data, 6,044 Trajectories, 172 Games; Release Pending", "Model": "VLA, Synthetic-Game SFT + PPO, Keyboard/Mouse; Code & Weights Pending" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2610.20584" }, { "label": "Project", "url": "https://player-1-project.github.io/" }, { "label": "Repository", "url": "https://github.com/TencentARC/Player-1" } ], "summary": "Post-trains computer-use models with gameplay demonstrations and PPO in sixteen synthetic games to learn transferable keyboard-and-mouse control.", "release_status": "Paper and project page released; repository currently contains assets/README. Player-1-Data, Player-1-Gym, model checkpoints, and training/evaluation code are listed as planned releases.", "notes": "Correct paper title omits the search-result paraphrase 'An Agent That'. Player-1-Data contains 6,044 trajectories from 172 games. Player-1-Gym is a synthetic RL training suite, not a new held-out benchmark. Native controls include relative mouse movement, sustained inputs, and composite actions.", "source": "update", "focus": "player", "source_rows": [ { "section": "Model", "line": 93, "date": "2026-10", "title": "Player-1: Learning to Play from Synthetic Experience", "venue": "arXiv", "keywords": "VLA, Synthetic-Game SFT + PPO, Keyboard/Mouse; Code & Weights Pending", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2610.20584" }, { "label": "Project", "url": "https://player-1-project.github.io/" }, { "label": "Repository", "url": "https://github.com/TencentARC/Player-1" } ] }, { "section": "Dataset", "line": 231, "date": "2026-10", "title": "Player-1: Learning to Play from Synthetic Experience", "venue": "arXiv", "keywords": "Player-1-Data, 6,044 Trajectories, 172 Games; Release Pending", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2610.20584" }, { "label": "Project", "url": "https://player-1-project.github.io/" }, { "label": "Repository", "url": "https://github.com/TencentARC/Player-1" } ] } ], "sources": [ "https://arxiv.org/abs/2610.20584", "https://github.com/TencentARC/Player-1", "https://player-1-project.github.io/" ], "search_text": "player-1: learning to play from synthetic experience player-1 2026-10-14 arxiv dataset model player-1-data 6 044 trajectories 172 games; release pending vla synthetic-game sft + ppo keyboard/mouse; code & weights pending post-trains computer-use models with gameplay demonstrations and ppo in sixteen synthetic games to learn transferable keyboard-and-mouse control. correct paper title omits the search-result paraphrase 'an agent that'. player-1-data contains 6,044 trajectories from 172 games. player-1-gym is a synthetic rl training suite, not a new held-out benchmark. native controls include relative mouse movement, sustained inputs, and composite actions. player" }, { "id": "work-267945c6f012", "name": "Preference Model Slay the Spire benchmark", "title": "Slay the Spire — A Long-Horizon Reasoning Benchmark", "date": "2026-10-12", "date_precision": "day", "date_basis": "Official research blog displayed date 10.12.26 (October 12, corroborated by October publication search result); no arXiv paper identified", "venue": "Research blog", "categories": [ "Harness", "Benchmark" ], "primary_category": "Benchmark", "keywords": [ "Karotte RL Framework", "Headless StS", "MCP Tools", "Mechanic Lookup", "Rollout Logging", "Slay the Spire", "11 Models", "5 Scenarios", "1650 Runs", "Reasoning Effort" ], "keywords_by_category": { "Harness": "Karotte RL Framework, Headless StS, MCP Tools, Mechanic Lookup, Rollout Logging", "Benchmark": "Slay the Spire, 11 Models, 5 Scenarios, 1650 Runs, Reasoning Effort" }, "links": [ { "label": "Report", "url": "https://www.preferencemodel.com/blog/slay-the-spire/" }, { "label": "Project", "url": "https://ai-plays-slay-the-spire.com/" } ], "summary": "Benchmarks 11 models at three reasoning levels on five curated Slay the Spire scenarios, totaling 1,650 main runs. Its headless, text-tool harness adds mechanic lookup and action supervision; a public viewer presents run transcripts and videos.", "release_status": "Official report and transcript/video viewer public. Karotte framework is open source, but no benchmark-specific environment code or bulk trajectory download was verified; do not label the generic framework as the benchmark code.", "notes": "The public viewer is an evaluation artifact; a bulk dataset release was not verified. This study is separate from SpireBench and SpireClimb.", "source": "update", "focus": "player", "source_rows": [ { "section": "Harness", "line": 176, "date": "2026-10", "title": "Slay the Spire — A Long-Horizon Reasoning Benchmark", "venue": "Research blog", "keywords": "Karotte RL Framework, Headless StS, MCP Tools, Mechanic Lookup, Rollout Logging", "links": [ { "label": "Report", "url": "https://www.preferencemodel.com/blog/slay-the-spire/" }, { "label": "Project", "url": "https://ai-plays-slay-the-spire.com/" } ] }, { "section": "Benchmark", "line": 271, "date": "2026-10", "title": "Slay the Spire — A Long-Horizon Reasoning Benchmark", "venue": "Research blog", "keywords": "Slay the Spire, 11 Models, 5 Scenarios, 1650 Runs, Reasoning Effort", "links": [ { "label": "Report", "url": "https://www.preferencemodel.com/blog/slay-the-spire/" }, { "label": "Project", "url": "https://ai-plays-slay-the-spire.com/" } ] } ], "sources": [ "https://www.preferencemodel.com/blog/slay-the-spire/", "https://ai-plays-slay-the-spire.com/", "https://github.com/preferencemodel/karotte" ], "search_text": "slay the spire — a long-horizon reasoning benchmark preference model slay the spire benchmark 2026-10-12 research blog harness benchmark karotte rl framework headless sts mcp tools mechanic lookup rollout logging slay the spire 11 models 5 scenarios 1650 runs reasoning effort benchmarks 11 models at three reasoning levels on five curated slay the spire scenarios, totaling 1,650 main runs. its headless, text-tool harness adds mechanic lookup and action supervision; a public viewer presents run transcripts and videos. the public viewer is an evaluation artifact; a bulk dataset release was not verified. this study is separate from spirebench and spireclimb. player" }, { "id": "arxiv-2610.14158", "name": "Video-Arena", "title": "Video-Arena: A Scalable Agentic Video Benchmark for Spatiotemporal Capability", "date": "2026-10-09", "date_precision": "day", "date_basis": "arXiv v1 submission: 2026-10-09 17:27:16 UTC", "venue": "arXiv", "categories": [ "Dataset", "Benchmark" ], "primary_category": "Benchmark", "keywords": [ "Evaluation Videos & Ground Truth", "180 Cases", "9 Games", "Recorded-Game Video Understanding", "State Tracking", "Deterministic Judges" ], "keywords_by_category": { "Dataset": "Evaluation Videos & Ground Truth, 180 Cases, 9 Games", "Benchmark": "Recorded-Game Video Understanding, State Tracking, 9 Games, Deterministic Judges" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2610.14158" }, { "label": "Project", "url": "https://videoarena-bench.github.io/" }, { "label": "Code", "url": "https://github.com/THUSI-Lab/VideoArena" }, { "label": "Data", "url": "https://huggingface.co/datasets/THUSI-Lab/Video-Arena" } ], "summary": "A game-video benchmark of event perception, state-update interpretation, and hidden-state tracking, with native video and tool-assisted agentic protocols. It provides 180 cases across nine games, deterministic game-specific judges, and released video/ground-truth data.", "release_status": "Evaluation code, task definitions, judges, videos, and ground truth released; case-building pipelines not released.", "notes": "This is the THUSI-Lab game-video benchmark, not Arena.ai's video-generation arena. The nine games are Go, Chess, Gomoku, Doudizhu, CS2, DOTA 2, StarCraft II, Google Research Football, and Badminton. It evaluates understanding of recorded games rather than closed-loop game playing.", "source": "update", "focus": "player", "source_rows": [ { "section": "Dataset", "line": 232, "date": "2026-10", "title": "Video-Arena: A Scalable Agentic Video Benchmark for Spatiotemporal Capability", "venue": "arXiv", "keywords": "Evaluation Videos & Ground Truth, 180 Cases, 9 Games", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2610.14158" }, { "label": "Project", "url": "https://videoarena-bench.github.io/" }, { "label": "Code", "url": "https://github.com/THUSI-Lab/VideoArena" }, { "label": "Data", "url": "https://huggingface.co/datasets/THUSI-Lab/Video-Arena" } ] }, { "section": "Benchmark", "line": 272, "date": "2026-10", "title": "Video-Arena: A Scalable Agentic Video Benchmark for Spatiotemporal Capability", "venue": "arXiv", "keywords": "Recorded-Game Video Understanding, State Tracking, 9 Games, Deterministic Judges", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2610.14158" }, { "label": "Project", "url": "https://videoarena-bench.github.io/" }, { "label": "Code", "url": "https://github.com/THUSI-Lab/VideoArena" } ] } ], "sources": [ "https://arxiv.org/abs/2610.14158", "https://github.com/THUSI-Lab/VideoArena", "https://huggingface.co/datasets/THUSI-Lab/Video-Arena" ], "search_text": "video-arena: a scalable agentic video benchmark for spatiotemporal capability video-arena 2026-10-09 arxiv dataset benchmark evaluation videos & ground truth 180 cases 9 games recorded-game video understanding state tracking deterministic judges a game-video benchmark of event perception, state-update interpretation, and hidden-state tracking, with native video and tool-assisted agentic protocols. it provides 180 cases across nine games, deterministic game-specific judges, and released video/ground-truth data. this is the thusi-lab game-video benchmark, not arena.ai's video-generation arena. the nine games are go, chess, gomoku, doudizhu, cs2, dota 2, starcraft ii, google research football, and badminton. it evaluates understanding of recorded games rather than closed-loop game playing. player" }, { "id": "arxiv-2610.12596", "name": "SpireBench", "title": "SpireBench: Evaluating Long-Horizon Strategic Play of Agentic Models with Slay the Spire", "date": "2026-10-08", "date_precision": "day", "date_basis": "arXiv v1 submission 2026-10-08 18:18:19 UTC", "venue": "arXiv", "categories": [ "Harness", "Benchmark" ], "primary_category": "Benchmark", "keywords": [ "Full-Game Tool Interface", "Batched Actions", "State Diffs", "Cross-Run Memory", "Complete Slay the Spire", "A20", "20 LLMs", "Knowledge-Reskin and Memory Ablations" ], "keywords_by_category": { "Harness": "Full-Game Tool Interface, Batched Actions, State Diffs, Cross-Run Memory", "Benchmark": "Complete Slay the Spire, A20, 20 LLMs, Knowledge-Reskin and Memory Ablations" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2610.12596" } ], "summary": "Evaluates 20 LLMs on complete Slay the Spire runs at Ascension 20 using ten held-out seeds and human baselines. A full-game tool harness supports action batches and memory studies; consistent renaming of game terms probes prior-knowledge dependence.", "release_status": "Paper and detailed harness specification public; no author-linked public code or standalone dataset release verified.", "notes": "Original Slay the Spire (StS1) study by Lowe and Keemink, distinct from the spirebench.com / HermesBridge community project for Slay the Spire 2. No new learned policy or downloadable standalone corpus is claimed.", "source": "update", "focus": "player", "source_rows": [ { "section": "Harness", "line": 177, "date": "2026-10", "title": "SpireBench: Evaluating Long-Horizon Strategic Play of Agentic Models with Slay the Spire", "venue": "arXiv", "keywords": "Full-Game Tool Interface, Batched Actions, State Diffs, Cross-Run Memory", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2610.12596" } ] }, { "section": "Benchmark", "line": 273, "date": "2026-10", "title": "SpireBench: Evaluating Long-Horizon Strategic Play of Agentic Models with Slay the Spire", "venue": "arXiv", "keywords": "Complete Slay the Spire, A20, 20 LLMs, Knowledge-Reskin and Memory Ablations", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2610.12596" } ] } ], "sources": [ "https://arxiv.org/abs/2610.12596", "https://arxiv.org/html/2610.12596v1" ], "search_text": "spirebench: evaluating long-horizon strategic play of agentic models with slay the spire spirebench 2026-10-08 arxiv harness benchmark full-game tool interface batched actions state diffs cross-run memory complete slay the spire a20 20 llms knowledge-reskin and memory ablations evaluates 20 llms on complete slay the spire runs at ascension 20 using ten held-out seeds and human baselines. a full-game tool harness supports action batches and memory studies; consistent renaming of game terms probes prior-knowledge dependence. original slay the spire (sts1) study by lowe and keemink, distinct from the spirebench.com / hermesbridge community project for slay the spire 2. no new learned policy or downloadable standalone corpus is claimed. player" }, { "id": "arxiv-2610.12341", "name": "AAArena / Adversarial Heuristic Learning", "title": "Can AI Agents Learn Their Way to the Top? Evaluating Heuristic Learning in a Long-Running Game Agent Competition", "date": "2026-10-08", "date_precision": "day", "date_basis": "First arXiv submission date", "venue": "arXiv", "categories": [ "Dataset", "Harness", "Benchmark" ], "primary_category": "Benchmark", "keywords": [ "1", "920 Archived Human Programs; Public Subset: 909 Packages", "Opponent Selection", "Replay Analysis", "Executable Policy Improvement", "12 Adversarial Games", "Historical Human Opponents", "Budgeted Heuristic Learning" ], "keywords_by_category": { "Dataset": "1,920 Archived Human Programs; Public Subset: 909 Packages", "Harness": "Opponent Selection, Replay Analysis, Executable Policy Improvement", "Benchmark": "12 Adversarial Games, Historical Human Opponents, Budgeted Heuristic Learning" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2610.12341" }, { "label": "Project", "url": "https://aaarena.net/" }, { "label": "Code", "url": "https://github.com/THU-CST-SAST/AAArena" } ], "summary": "Twelve adversarial games and 1,920 archived human programs test coding agents that choose opponents, analyze replays, and improve executable game policies within fixed budgets while keeping model weights frozen.", "release_status": "Evaluation code and 909 historical human-player packages are public; the complete 1,920-program frozen opponent pool remains on the authenticated evaluation server.", "notes": "The arXiv abstract has an unexpanded configuration-count macro, so do not report a count of evaluated configurations. The 1,920-program archive is not fully downloadable; distinguish it from the 909 public packages.", "source": "update", "focus": "player", "source_rows": [ { "section": "Harness", "line": 178, "date": "2026-10", "title": "Can AI Agents Learn Their Way to the Top? Evaluating Heuristic Learning in a Long-Running Game Agent Competition", "venue": "arXiv", "keywords": "Opponent Selection, Replay Analysis, Executable Policy Improvement", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2610.12341" }, { "label": "Project", "url": "https://aaarena.net/" }, { "label": "Code", "url": "https://github.com/THU-CST-SAST/AAArena" } ] }, { "section": "Dataset", "line": 233, "date": "2026-10", "title": "Can AI Agents Learn Their Way to the Top? Evaluating Heuristic Learning in a Long-Running Game Agent Competition", "venue": "arXiv", "keywords": "1,920 Archived Human Programs; Public Subset: 909 Packages", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2610.12341" }, { "label": "Project", "url": "https://aaarena.net/" }, { "label": "Code", "url": "https://github.com/THU-CST-SAST/AAArena" } ] }, { "section": "Benchmark", "line": 274, "date": "2026-10", "title": "Can AI Agents Learn Their Way to the Top? Evaluating Heuristic Learning in a Long-Running Game Agent Competition", "venue": "arXiv", "keywords": "12 Adversarial Games, Historical Human Opponents, Budgeted Heuristic Learning", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2610.12341" }, { "label": "Project", "url": "https://aaarena.net/" }, { "label": "Code", "url": "https://github.com/THU-CST-SAST/AAArena" } ] } ], "sources": [ "https://arxiv.org/abs/2610.12341", "https://github.com/THU-CST-SAST/AAArena", "https://aaarena.net/" ], "search_text": "can ai agents learn their way to the top? evaluating heuristic learning in a long-running game agent competition aaarena / adversarial heuristic learning 2026-10-08 arxiv dataset harness benchmark 1 920 archived human programs; public subset: 909 packages opponent selection replay analysis executable policy improvement 12 adversarial games historical human opponents budgeted heuristic learning twelve adversarial games and 1,920 archived human programs test coding agents that choose opponents, analyze replays, and improve executable game policies within fixed budgets while keeping model weights frozen. the arxiv abstract has an unexpanded configuration-count macro, so do not report a count of evaluated configurations. the 1,920-program archive is not fully downloadable; distinguish it from the 909 public packages. player" }, { "id": "work-8dbf6af4ac59", "name": "Scaling in Games / GameScaling", "title": "Scaling in Games: Continued Pre-Training for Embodied Agents in Diverse Virtual Worlds", "date": "2026-10", "date_precision": "month", "date_basis": "First project-page commit 2026-10-09 17:01:38 UTC (Oct 10 Asia/Shanghai); initial code release commit 2026-10-10 16:05:19 UTC. No exact paper-publication day asserted.", "venue": "Technical report", "categories": [ "Dataset", "Model" ], "primary_category": "Model", "keywords": [ "Data-Scaling Study", "Volume vs. Game Diversity; Training Corpus Not Released", "Continued Pre-Training", "120 Configurations", "Cross-Game Transfer", "8 Released Policies" ], "keywords_by_category": { "Dataset": "Data-Scaling Study, Volume vs. Game Diversity; Training Corpus Not Released", "Model": "Continued Pre-Training, 120 Configurations, Cross-Game Transfer, 8 Released Policies" }, "links": [ { "label": "Paper", "url": "https://gamescaling.github.io/assets/paper.pdf" }, { "label": "Project", "url": "https://gamescaling.github.io/" }, { "label": "Code", "url": "https://github.com/THUSI-Lab/GameScaling" }, { "label": "Models", "url": "https://huggingface.co/THUSI-Lab/GameScaling" } ], "summary": "A controlled study of continued pre-training over 120 configurations varying gameplay volume, game diversity, and model size (0.8B–27B). GameScaling releases eight keyboard-and-mouse policies and evaluates transfer from in-corpus games to related and distant unseen worlds.", "release_status": "Official paper PDF, inference client, and eight checkpoints released; training code and training corpus not released; arXiv marked coming soon in current README.", "notes": "Use project/PDF citation, not a fabricated arXiv ID. Latest repository citation (Oct 20) uses journal={URL: https://gamescaling.github.io/}; older web-search cache calls it an arXiv preprint. All released checkpoints use 1,600 hours; the study spans 50–1,600 hours and five mixtures. Data-volume/diversity relevance warrants a secondary Dataset tag but no claim of a public corpus.", "source": "update", "focus": "player", "source_rows": [ { "section": "Model", "line": 94, "date": "2026-10", "title": "Scaling in Games: Continued Pre-Training for Embodied Agents in Diverse Virtual Worlds", "venue": "Technical report", "keywords": "Continued Pre-Training, 120 Configurations, Cross-Game Transfer, 8 Released Policies", "links": [ { "label": "Paper", "url": "https://gamescaling.github.io/assets/paper.pdf" }, { "label": "Project", "url": "https://gamescaling.github.io/" }, { "label": "Code", "url": "https://github.com/THUSI-Lab/GameScaling" }, { "label": "Models", "url": "https://huggingface.co/THUSI-Lab/GameScaling" } ] }, { "section": "Dataset", "line": 234, "date": "2026-10", "title": "Scaling in Games: Continued Pre-Training for Embodied Agents in Diverse Virtual Worlds", "venue": "Technical report", "keywords": "Data-Scaling Study, Volume vs. Game Diversity; Training Corpus Not Released", "links": [ { "label": "Paper", "url": "https://gamescaling.github.io/assets/paper.pdf" }, { "label": "Project", "url": "https://gamescaling.github.io/" }, { "label": "Code", "url": "https://github.com/THUSI-Lab/GameScaling" } ] } ], "sources": [ "https://gamescaling.github.io/assets/paper.pdf", "https://github.com/THUSI-Lab/GameScaling", "https://github.com/gamescaling/gamescaling.github.io/commit/08837b81e6a2dc1f71e898840cbdb52bc31546af", "https://github.com/THUSI-Lab/GameScaling/commit/43794d63fd13dd511d8378261e2418347ee5acc1", "https://github.com/gamescaling/gamescaling.github.io/blob/main/abstract.html" ], "search_text": "scaling in games: continued pre-training for embodied agents in diverse virtual worlds scaling in games / gamescaling 2026-10 technical report dataset model data-scaling study volume vs. game diversity; training corpus not released continued pre-training 120 configurations cross-game transfer 8 released policies a controlled study of continued pre-training over 120 configurations varying gameplay volume, game diversity, and model size (0.8b–27b). gamescaling releases eight keyboard-and-mouse policies and evaluates transfer from in-corpus games to related and distant unseen worlds. use project/pdf citation, not a fabricated arxiv id. latest repository citation (oct 20) uses journal={url: https://gamescaling.github.io/}; older web-search cache calls it an arxiv preprint. all released checkpoints use 1,600 hours; the study spans 50–1,600 hours and five mixtures. data-volume/diversity relevance warrants a secondary dataset tag but no claim of a public corpus. player" }, { "id": "arxiv-2609.25652", "name": "GameDirector", "title": "GameDirector: Decoupling Gameplay Logic from Rendering for Player-Configurable Game World Models", "date": "2026-09-22", "date_precision": "day", "date_basis": "arXiv v1", "venue": "arXiv", "categories": [ "Dataset", "Model", "Harness" ], "primary_category": "Harness", "keywords": [ "51", "758 Clips", "3 Games", "Video/Control/NPC/State Alignment", "State-Conditioned Game WM", "NPC Control", "Neural Rendering", "Explicit State & Rules", "NPC Planning", "Decoupled Rendering" ], "keywords_by_category": { "Dataset": "51,758 Clips, 3 Games, Video/Control/NPC/State Alignment", "Model": "State-Conditioned Game WM, NPC Control, Neural Rendering", "Harness": "Explicit State & Rules, NPC Planning, Decoupled Rendering" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2609.25652" }, { "label": "Project", "url": "https://jimntu.github.io/gamedirector/" }, { "label": "Code", "url": "https://github.com/Jimntu/GameDirector_code" }, { "label": "Models", "url": "https://huggingface.co/onepiece1999/GameDirector" }, { "label": "Data", "url": "https://huggingface.co/datasets/onepiece1999/GameDirector-Training-Data" } ], "summary": "Separates explicit game-state updates, rule enforcement, and NPC planning from video rendering to support configurable and mechanically consistent gameplay.", "release_status": "Code, model and dataset repositories available.", "notes": "Dataset has 51,758 clips across three games and 13 characters, including 300 evaluation clips. Uses a 2B language planner and trained visual components with a Wan2.2-TI2V-5B renderer. Related NPC/world-generation infrastructure, not a cross-game native-action player policy.", "source": "update", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 95, "date": "2026-09", "title": "GameDirector: Decoupling Gameplay Logic from Rendering for Player-Configurable Game World Models", "venue": "arXiv", "keywords": "State-Conditioned Game WM, NPC Control, Neural Rendering", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2609.25652" }, { "label": "Project", "url": "https://jimntu.github.io/gamedirector/" }, { "label": "Code", "url": "https://github.com/Jimntu/GameDirector_code" }, { "label": "Models", "url": "https://huggingface.co/onepiece1999/GameDirector" } ] }, { "section": "Harness", "line": 179, "date": "2026-09", "title": "GameDirector: Decoupling Gameplay Logic from Rendering for Player-Configurable Game World Models", "venue": "arXiv", "keywords": "Explicit State & Rules, NPC Planning, Decoupled Rendering", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2609.25652" }, { "label": "Project", "url": "https://jimntu.github.io/gamedirector/" }, { "label": "Code", "url": "https://github.com/Jimntu/GameDirector_code" } ] }, { "section": "Dataset", "line": 235, "date": "2026-09", "title": "GameDirector: Decoupling Gameplay Logic from Rendering for Player-Configurable Game World Models", "venue": "arXiv", "keywords": "51,758 Clips, 3 Games, Video/Control/NPC/State Alignment", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2609.25652" }, { "label": "Project", "url": "https://jimntu.github.io/gamedirector/" }, { "label": "Code", "url": "https://github.com/Jimntu/GameDirector_code" }, { "label": "Data", "url": "https://huggingface.co/datasets/onepiece1999/GameDirector-Training-Data" } ] } ], "sources": [ "https://arxiv.org/abs/2609.25652", "https://jimntu.github.io/gamedirector/", "https://github.com/Jimntu/GameDirector_code", "https://huggingface.co/datasets/onepiece1999/GameDirector-Training-Data" ], "search_text": "gamedirector: decoupling gameplay logic from rendering for player-configurable game world models gamedirector 2026-09-22 arxiv dataset model harness 51 758 clips 3 games video/control/npc/state alignment state-conditioned game wm npc control neural rendering explicit state & rules npc planning decoupled rendering separates explicit game-state updates, rule enforcement, and npc planning from video rendering to support configurable and mechanically consistent gameplay. dataset has 51,758 clips across three games and 13 characters, including 300 evaluation clips. uses a 2b language planner and trained visual components with a wan2.2-ti2v-5b renderer. related npc/world-generation infrastructure, not a cross-game native-action player policy. supporting environment" }, { "id": "arxiv-2609.25001", "name": "GameHorizon Suite", "title": "GameHorizon Suite: Multi-Horizon Data and Evaluation in Gameplay", "date": "2026-09-21", "date_precision": "day", "date_basis": "arXiv v1 submission 2026-09-21 17:59:33 UTC; v2 2026-10-12; project announced Sep 22", "venue": "arXiv", "categories": [ "Dataset", "Benchmark" ], "primary_category": "Dataset", "keywords": [ "GameHorizon-Data", "5", "000 Hours", "21 Games", "Multi-Horizon Instructions; Release Pending", "GameHorizon-Bench", "Offline QA + Online Tasks", "Multi-Horizon Evaluation; Release Pending" ], "keywords_by_category": { "Dataset": "GameHorizon-Data, 5,000 Hours, 21 Games, Multi-Horizon Instructions; Release Pending", "Benchmark": "GameHorizon-Bench, Offline QA + Online Tasks, Multi-Horizon Evaluation; Release Pending" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2609.25001" }, { "label": "Project", "url": "https://gamehorizon-suite.github.io/" }, { "label": "Repository", "url": "https://github.com/TencentARC/GameHorizon" } ], "summary": "A suite combining an automatic multi-horizon instruction annotator, 5,000 hours of expert gameplay across 21 games, and offline plus stepwise online evaluation. GameHorizon-Bench contains 5,000 offline questions and 20 online tasks with 62 subtasks.", "release_status": "Paper, project page, and leaderboard public. Annotator, benchmark code/environments, and dataset still coming soon as of 2026-10-21; official README says progressive releases start around 2026-10-25.", "notes": "Can use two rows, GameHorizon-Data under Dataset and GameHorizon-Bench under Benchmark, linked to the same paper and named suite. Do not label the placeholder repository as released code or invent a dataset URL. Annotator is data construction tooling, not the survey's deployed-agent Harness.", "source": "update", "focus": "player", "source_rows": [ { "section": "Dataset", "line": 236, "date": "2026-09", "title": "GameHorizon Suite: Multi-Horizon Data and Evaluation in Gameplay", "venue": "arXiv", "keywords": "GameHorizon-Data, 5,000 Hours, 21 Games, Multi-Horizon Instructions; Release Pending", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2609.25001" }, { "label": "Project", "url": "https://gamehorizon-suite.github.io/" }, { "label": "Repository", "url": "https://github.com/TencentARC/GameHorizon" } ] }, { "section": "Benchmark", "line": 275, "date": "2026-09", "title": "GameHorizon Suite: Multi-Horizon Data and Evaluation in Gameplay", "venue": "arXiv", "keywords": "GameHorizon-Bench, Offline QA + Online Tasks, Multi-Horizon Evaluation; Release Pending", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2609.25001" }, { "label": "Project", "url": "https://gamehorizon-suite.github.io/" }, { "label": "Repository", "url": "https://github.com/TencentARC/GameHorizon" } ] } ], "sources": [ "https://arxiv.org/abs/2609.25001", "https://github.com/TencentARC/GameHorizon/blob/main/README.md" ], "search_text": "gamehorizon suite: multi-horizon data and evaluation in gameplay gamehorizon suite 2026-09-21 arxiv dataset benchmark gamehorizon-data 5 000 hours 21 games multi-horizon instructions; release pending gamehorizon-bench offline qa + online tasks multi-horizon evaluation; release pending a suite combining an automatic multi-horizon instruction annotator, 5,000 hours of expert gameplay across 21 games, and offline plus stepwise online evaluation. gamehorizon-bench contains 5,000 offline questions and 20 online tasks with 62 subtasks. can use two rows, gamehorizon-data under dataset and gamehorizon-bench under benchmark, linked to the same paper and named suite. do not label the placeholder repository as released code or invent a dataset url. annotator is data construction tooling, not the survey's deployed-agent harness. player" }, { "id": "work-d6c4251d9567", "name": "Agora-2 (Odyssey)", "title": "Agora-2: An Action-Conditioned World Model for Real-Time Multiplayer Environments", "date": "2026-09-21", "date_precision": "day", "date_basis": "Official research announcement date", "venue": "Technical report", "categories": [ "Model" ], "primary_category": "Model", "keywords": [ "Action-Conditioned Multiplayer WM", "Shared State", "Diablo II" ], "keywords_by_category": { "Model": "Action-Conditioned Multiplayer WM, Shared State, Diablo II" }, "links": [ { "label": "Paper", "url": "https://agora-2.odyssey.systems/agora-2.pdf" }, { "label": "Project", "url": "https://odyssey.systems/introducing-agora-2" } ], "summary": "A shared-state multiplayer world model developed with Diablo II captures, separating learned simulation from per-player rendering and supporting four human-controlled entities alongside sixteen autonomous entities.", "release_status": "code: No author-linked public code release verified; data: No standalone downloadable dataset release verified in this review; weights: No downloadable model weights verified in this review", "notes": "Date is the official announcement release date; technical report PDF itself has no date on its first page. The report clarifies the advertised 20 humans/agents means four human-controlled plus sixteen autonomous entities in its deployed setup. Learned state predictions coexist with server integration/lifecycle rules; do not describe every gameplay rule as learned.", "source": "update", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 96, "date": "2026-09", "title": "Agora-2: An Action-Conditioned World Model for Real-Time Multiplayer Environments", "venue": "Technical report", "keywords": "Action-Conditioned Multiplayer WM, Shared State, Diablo II", "links": [ { "label": "Paper", "url": "https://agora-2.odyssey.systems/agora-2.pdf" }, { "label": "Project", "url": "https://odyssey.systems/introducing-agora-2" } ] } ], "sources": [ "https://agora-2.odyssey.systems/agora-2.pdf", "https://odyssey.systems/introducing-agora-2" ], "search_text": "agora-2: an action-conditioned world model for real-time multiplayer environments agora-2 (odyssey) 2026-09-21 technical report model action-conditioned multiplayer wm shared state diablo ii a shared-state multiplayer world model developed with diablo ii captures, separating learned simulation from per-player rendering and supporting four human-controlled entities alongside sixteen autonomous entities. date is the official announcement release date; technical report pdf itself has no date on its first page. the report clarifies the advertised 20 humans/agents means four human-controlled plus sixteen autonomous entities in its deployed setup. learned state predictions coexist with server integration/lifecycle rules; do not describe every gameplay rule as learned. supporting environment" }, { "id": "work-702c67caa01b", "name": "Odyssey-3", "title": "Introducing Odyssey-3: A General-Purpose Physical Intelligence", "date": "2026-09-15", "date_precision": "day", "date_basis": "Official research announcement date", "venue": "Research announcement", "categories": [ "Model" ], "primary_category": "Model", "keywords": [ "Frozen WM Backbone", "Learned Gameplay Action Decoders", "GTA V; Research Preview" ], "keywords_by_category": { "Model": "Frozen WM Backbone, Learned Gameplay Action Decoders, GTA V; Research Preview" }, "links": [ { "label": "Project", "url": "https://odyssey.systems/introducing-odyssey-3" } ], "summary": "An autoregressive diffusion world-model backbone used to train gameplay action decoders while keeping the backbone frozen, with GTA V demonstrations and qualitative cross-title movement transfer to RDR2 and Sleeping Dogs.", "release_status": "code: No author-linked public code release verified; data: No standalone downloadable dataset release verified in this review; weights: No downloadable model weights verified in this review", "notes": "Official announcement reports qualitative gameplay examples, not a matched benchmark establishing cross-game mastery. No separate Odyssey-3 paper or released gameplay dataset/model weights verified.", "source": "update", "focus": "player", "source_rows": [ { "section": "Model", "line": 97, "date": "2026-09", "title": "Introducing Odyssey-3: A General-Purpose Physical Intelligence", "venue": "Research announcement", "keywords": "Frozen WM Backbone, Learned Gameplay Action Decoders, GTA V; Research Preview", "links": [ { "label": "Project", "url": "https://odyssey.systems/introducing-odyssey-3" } ] } ], "sources": [ "https://odyssey.systems/introducing-odyssey-3" ], "search_text": "introducing odyssey-3: a general-purpose physical intelligence odyssey-3 2026-09-15 research announcement model frozen wm backbone learned gameplay action decoders gta v; research preview an autoregressive diffusion world-model backbone used to train gameplay action decoders while keeping the backbone frozen, with gta v demonstrations and qualitative cross-title movement transfer to rdr2 and sleeping dogs. official announcement reports qualitative gameplay examples, not a matched benchmark establishing cross-game mastery. no separate odyssey-3 paper or released gameplay dataset/model weights verified. player" }, { "id": "arxiv-2608.26200", "name": "GameWAM", "title": "GameWAM: A World Action Model for Video Games", "date": "2026-08-25", "date_precision": "day", "date_basis": "arXiv v1; dataset exact initial upload dates not separately established", "venue": "arXiv", "categories": [ "Dataset", "Model" ], "primary_category": "Model", "keywords": [ "Minecraft & ViZDoom", "Video/Action Trajectories", "GUI Tasks", "WAM", "Joint Video/Action Training", "Action-Only Inference" ], "keywords_by_category": { "Dataset": "Minecraft & ViZDoom, Video/Action Trajectories, GUI Tasks", "Model": "WAM, Joint Video/Action Training, Action-Only Inference" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2608.26200" }, { "label": "Project", "url": "https://yunncheng.github.io/GameWAM/" }, { "label": "Code", "url": "https://github.com/yunncheng/GameWAM" }, { "label": "Models", "url": "https://huggingface.co/Yunncheng/gamewam-models" }, { "label": "Data 1", "url": "https://huggingface.co/datasets/Yunncheng/gamewam-minecraft" }, { "label": "Data 2", "url": "https://huggingface.co/datasets/Yunncheng/gamewam-vizdoom" } ], "summary": "Jointly learns future-video prediction and native keyboard-and-mouse actions, then uses action-only closed-loop inference for gameplay and in-game GUI tasks.", "release_status": "Code, model checkpoints, and Minecraft/ViZDoom training datasets available.", "notes": "Project explicitly clarifies that future-video prediction is training supervision and future video is not denoised online. Do not imply online imagined-video planning. Minecraft data includes VPT-derived streams and newly scripted GUI trajectories; it is not all newly collected data. ViZDoom provides four scenarios and one training split.", "source": "update", "focus": "player", "source_rows": [ { "section": "Model", "line": 98, "date": "2026-08", "title": "GameWAM: A World Action Model for Video Games", "venue": "arXiv", "keywords": "WAM, Joint Video/Action Training, Action-Only Inference", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2608.26200" }, { "label": "Project", "url": "https://yunncheng.github.io/GameWAM/" }, { "label": "Code", "url": "https://github.com/yunncheng/GameWAM" }, { "label": "Models", "url": "https://huggingface.co/Yunncheng/gamewam-models" } ] }, { "section": "Dataset", "line": 237, "date": "2026-08", "title": "GameWAM: A World Action Model for Video Games", "venue": "arXiv", "keywords": "Minecraft & ViZDoom, Video/Action Trajectories, GUI Tasks", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2608.26200" }, { "label": "Project", "url": "https://yunncheng.github.io/GameWAM/" }, { "label": "Code", "url": "https://github.com/yunncheng/GameWAM" }, { "label": "Data 1", "url": "https://huggingface.co/datasets/Yunncheng/gamewam-minecraft" }, { "label": "Data 2", "url": "https://huggingface.co/datasets/Yunncheng/gamewam-vizdoom" } ] } ], "sources": [ "https://arxiv.org/abs/2608.26200", "https://yunncheng.github.io/GameWAM/", "https://github.com/yunncheng/GameWAM", "https://huggingface.co/datasets/Yunncheng/gamewam-minecraft", "https://huggingface.co/datasets/Yunncheng/gamewam-vizdoom" ], "search_text": "gamewam: a world action model for video games gamewam 2026-08-25 arxiv dataset model minecraft & vizdoom video/action trajectories gui tasks wam joint video/action training action-only inference jointly learns future-video prediction and native keyboard-and-mouse actions, then uses action-only closed-loop inference for gameplay and in-game gui tasks. project explicitly clarifies that future-video prediction is training supervision and future video is not denoised online. do not imply online imagined-video planning. minecraft data includes vpt-derived streams and newly scripted gui trajectories; it is not all newly collected data. vizdoom provides four scenarios and one training split. player" }, { "id": "arxiv-2608.21439", "name": "WorldMind", "title": "WorldMind: Decoupled Game World Model for State-Aware NPC Behavior", "date": "2026-08-18", "date_precision": "day", "date_basis": "arXiv v1", "venue": "arXiv", "categories": [ "Dataset", "Model", "Harness" ], "primary_category": "Harness", "keywords": [ "BOSS-140K", "144", "631 Clips", "14 Bosses", "3 Games; Release Pending", "Decoupled Game WM", "State-Aware NPCs; Weights Pending", "State Reconstruction", "LLM NPC Planning", "Closed-Loop World Simulation; Code Pending" ], "keywords_by_category": { "Dataset": "BOSS-140K, 144,631 Clips, 14 Bosses, 3 Games; Release Pending", "Model": "Decoupled Game WM, State-Aware NPCs; Weights Pending", "Harness": "State Reconstruction, LLM NPC Planning, Closed-Loop World Simulation; Code Pending" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2608.21439" }, { "label": "Project", "url": "https://teawhite.cn/WorldMind/" }, { "label": "Repository", "url": "https://github.com/TeaWhiteBro/WorldMind" } ], "summary": "Connects visual state reconstruction, language-model NPC planning, action conditioning, and video generation in a closed loop for state-aware boss behavior.", "release_status": "Paper and project available; official repository says code and weights are being prepared, and project says data will be released.", "notes": "BOSS-140K is announced, not yet a downloadable released dataset. It contains 144,631 video clips and 279,721 non-death decision rows, spans 14 bosses/3 games, and synchronizes controls, NPC skills, and internal states. One game is anonymized as Game A; do not infer its name. Language-model decision layer is untuned; learned understanding/generation modules are separated by runtime orchestration.", "source": "update", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 99, "date": "2026-08", "title": "WorldMind: Decoupled Game World Model for State-Aware NPC Behavior", "venue": "arXiv", "keywords": "Decoupled Game WM, State-Aware NPCs; Weights Pending", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2608.21439" }, { "label": "Project", "url": "https://teawhite.cn/WorldMind/" }, { "label": "Repository", "url": "https://github.com/TeaWhiteBro/WorldMind" } ] }, { "section": "Harness", "line": 180, "date": "2026-08", "title": "WorldMind: Decoupled Game World Model for State-Aware NPC Behavior", "venue": "arXiv", "keywords": "State Reconstruction, LLM NPC Planning, Closed-Loop World Simulation; Code Pending", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2608.21439" }, { "label": "Project", "url": "https://teawhite.cn/WorldMind/" }, { "label": "Repository", "url": "https://github.com/TeaWhiteBro/WorldMind" } ] }, { "section": "Dataset", "line": 238, "date": "2026-08", "title": "WorldMind: Decoupled Game World Model for State-Aware NPC Behavior", "venue": "arXiv", "keywords": "BOSS-140K, 144,631 Clips, 14 Bosses, 3 Games; Release Pending", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2608.21439" }, { "label": "Project", "url": "https://teawhite.cn/WorldMind/" }, { "label": "Repository", "url": "https://github.com/TeaWhiteBro/WorldMind" } ] } ], "sources": [ "https://arxiv.org/abs/2608.21439", "https://teawhite.cn/WorldMind/", "https://github.com/TeaWhiteBro/WorldMind" ], "search_text": "worldmind: decoupled game world model for state-aware npc behavior worldmind 2026-08-18 arxiv dataset model harness boss-140k 144 631 clips 14 bosses 3 games; release pending decoupled game wm state-aware npcs; weights pending state reconstruction llm npc planning closed-loop world simulation; code pending connects visual state reconstruction, language-model npc planning, action conditioning, and video generation in a closed loop for state-aware boss behavior. boss-140k is announced, not yet a downloadable released dataset. it contains 144,631 video clips and 279,721 non-death decision rows, spans 14 bosses/3 games, and synchronizes controls, npc skills, and internal states. one game is anonymized as game a; do not infer its name. language-model decision layer is untuned; learned understanding/generation modules are separated by runtime orchestration. supporting environment" }, { "id": "arxiv-2607.26754", "name": "StatePlay", "title": "StatePlay: State-Aware Game World Models for Mechanics-Consistent Generation", "date": "2026-07-29", "date_precision": "day", "date_basis": "arXiv v1", "venue": "arXiv", "categories": [ "Dataset", "Model" ], "primary_category": "Model", "keywords": [ "Street Fighter", "Synchronized Video/Actions/States", "NPC Strategies", "State-Aware WM", "Joint Frame/State Prediction", "Mechanics Consistency" ], "keywords_by_category": { "Dataset": "Street Fighter, Synchronized Video/Actions/States, NPC Strategies", "Model": "State-Aware WM, Joint Frame/State Prediction, Mechanics Consistency" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2607.26754" }, { "label": "Project", "url": "https://jimntu.github.io/stateplay_page/" }, { "label": "Code", "url": "https://github.com/Jimntu/StatePlay" }, { "label": "Models", "url": "https://huggingface.co/onepiece1999/StatePlay" }, { "label": "Data", "url": "https://huggingface.co/datasets/onepiece1999/StatePlay-Dataset" } ], "summary": "Jointly predicts game frames and internal states through specialized transformer branches to improve consistency with health, skill, and termination mechanics.", "release_status": "Code, checkpoints, and dataset available.", "notes": "Action-conditioned game world generation rather than an autonomous action policy. Dataset pairs Street Fighter clips with actions, prompts, strategies, and states; visible dataset viewer reports 10k train rows.", "source": "update", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 100, "date": "2026-07", "title": "StatePlay: State-Aware Game World Models for Mechanics-Consistent Generation", "venue": "arXiv", "keywords": "State-Aware WM, Joint Frame/State Prediction, Mechanics Consistency", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2607.26754" }, { "label": "Project", "url": "https://jimntu.github.io/stateplay_page/" }, { "label": "Code", "url": "https://github.com/Jimntu/StatePlay" }, { "label": "Models", "url": "https://huggingface.co/onepiece1999/StatePlay" } ] }, { "section": "Dataset", "line": 239, "date": "2026-07", "title": "StatePlay: State-Aware Game World Models for Mechanics-Consistent Generation", "venue": "arXiv", "keywords": "Street Fighter, Synchronized Video/Actions/States, NPC Strategies", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2607.26754" }, { "label": "Project", "url": "https://jimntu.github.io/stateplay_page/" }, { "label": "Code", "url": "https://github.com/Jimntu/StatePlay" }, { "label": "Data", "url": "https://huggingface.co/datasets/onepiece1999/StatePlay-Dataset" } ] } ], "sources": [ "https://arxiv.org/abs/2607.26754", "https://jimntu.github.io/stateplay_page/", "https://github.com/Jimntu/StatePlay", "https://huggingface.co/datasets/onepiece1999/StatePlay-Dataset" ], "search_text": "stateplay: state-aware game world models for mechanics-consistent generation stateplay 2026-07-29 arxiv dataset model street fighter synchronized video/actions/states npc strategies state-aware wm joint frame/state prediction mechanics consistency jointly predicts game frames and internal states through specialized transformer branches to improve consistency with health, skill, and termination mechanics. action-conditioned game world generation rather than an autonomous action policy. dataset pairs street fighter clips with actions, prompts, strategies, and states; visible dataset viewer reports 10k train rows. supporting environment" }, { "id": "arxiv-2607.26712", "name": "ActSWM", "title": "ActSWM: Action-Sensitive World Models for Long-Horizon Planning in Open-World Games", "date": "2026-07-29", "date_precision": "day", "date_basis": "arXiv v1", "venue": "arXiv", "categories": [ "Model" ], "primary_category": "Model", "keywords": [ "Latent WM", "Action Sensitivity", "Long-Horizon Planning", "Minecraft" ], "keywords_by_category": { "Model": "Latent WM, Action Sensitivity, Long-Horizon Planning, Minecraft" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2607.26712" } ], "summary": "Trains action-sensitive latent dynamics to keep alternative futures distinguishable during long-horizon Minecraft planning and cross-game action recovery.", "release_status": "Paper verified; official code attribution not established.", "notes": "A matching public repository at https://github.com/KeithBlackmore/ActSWM- exists but cites Anonymous Authors / 2027 and is not linked from arXiv; omit code link unless authorship is established. Lightweight JEPA-style model, best treated as an adjacent planning/world-model contribution rather than a new LFM policy.", "source": "update", "focus": "player", "source_rows": [ { "section": "Model", "line": 101, "date": "2026-07", "title": "ActSWM: Action-Sensitive World Models for Long-Horizon Planning in Open-World Games", "venue": "arXiv", "keywords": "Latent WM, Action Sensitivity, Long-Horizon Planning, Minecraft", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2607.26712" } ] } ], "sources": [ "https://arxiv.org/abs/2607.26712", "https://arxiv.org/html/2607.26712v2" ], "search_text": "actswm: action-sensitive world models for long-horizon planning in open-world games actswm 2026-07-29 arxiv model latent wm action sensitivity long-horizon planning minecraft trains action-sensitive latent dynamics to keep alternative futures distinguishable during long-horizon minecraft planning and cross-game action recovery. a matching public repository at https://github.com/keithblackmore/actswm- exists but cites anonymous authors / 2027 and is not linked from arxiv; omit code link unless authorship is established. lightweight jepa-style model, best treated as an adjacent planning/world-model contribution rather than a new lfm policy. player" }, { "id": "arxiv-2609.18996", "name": "Compiled Agency / Gauntlet", "title": "Compiled Agency: Coding Agents as Game AI Researchers -- from a Roguelike to StarCraft II and Civilization", "date": "2026-07-17", "date_precision": "day", "date_basis": "First arXiv submission date", "venue": "arXiv", "categories": [ "Harness", "Benchmark" ], "primary_category": "Benchmark", "keywords": [ "Coding Agents", "Develop–Freeze–Evaluate", "Standalone Controllers", "Gauntlet", "Roguelike/StarCraft II/Civilization", "Held-Out Evaluation", "Zero LLM Calls During Play" ], "keywords_by_category": { "Harness": "Coding Agents, Develop–Freeze–Evaluate, Standalone Controllers", "Benchmark": "Gauntlet, Roguelike/StarCraft II/Civilization, Held-Out Evaluation, Zero LLM Calls During Play" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2609.18996" } ], "summary": "A develop-freeze-evaluate protocol in which coding agents use game interaction to build standalone controllers, then evaluate frozen programs on held-out instances with zero model calls during play.", "release_status": "code: No author-linked public code release verified; data: No standalone downloadable dataset release verified in this review; weights: No downloadable model weights verified in this review", "notes": "Use the current v2 title above, not the older search-index title 'Frontier General-Purpose Coding Agents Build Winning Game Players from Bare Interaction'. ArXiv header/history explicitly state July 17 despite the 2609 identifier. No author-linked project/code URL was verified.", "source": "update", "focus": "player", "source_rows": [ { "section": "Harness", "line": 181, "date": "2026-07", "title": "Compiled Agency: Coding Agents as Game AI Researchers -- from a Roguelike to StarCraft II and Civilization", "venue": "arXiv", "keywords": "Coding Agents, Develop–Freeze–Evaluate, Standalone Controllers", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2609.18996" } ] }, { "section": "Benchmark", "line": 276, "date": "2026-07", "title": "Compiled Agency: Coding Agents as Game AI Researchers -- from a Roguelike to StarCraft II and Civilization", "venue": "arXiv", "keywords": "Gauntlet, Roguelike/StarCraft II/Civilization, Held-Out Evaluation, Zero LLM Calls During Play", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2609.18996" } ] } ], "sources": [ "https://arxiv.org/abs/2609.18996", "https://arxiv.org/html/2609.18996v2" ], "search_text": "compiled agency: coding agents as game ai researchers -- from a roguelike to starcraft ii and civilization compiled agency / gauntlet 2026-07-17 arxiv harness benchmark coding agents develop–freeze–evaluate standalone controllers gauntlet roguelike/starcraft ii/civilization held-out evaluation zero llm calls during play a develop-freeze-evaluate protocol in which coding agents use game interaction to build standalone controllers, then evaluate frozen programs on held-out instances with zero model calls during play. use the current v2 title above, not the older search-index title 'frontier general-purpose coding agents build winning game players from bare interaction'. arxiv header/history explicitly state july 17 despite the 2609 identifier. no author-linked project/code url was verified. player" }, { "id": "arxiv-2607.02255", "name": "AgenticSTS", "title": "AgenticSTS: A Bounded-Memory Testbed for Long-Horizon LLM Agents", "date": "2026-07-02", "date_precision": "day", "date_basis": "arXiv v1 submission 2026-07-02 14:44:32 UTC", "venue": "arXiv", "categories": [ "Dataset", "Harness", "Benchmark" ], "primary_category": "Harness", "keywords": [ "298 Completed Games", "Condition-Tagged Trajectories", "Frozen Memory/Skill Stores", "Bounded Typed Memory", "Episodic Retrieval", "Strategic Skills", "Slay the Spire 2", "Long-Horizon Memory Ablations", "State/Action API" ], "keywords_by_category": { "Dataset": "298 Completed Games, Condition-Tagged Trajectories, Frozen Memory/Skill Stores", "Harness": "Bounded Typed Memory, Episodic Retrieval, Strategic Skills, Slay the Spire 2", "Benchmark": "Long-Horizon Memory Ablations, State/Action API, Slay the Spire 2" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2607.02255" }, { "label": "Project", "url": "https://alayalab.github.io/AgenticSTS/" }, { "label": "Code", "url": "https://github.com/AlayaLab/AgenticSTS" }, { "label": "Data", "url": "https://huggingface.co/datasets/AlayaLab/AgenticSTS-trajectories" } ], "summary": "A Slay the Spire 2 agent/testbed that rebuilds bounded decision prompts from typed rules, episodic memories, and triggered strategic skills. It releases condition-tagged trajectories, frozen memory/skill stores, prompt records, and analysis scripts for ablations of long-horizon memory.", "release_status": "Agent, game-mod interface, analysis scripts, and trajectories public.", "notes": "Paper analyzes 298 completed games; HF has 312 analysis records (298 completed plus 14 capped), and repo history has 385 audit rows. Describe as 298 completed trajectories to avoid count confusion. Interface exposes state/action APIs via game mod, not raw-screen gameplay. Small fixed-A0 sample means avoid unqualified superiority/performance claims.", "source": "update", "focus": "player", "source_rows": [ { "section": "Harness", "line": 182, "date": "2026-07", "title": "AgenticSTS: A Bounded-Memory Testbed for Long-Horizon LLM Agents", "venue": "arXiv", "keywords": "Bounded Typed Memory, Episodic Retrieval, Strategic Skills, Slay the Spire 2", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2607.02255" }, { "label": "Project", "url": "https://alayalab.github.io/AgenticSTS/" }, { "label": "Code", "url": "https://github.com/AlayaLab/AgenticSTS" } ] }, { "section": "Dataset", "line": 240, "date": "2026-07", "title": "AgenticSTS: A Bounded-Memory Testbed for Long-Horizon LLM Agents", "venue": "arXiv", "keywords": "298 Completed Games, Condition-Tagged Trajectories, Frozen Memory/Skill Stores", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2607.02255" }, { "label": "Project", "url": "https://alayalab.github.io/AgenticSTS/" }, { "label": "Code", "url": "https://github.com/AlayaLab/AgenticSTS" }, { "label": "Data", "url": "https://huggingface.co/datasets/AlayaLab/AgenticSTS-trajectories" } ] }, { "section": "Benchmark", "line": 277, "date": "2026-07", "title": "AgenticSTS: A Bounded-Memory Testbed for Long-Horizon LLM Agents", "venue": "arXiv", "keywords": "Long-Horizon Memory Ablations, State/Action API, Slay the Spire 2", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2607.02255" }, { "label": "Project", "url": "https://alayalab.github.io/AgenticSTS/" }, { "label": "Code", "url": "https://github.com/AlayaLab/AgenticSTS" } ] } ], "sources": [ "https://arxiv.org/abs/2607.02255v1", "https://github.com/AlayaLab/AgenticSTS", "https://huggingface.co/datasets/AlayaLab/AgenticSTS-trajectories" ], "search_text": "agenticsts: a bounded-memory testbed for long-horizon llm agents agenticsts 2026-07-02 arxiv dataset harness benchmark 298 completed games condition-tagged trajectories frozen memory/skill stores bounded typed memory episodic retrieval strategic skills slay the spire 2 long-horizon memory ablations state/action api a slay the spire 2 agent/testbed that rebuilds bounded decision prompts from typed rules, episodic memories, and triggered strategic skills. it releases condition-tagged trajectories, frozen memory/skill stores, prompt records, and analysis scripts for ablations of long-horizon memory. paper analyzes 298 completed games; hf has 312 analysis records (298 completed plus 14 capped), and repo history has 385 audit rows. describe as 298 completed trajectories to avoid count confusion. interface exposes state/action apis via game mod, not raw-screen gameplay. small fixed-a0 sample means avoid unqualified superiority/performance claims. player" }, { "id": "arxiv-2607.00642", "name": "Coachable agents", "title": "Coachable agents for interactive gameplay", "date": "2026-07-01", "date_precision": "day", "date_basis": "First arXiv submission date; v2 was submitted on 2026-08-10", "venue": "arXiv", "categories": [ "Model" ], "primary_category": "Model", "keywords": [ "Related Non-LFM RL", "Style-Conditioned Policies", "Runtime Coaching", "HFW / GT7 / Humanoid" ], "keywords_by_category": { "Model": "Related Non-LFM RL, Style-Conditioned Policies, Runtime Coaching, HFW / GT7 / Humanoid" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2607.00642" }, { "label": "Code (humanoid)", "url": "https://github.com/SonyResearch/coachable_agents" } ], "summary": "Studies reinforcement-learning policies whose behavior can be changed at runtime through structured style controls while retaining a fixed core task. Domain-specific experiments cover Horizon Forbidden West, Gran Turismo 7, and a humanoid control environment.", "release_status": "Humanoid training/evaluation code and an example policy are public. The linked repository does not release the Horizon Forbidden West or Gran Turismo 7 environments and controllers.", "notes": "Included as related non-foundation-model RL research, not an LFM-based generalist agent. Coaching uses structured style parameters rather than natural-language instructions. Style conditioning is part of the learned policy; no separate Dataset, Harness, or Benchmark contribution is claimed.", "source": "update", "focus": "player", "source_rows": [ { "section": "Model", "line": 102, "date": "2026-07", "title": "Coachable agents for interactive gameplay", "venue": "arXiv", "keywords": "Related Non-LFM RL, Style-Conditioned Policies, Runtime Coaching, HFW / GT7 / Humanoid", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2607.00642" }, { "label": "Code (humanoid)", "url": "https://github.com/SonyResearch/coachable_agents" } ] } ], "sources": [ "https://arxiv.org/abs/2607.00642", "https://arxiv.org/html/2607.00642v2", "https://github.com/SonyResearch/coachable_agents" ], "search_text": "coachable agents for interactive gameplay coachable agents 2026-07-01 arxiv model related non-lfm rl style-conditioned policies runtime coaching hfw / gt7 / humanoid studies reinforcement-learning policies whose behavior can be changed at runtime through structured style controls while retaining a fixed core task. domain-specific experiments cover horizon forbidden west, gran turismo 7, and a humanoid control environment. included as related non-foundation-model rl research, not an lfm-based generalist agent. coaching uses structured style parameters rather than natural-language instructions. style conditioning is part of the learned policy; no separate dataset, harness, or benchmark contribution is claimed. player" }, { "id": "arxiv-2606.17838", "name": "Environment-Grounded APO", "title": "Environment-Grounded Automated Prompt Optimization for LLM Game Agents", "date": "2026-06-16", "date_precision": "day", "date_basis": "First arXiv submission date", "venue": "arXiv", "categories": [ "Harness" ], "primary_category": "Harness", "keywords": [ "Prompt Optimization", "Perception/Action Decomposition", "Evolutionary Search", "BabyAI" ], "keywords_by_category": { "Harness": "Prompt Optimization, Perception/Action Decomposition, Evolutionary Search, BabyAI" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2606.17838" }, { "label": "Code", "url": "https://github.com/ReanFernandes/rapoa" } ], "summary": "Decomposes perception and action selection into separate agents, then optimizes their prompts using behavior analysis, evolutionary mutations, and environment returns on BALROG's BabyAI tasks.", "release_status": "code: Author-linked public repository verified; data: No standalone downloadable dataset release verified in this review; weights: No downloadable model weights verified in this review", "notes": "Code URL explicitly linked by the paper. All five BabyAI tasks, fixed model weights.", "source": "update", "focus": "player", "source_rows": [ { "section": "Harness", "line": 183, "date": "2026-06", "title": "Environment-Grounded Automated Prompt Optimization for LLM Game Agents", "venue": "arXiv", "keywords": "Prompt Optimization, Perception/Action Decomposition, Evolutionary Search, BabyAI", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2606.17838" }, { "label": "Code", "url": "https://github.com/ReanFernandes/rapoa" } ] } ], "sources": [ "https://arxiv.org/abs/2606.17838", "https://arxiv.org/html/2606.17838v1" ], "search_text": "environment-grounded automated prompt optimization for llm game agents environment-grounded apo 2026-06-16 arxiv harness prompt optimization perception/action decomposition evolutionary search babyai decomposes perception and action selection into separate agents, then optimizes their prompts using behavior analysis, evolutionary mutations, and environment returns on balrog's babyai tasks. code url explicitly linked by the paper. all five babyai tasks, fixed model weights. player" }, { "id": "arxiv-2606.09826", "name": "OmniGameArena / IDC", "title": "OmniGameArena: A Unified UE5 Benchmark for VLM Game Agents with Improvement Dynamics", "date": "2026-06-08", "date_precision": "day", "date_basis": "First arXiv submission date", "venue": "arXiv", "categories": [ "Harness", "Benchmark" ], "primary_category": "Benchmark", "keywords": [ "IDC", "Repeated-Round Reflection", "Skill-Prompt Refinement", "12 UE5 Games", "Solo/PvP/Coop", "Improvement Dynamics", "Held-Out Variants" ], "keywords_by_category": { "Harness": "IDC, Repeated-Round Reflection, Skill-Prompt Refinement", "Benchmark": "12 UE5 Games, Solo/PvP/Coop, Improvement Dynamics, Held-Out Variants" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2606.09826" }, { "label": "Project", "url": "https://mxlin043.github.io/OmniGameArena/" }, { "label": "Code", "url": "https://github.com/mxlin043/OmniGameArena" } ], "summary": "Twelve new UE5 games spanning solo, competitive, and cooperative play, with an Improvement Dynamics Curve harness that refines skill prompts over repeated rounds and tests transfer to held-out variants.", "release_status": "code: Author-linked public repository verified; data: Official project links UE5 environment downloads on Hugging Face and ModelScope; not a gameplay training dataset; weights: No downloadable model weights verified in this review", "notes": "Seven solo, three PvP, two cooperative games. README distinguishes paused decision-quality and latency-charged tracks. Current project page reports expanded reflection analysis relative to original arXiv abstract; avoid copying inconsistent agent counts.", "source": "update", "focus": "player", "source_rows": [ { "section": "Harness", "line": 184, "date": "2026-06", "title": "OmniGameArena: A Unified UE5 Benchmark for VLM Game Agents with Improvement Dynamics", "venue": "arXiv", "keywords": "IDC, Repeated-Round Reflection, Skill-Prompt Refinement", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2606.09826" }, { "label": "Project", "url": "https://mxlin043.github.io/OmniGameArena/" }, { "label": "Code", "url": "https://github.com/mxlin043/OmniGameArena" } ] }, { "section": "Benchmark", "line": 278, "date": "2026-06", "title": "OmniGameArena: A Unified UE5 Benchmark for VLM Game Agents with Improvement Dynamics", "venue": "arXiv", "keywords": "12 UE5 Games, Solo/PvP/Coop, Improvement Dynamics, Held-Out Variants", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2606.09826" }, { "label": "Project", "url": "https://mxlin043.github.io/OmniGameArena/" }, { "label": "Code", "url": "https://github.com/mxlin043/OmniGameArena" } ] } ], "sources": [ "https://arxiv.org/abs/2606.09826", "https://mxlin043.github.io/OmniGameArena/", "https://github.com/mxlin043/OmniGameArena" ], "search_text": "omnigamearena: a unified ue5 benchmark for vlm game agents with improvement dynamics omnigamearena / idc 2026-06-08 arxiv harness benchmark idc repeated-round reflection skill-prompt refinement 12 ue5 games solo/pvp/coop improvement dynamics held-out variants twelve new ue5 games spanning solo, competitive, and cooperative play, with an improvement dynamics curve harness that refines skill prompts over repeated rounds and tests transfer to held-out variants. seven solo, three pvp, two cooperative games. readme distinguishes paused decision-quality and latency-charged tracks. current project page reports expanded reflection analysis relative to original arxiv abstract; avoid copying inconsistent agent counts. player" }, { "id": "arxiv-2606.24893", "name": "AgentOdyssey", "title": "AgentOdyssey: Open-Ended Long-Horizon Text Game Generation for Test-Time Continual Learning Agents", "date": "2026-05-29", "date_precision": "day", "date_basis": "First arXiv submission date", "venue": "NeurIPS 2026", "categories": [ "Benchmark" ], "primary_category": "Benchmark", "keywords": [ "Generated Text Worlds", "Test-Time Continual Learning", "Memory", "Long-Horizon Planning" ], "keywords_by_category": { "Benchmark": "Generated Text Worlds, Test-Time Continual Learning, Memory, Long-Horizon Planning" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2606.24893" }, { "label": "Project", "url": "https://agentodyssey.github.io/" }, { "label": "Code", "url": "https://github.com/agentodyssey/agentodyssey" } ], "summary": "Procedurally generated open-ended text games for test-time continual learning, measuring long-horizon progress, knowledge acquisition, episodic memory, exploration, and cost across multiple agent paradigms.", "release_status": "code: Author-linked public repository verified; data: Generated game environments and generator included in the official repository; not a gameplay training dataset; weights: No downloadable model weights verified in this review", "notes": "ArXiv header and submission history explicitly state May 29 despite the 2606 identifier; latest abstract evaluates six games and distinguishes SFT catastrophic forgetting from memory gains. Accepted NeurIPS 2026 is stated by the authors in arXiv comments.", "source": "update", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 279, "date": "2026-05", "title": "AgentOdyssey: Open-Ended Long-Horizon Text Game Generation for Test-Time Continual Learning Agents", "venue": "NeurIPS 2026", "keywords": "Generated Text Worlds, Test-Time Continual Learning, Memory, Long-Horizon Planning", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2606.24893" }, { "label": "Project", "url": "https://agentodyssey.github.io/" }, { "label": "Code", "url": "https://github.com/agentodyssey/agentodyssey" } ] } ], "sources": [ "https://arxiv.org/abs/2606.24893", "https://agentodyssey.github.io/", "https://github.com/agentodyssey/agentodyssey" ], "search_text": "agentodyssey: open-ended long-horizon text game generation for test-time continual learning agents agentodyssey 2026-05-29 neurips 2026 benchmark generated text worlds test-time continual learning memory long-horizon planning procedurally generated open-ended text games for test-time continual learning, measuring long-horizon progress, knowledge acquisition, episodic memory, exploration, and cost across multiple agent paradigms. arxiv header and submission history explicitly state may 29 despite the 2606 identifier; latest abstract evaluates six games and distinguishes sft catastrophic forgetting from memory gains. accepted neurips 2026 is stated by the authors in arxiv comments. player" }, { "id": "arxiv-2605.29653", "name": "PTCG-Bench", "title": "PTCG-Bench: Can LLM Agents Master Pokémon Trading Card Game?", "date": "2026-05-28", "date_precision": "day", "date_basis": "First arXiv submission date", "venue": "EMNLP 2026 Findings", "categories": [ "Harness", "Benchmark" ], "primary_category": "Benchmark", "keywords": [ "Modular Agent Harness", "Experience Accumulation", "Self-Evolution", "Pokémon TCG", "Strategic Reasoning", "Harness Ablations" ], "keywords_by_category": { "Harness": "Modular Agent Harness, Experience Accumulation, Self-Evolution", "Benchmark": "Pokémon TCG, Strategic Reasoning, Self-Evolution, Harness Ablations" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.29653" }, { "label": "Code", "url": "https://github.com/zjunet/PTCG-Bench" } ], "summary": "Pokémon Trading Card Game evaluation separates single-environment decision making from improvement across play and examines sensitivity to modular harness design.", "release_status": "Code available; September 19 revision adds explicit harness-aware framing.", "notes": "", "source": "update", "focus": "player", "source_rows": [ { "section": "Harness", "line": 185, "date": "2026-05", "title": "PTCG-Bench: Can LLM Agents Master Pokémon Trading Card Game?", "venue": "EMNLP 2026 Findings", "keywords": "Modular Agent Harness, Experience Accumulation, Self-Evolution", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.29653" }, { "label": "Code", "url": "https://github.com/zjunet/PTCG-Bench" } ] }, { "section": "Benchmark", "line": 280, "date": "2026-05", "title": "PTCG-Bench: Can LLM Agents Master Pokémon Trading Card Game?", "venue": "EMNLP 2026 Findings", "keywords": "Pokémon TCG, Strategic Reasoning, Self-Evolution, Harness Ablations", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.29653" }, { "label": "Code", "url": "https://github.com/zjunet/PTCG-Bench" } ] } ], "sources": [ "https://arxiv.org/abs/2605.29653", "https://github.com/zjunet/PTCG-Bench" ], "search_text": "ptcg-bench: can llm agents master pokémon trading card game? ptcg-bench 2026-05-28 emnlp 2026 findings harness benchmark modular agent harness experience accumulation self-evolution pokémon tcg strategic reasoning harness ablations pokémon trading card game evaluation separates single-environment decision making from improvement across play and examines sensitivity to modular harness design. player" }, { "id": "arxiv-2605.29512", "name": "MINDGAMES", "title": "MINDGAMES: A Live Arena for Evaluating Social and Strategic Reasoning in Multi-Agent LLMs", "date": "2026-05-28", "date_precision": "day", "date_basis": "First arXiv submission date", "venue": "arXiv / NeurIPS 2025 competition report", "categories": [ "Dataset", "Benchmark" ], "primary_category": "Benchmark", "keywords": [ "MGC2025", "29", "571 Games", "Turn-Level Multiagent Trajectories", "4 Social/Strategic Games", "TrueSkill", "MG-Ref", "Error Attribution" ], "keywords_by_category": { "Dataset": "MGC2025, 29,571 Games, Turn-Level Multiagent Trajectories", "Benchmark": "4 Social/Strategic Games, TrueSkill, MG-Ref, Error Attribution" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.29512" }, { "label": "Code", "url": "https://github.com/mind-games-challenge/mindgames-starter-kit" }, { "label": "Data", "url": "https://huggingface.co/datasets/mindgameschallenge/MGC2025" } ], "summary": "Four-game social and strategic evaluation with a fixed reference-agent protocol and 29,571 multi-agent games containing turn-level observations, actions and rewards.", "release_status": "Report, reference-agent starter kit and dataset available; competition originated in 2025, paper first posted May 2026.", "notes": "", "source": "update", "focus": "player", "source_rows": [ { "section": "Dataset", "line": 241, "date": "2026-05", "title": "MINDGAMES: A Live Arena for Evaluating Social and Strategic Reasoning in Multi-Agent LLMs", "venue": "arXiv / NeurIPS 2025 competition report", "keywords": "MGC2025, 29,571 Games, Turn-Level Multiagent Trajectories", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.29512" }, { "label": "Code", "url": "https://github.com/mind-games-challenge/mindgames-starter-kit" }, { "label": "Data", "url": "https://huggingface.co/datasets/mindgameschallenge/MGC2025" } ] }, { "section": "Benchmark", "line": 281, "date": "2026-05", "title": "MINDGAMES: A Live Arena for Evaluating Social and Strategic Reasoning in Multi-Agent LLMs", "venue": "arXiv / NeurIPS 2025 competition report", "keywords": "4 Social/Strategic Games, TrueSkill, MG-Ref, Error Attribution", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.29512" }, { "label": "Code", "url": "https://github.com/mind-games-challenge/mindgames-starter-kit" } ] } ], "sources": [ "https://arxiv.org/abs/2605.29512", "https://github.com/mind-games-challenge/mindgames-starter-kit", "https://huggingface.co/datasets/mindgameschallenge/MGC2025" ], "search_text": "mindgames: a live arena for evaluating social and strategic reasoning in multi-agent llms mindgames 2026-05-28 arxiv / neurips 2025 competition report dataset benchmark mgc2025 29 571 games turn-level multiagent trajectories 4 social/strategic games trueskill mg-ref error attribution four-game social and strategic evaluation with a fixed reference-agent protocol and 29,571 multi-agent games containing turn-level observations, actions and rewards. player" }, { "id": "arxiv-2605.23345", "name": "SCOPE / CrossFPS", "title": "SCOPE: Simulating Cross-game Operations in Playable Environments for FPS World Models", "date": "2026-05-22", "date_precision": "day", "date_basis": "First arXiv submission date", "venue": "arXiv", "categories": [ "Dataset", "Model" ], "primary_category": "Model", "keywords": [ "CrossFPS", "69K Clips", "7 FPS Games", "Frame-Aligned Controller Signals", "FPS World Model", "Spatially Selective Action Conditioning", "Cross-Game Transfer" ], "keywords_by_category": { "Dataset": "CrossFPS, 69K Clips, 7 FPS Games, Frame-Aligned Controller Signals", "Model": "FPS World Model, Spatially Selective Action Conditioning, Cross-Game Transfer" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.23345" }, { "label": "Project", "url": "https://z2tong.github.io/SCOPE/" }, { "label": "Code", "url": "https://github.com/z2tong/SCOPE" }, { "label": "Data", "url": "https://huggingface.co/datasets/zizhaotong/CrossFPS-train" } ], "summary": "Action-conditioned FPS world model with spatially selective control; CrossFPS provides 69K clips from seven games with frame-aligned 10-DoF controller signals.", "release_status": "Code and CrossFPS training split available; a world simulator, not an autonomous game-playing policy.", "notes": "", "source": "update", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 103, "date": "2026-05", "title": "SCOPE: Simulating Cross-game Operations in Playable Environments for FPS World Models", "venue": "arXiv", "keywords": "FPS World Model, Spatially Selective Action Conditioning, Cross-Game Transfer", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.23345" }, { "label": "Project", "url": "https://z2tong.github.io/SCOPE/" }, { "label": "Code", "url": "https://github.com/z2tong/SCOPE" } ] }, { "section": "Dataset", "line": 242, "date": "2026-05", "title": "SCOPE: Simulating Cross-game Operations in Playable Environments for FPS World Models", "venue": "arXiv", "keywords": "CrossFPS, 69K Clips, 7 FPS Games, Frame-Aligned Controller Signals", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.23345" }, { "label": "Project", "url": "https://z2tong.github.io/SCOPE/" }, { "label": "Code", "url": "https://github.com/z2tong/SCOPE" }, { "label": "Data", "url": "https://huggingface.co/datasets/zizhaotong/CrossFPS-train" } ] } ], "sources": [ "https://arxiv.org/abs/2605.23345", "https://z2tong.github.io/SCOPE/", "https://github.com/z2tong/SCOPE", "https://huggingface.co/datasets/zizhaotong/CrossFPS-train" ], "search_text": "scope: simulating cross-game operations in playable environments for fps world models scope / crossfps 2026-05-22 arxiv dataset model crossfps 69k clips 7 fps games frame-aligned controller signals fps world model spatially selective action conditioning cross-game transfer action-conditioned fps world model with spatially selective control; crossfps provides 69k clips from seven games with frame-aligned 10-dof controller signals. supporting environment" }, { "id": "arxiv-2605.21240", "name": "APEX", "title": "APEX: Autonomous Policy Exploration for Self-Evolving LLM Agents", "date": "2026-05-20", "date_precision": "day", "date_basis": "First arXiv submission date", "venue": "arXiv", "categories": [ "Harness" ], "primary_category": "Harness", "keywords": [ "Strategy DAG", "Exploration Management", "Cross-Episode Self-Evolution", "Jericho" ], "keywords_by_category": { "Harness": "Strategy DAG, Exploration Management, Cross-Episode Self-Evolution, Jericho" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.21240" }, { "label": "Code", "url": "https://github.com/liushiliushi/APEX1" } ], "summary": "Maintains a directed acyclic strategy map with fork discovery and exploration-aware policy selection to prevent exploration collapse across repeated Jericho text-adventure episodes, without updating model weights.", "release_status": "code: Author-linked public repository verified; data: No standalone downloadable dataset release verified in this review; weights: No downloadable model weights verified in this review", "notes": "Evaluated on nine Jericho games and WebArena. Code URL explicitly linked by the paper.", "source": "update", "focus": "player", "source_rows": [ { "section": "Harness", "line": 186, "date": "2026-05", "title": "APEX: Autonomous Policy Exploration for Self-Evolving LLM Agents", "venue": "arXiv", "keywords": "Strategy DAG, Exploration Management, Cross-Episode Self-Evolution, Jericho", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.21240" }, { "label": "Code", "url": "https://github.com/liushiliushi/APEX1" } ] } ], "sources": [ "https://arxiv.org/abs/2605.21240", "https://arxiv.org/html/2605.21240v1", "https://github.com/liushiliushi/APEX1" ], "search_text": "apex: autonomous policy exploration for self-evolving llm agents apex 2026-05-20 arxiv harness strategy dag exploration management cross-episode self-evolution jericho maintains a directed acyclic strategy map with fork discovery and exploration-aware policy selection to prevent exploration collapse across repeated jericho text-adventure episodes, without updating model weights. evaluated on nine jericho games and webarena. code url explicitly linked by the paper. player" }, { "id": "arxiv-2605.18636", "name": "SPIKE", "title": "SPIKE: An Adaptive Dual Controller Framework for Cost-Efficient Long-Horizon Game Agents", "date": "2026-05-18", "date_precision": "day", "date_basis": "arXiv v1", "venue": "arXiv", "categories": [ "Harness" ], "primary_category": "Harness", "keywords": [ "Adaptive Dual Controllers", "Event-Triggered Planning", "Memory", "Stardew Valley" ], "keywords_by_category": { "Harness": "Adaptive Dual Controllers, Event-Triggered Planning, Memory, Stardew Valley" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.18636" }, { "label": "Project", "url": "https://wencanjiang.github.io/projects/SPIKE/" }, { "label": "Code", "url": "https://github.com/wencanjiang/SPIKE" } ], "summary": "Uses event-triggered strategic planning, fast reactive execution, and controller-specific memory to reduce token and latency costs in long-horizon Stardew Valley tasks.", "release_status": "Source code available.", "notes": "Evaluated on existing StarDojo Lite-100; do not classify as a new benchmark or a trained new foundation model.", "source": "update", "focus": "player", "source_rows": [ { "section": "Harness", "line": 187, "date": "2026-05", "title": "SPIKE: An Adaptive Dual Controller Framework for Cost-Efficient Long-Horizon Game Agents", "venue": "arXiv", "keywords": "Adaptive Dual Controllers, Event-Triggered Planning, Memory, Stardew Valley", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.18636" }, { "label": "Project", "url": "https://wencanjiang.github.io/projects/SPIKE/" }, { "label": "Code", "url": "https://github.com/wencanjiang/SPIKE" } ] } ], "sources": [ "https://arxiv.org/abs/2605.18636", "https://wencanjiang.github.io/projects/SPIKE/", "https://github.com/wencanjiang/SPIKE" ], "search_text": "spike: an adaptive dual controller framework for cost-efficient long-horizon game agents spike 2026-05-18 arxiv harness adaptive dual controllers event-triggered planning memory stardew valley uses event-triggered strategic planning, fast reactive execution, and controller-specific memory to reduce token and latency costs in long-horizon stardew valley tasks. evaluated on existing stardojo lite-100; do not classify as a new benchmark or a trained new foundation model. player" }, { "id": "work-568f87de2f48", "name": "Agora-1 (Odyssey)", "title": "Agora-1: The Multi-Agent World Model", "date": "2026-05-18", "date_precision": "day", "date_basis": "Official research announcement date", "venue": "Research announcement", "categories": [ "Model" ], "primary_category": "Model", "keywords": [ "Multiplayer WM", "Shared Simulation", "Separate State/Rendering", "GoldenEye" ], "keywords_by_category": { "Model": "Multiplayer WM, Shared Simulation, Separate State/Rendering, GoldenEye" }, "links": [ { "label": "Project", "url": "https://odyssey.systems/introducing-agora-1" } ], "summary": "A multiplayer world-model research preview based on GoldenEye that separates learned state evolution from rendering, maintaining a shared simulation for up to four simultaneous participants.", "release_status": "code: No author-linked public code release verified; data: No standalone downloadable dataset release verified in this review; weights: No downloadable model weights verified in this review", "notes": "Official announcement dates release May 18, not May 19 as some secondary reports say. Treat this as a learned simulator, not evidence of generalist gameplay competence.", "source": "update", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 104, "date": "2026-05", "title": "Agora-1: The Multi-Agent World Model", "venue": "Research announcement", "keywords": "Multiplayer WM, Shared Simulation, Separate State/Rendering, GoldenEye", "links": [ { "label": "Project", "url": "https://odyssey.systems/introducing-agora-1" } ] } ], "sources": [ "https://odyssey.systems/introducing-agora-1" ], "search_text": "agora-1: the multi-agent world model agora-1 (odyssey) 2026-05-18 research announcement model multiplayer wm shared simulation separate state/rendering goldeneye a multiplayer world-model research preview based on goldeneye that separates learned state evolution from rendering, maintaining a shared simulation for up to four simultaneous participants. official announcement dates release may 18, not may 19 as some secondary reports say. treat this as a learned simulator, not evidence of generalist gameplay competence. supporting environment" }, { "id": "arxiv-2605.15256", "name": "ReactiveGWM", "title": "ReactiveGWM: Flexible Control and NPC Reactivity in Game World Models", "date": "2026-05-14", "date_precision": "day", "date_basis": "arXiv v1; v2 code released September 29 and paper revised October 5", "venue": "arXiv", "categories": [ "Model" ], "primary_category": "Model", "keywords": [ "Reactive NPCs", "Spatial Character Binding", "Flexible Action/Rule Control; v2 Oct 2026" ], "keywords_by_category": { "Model": "Reactive NPCs, Spatial Character Binding, Flexible Action/Rule Control; v2 Oct 2026" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.15256" }, { "label": "Project", "url": "https://inv-wzq.github.io/ReactiveGWM/" }, { "label": "Code", "url": "https://github.com/INV-WZQ/ReactiveGWM" }, { "label": "Models", "url": "https://huggingface.co/INV-WZQ/ReactiveGWM-v2-Models" } ], "summary": "Uses character-specific spatial binding and unified action or NPC-rule conditioning to generate reactive multi-character gameplay with flexible control assignments.", "release_status": "Training/inference code and Hugging Face model/data support are released; current main is v2 and original implementation is retained on v1 branch.", "notes": "Use current arXiv title. Original May title was 'Steering NPC in Reactive Game World Models'. Project site remains on the older page branch, so current arXiv and main README are preferable for v2 claims. Raw external control and NPC reactivity do not make this a native-action game-playing WAM policy.", "source": "update", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 105, "date": "2026-05", "title": "ReactiveGWM: Flexible Control and NPC Reactivity in Game World Models", "venue": "arXiv", "keywords": "Reactive NPCs, Spatial Character Binding, Flexible Action/Rule Control; v2 Oct 2026", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.15256" }, { "label": "Project", "url": "https://inv-wzq.github.io/ReactiveGWM/" }, { "label": "Code", "url": "https://github.com/INV-WZQ/ReactiveGWM" }, { "label": "Models", "url": "https://huggingface.co/INV-WZQ/ReactiveGWM-v2-Models" } ] } ], "sources": [ "https://arxiv.org/abs/2605.15256", "https://github.com/INV-WZQ/ReactiveGWM" ], "search_text": "reactivegwm: flexible control and npc reactivity in game world models reactivegwm 2026-05-14 arxiv model reactive npcs spatial character binding flexible action/rule control; v2 oct 2026 uses character-specific spatial binding and unified action or npc-rule conditioning to generate reactive multi-character gameplay with flexible control assignments. use current arxiv title. original may title was 'steering npc in reactive game world models'. project site remains on the older page branch, so current arxiv and main readme are preferable for v2 claims. raw external control and npc reactivity do not make this a native-action game-playing wam policy. supporting environment" }, { "id": "arxiv-2605.09998", "name": "Continual Harness", "title": "Continual Harness: Online Adaptation for Self-Improving Foundation Agents", "date": "2026-05-11", "date_precision": "day", "date_basis": "First arXiv submission date", "venue": "arXiv", "categories": [ "Harness" ], "primary_category": "Harness", "keywords": [ "Reset-Free Adaptation", "Prompts", "Subagents", "Skills", "Memory", "Pokémon" ], "keywords_by_category": { "Harness": "Reset-Free Adaptation, Prompts, Subagents, Skills, Memory, Pokémon" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.09998" }, { "label": "Project", "url": "https://sethkarten.ai/continual-harness/" }, { "label": "Code", "url": "https://github.com/sethkarten/continual-harness" } ], "summary": "Reset-free online refinement of prompts, subagents, skills, and memory during Pokemon play; a separate process-reward co-learning experiment also updates an open-source model.", "release_status": "code: Author-linked public repository verified; data: No standalone downloadable dataset release verified in this review; weights: No downloadable model weights verified in this review", "notes": "Do not conflate its earlier human-in-the-loop Gemini Plays Pokemon completions with autonomous Continual Harness results. The primary paper evaluates Pokemon Red and Emerald; an official follow-up page at https://continual-harness.github.io/ reports ARC-AGI-3 experiments.", "source": "update", "focus": "player", "source_rows": [ { "section": "Harness", "line": 188, "date": "2026-05", "title": "Continual Harness: Online Adaptation for Self-Improving Foundation Agents", "venue": "arXiv", "keywords": "Reset-Free Adaptation, Prompts, Subagents, Skills, Memory, Pokémon", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.09998" }, { "label": "Project", "url": "https://sethkarten.ai/continual-harness/" }, { "label": "Code", "url": "https://github.com/sethkarten/continual-harness" } ] } ], "sources": [ "https://arxiv.org/abs/2605.09998", "https://sethkarten.ai/continual-harness/", "https://github.com/sethkarten/continual-harness" ], "search_text": "continual harness: online adaptation for self-improving foundation agents continual harness 2026-05-11 arxiv harness reset-free adaptation prompts subagents skills memory pokémon reset-free online refinement of prompts, subagents, skills, and memory during pokemon play; a separate process-reward co-learning experiment also updates an open-source model. do not conflate its earlier human-in-the-loop gemini plays pokemon completions with autonomous continual harness results. the primary paper evaluates pokemon red and emerald; an official follow-up page at https://continual-harness.github.io/ reports arc-agi-3 experiments. player" }, { "id": "arxiv-2605.04312", "name": "Agent Island", "title": "Agent Island: A Saturation- and Contamination-Resistant Benchmark from Multiagent Games", "date": "2026-05-05", "date_precision": "day", "date_basis": "First arXiv submission date", "venue": "arXiv", "categories": [ "Dataset", "Benchmark" ], "primary_category": "Benchmark", "keywords": [ "Public Multiagent Game Logs", "Dialogue", "Voting", "Behavior Analysis", "Social Reasoning", "Adaptive Opponents", "Uncertainty-Aware Skill Ranking" ], "keywords_by_category": { "Dataset": "Public Multiagent Game Logs, Dialogue, Voting, Behavior Analysis", "Benchmark": "Social Reasoning, Adaptive Opponents, Uncertainty-Aware Skill Ranking" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.04312" }, { "label": "Project", "url": "https://www.agentisland.ai/" }, { "label": "Data", "url": "https://www.agentisland.ai/logs" } ], "summary": "Dynamic multiplayer evaluation of cooperation, persuasion and conflict, with uncertainty-aware skill ranking and public game logs.", "release_status": "Paper, live benchmark and public logs available.", "notes": "", "source": "update", "focus": "player", "source_rows": [ { "section": "Dataset", "line": 243, "date": "2026-05", "title": "Agent Island: A Saturation- and Contamination-Resistant Benchmark from Multiagent Games", "venue": "arXiv", "keywords": "Public Multiagent Game Logs, Dialogue, Voting, Behavior Analysis", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.04312" }, { "label": "Project", "url": "https://www.agentisland.ai/" }, { "label": "Data", "url": "https://www.agentisland.ai/logs" } ] }, { "section": "Benchmark", "line": 282, "date": "2026-05", "title": "Agent Island: A Saturation- and Contamination-Resistant Benchmark from Multiagent Games", "venue": "arXiv", "keywords": "Social Reasoning, Adaptive Opponents, Uncertainty-Aware Skill Ranking", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.04312" }, { "label": "Project", "url": "https://www.agentisland.ai/" } ] } ], "sources": [ "https://arxiv.org/abs/2605.04312", "https://www.agentisland.ai/", "https://www.agentisland.ai/logs" ], "search_text": "agent island: a saturation- and contamination-resistant benchmark from multiagent games agent island 2026-05-05 arxiv dataset benchmark public multiagent game logs dialogue voting behavior analysis social reasoning adaptive opponents uncertainty-aware skill ranking dynamic multiplayer evaluation of cooperation, persuasion and conflict, with uncertainty-aware skill ranking and public game logs. player" }, { "id": "arxiv-2605.00347", "name": "Odysseus", "title": "Odysseus: Scaling VLMs to 100+ Turn Decision-Making in Games via Reinforcement Learning", "date": "2026-05-01", "date_precision": "day", "date_basis": "First arXiv submission date; COLM 2026 acceptance identified in the official training repository", "venue": "COLM 2026", "categories": [ "Model" ], "primary_category": "Model", "keywords": [ "VLM Policy", "PPO", "Turn-Level CNN Critic", "100+ Turns", "Mario", "Cross-Game Transfer" ], "keywords_by_category": { "Model": "VLM Policy, PPO, Turn-Level CNN Critic, 100+ Turns, Mario, Cross-Game Transfer" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.00347" }, { "label": "Project", "url": "https://odysseus-project.github.io/" }, { "label": "Code", "url": "https://github.com/odysseus-project/odysseus-mario-training" }, { "label": "Models", "url": "https://huggingface.co/collections/Odysseus-Project/odysseus" } ], "summary": "Trains VLM policies for long-horizon Super Mario Land gameplay using adapted PPO, a lightweight turn-level critic, positive-advantage filtering, and multi-level curriculum sampling. The study evaluates in-game and cross-game generalization while checking retention of general-domain capabilities.", "release_status": "Official RL training repository, companion evaluation repository, and Odysseus/Odysseus-Zero model checkpoints are public. The training README notes that GPU training and checkpoint export remain unvalidated for this code release.", "notes": "Distinct from the Odyssey company world-model releases and AgentOdyssey. Classified by its learned VLM policy and RL method; the training framework is not a separate deployment-time harness contribution. The released training presets initialize from Qwen3-VL-8B-Instruct; the paper also studies supervised initialization.", "source": "update", "focus": "player", "source_rows": [ { "section": "Model", "line": 106, "date": "2026-05", "title": "Odysseus: Scaling VLMs to 100+ Turn Decision-Making in Games via Reinforcement Learning", "venue": "COLM 2026", "keywords": "VLM Policy, PPO, Turn-Level CNN Critic, 100+ Turns, Mario, Cross-Game Transfer", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2605.00347" }, { "label": "Project", "url": "https://odysseus-project.github.io/" }, { "label": "Code", "url": "https://github.com/odysseus-project/odysseus-mario-training" }, { "label": "Models", "url": "https://huggingface.co/collections/Odysseus-Project/odysseus" } ] } ], "sources": [ "https://arxiv.org/abs/2605.00347", "https://arxiv.org/html/2605.00347v1", "https://odysseus-project.github.io/", "https://github.com/odysseus-project/odysseus-mario-training", "https://github.com/odysseus-project/odysseus-mario-evaluation", "https://huggingface.co/collections/Odysseus-Project/odysseus" ], "search_text": "odysseus: scaling vlms to 100+ turn decision-making in games via reinforcement learning odysseus 2026-05-01 colm 2026 model vlm policy ppo turn-level cnn critic 100+ turns mario cross-game transfer trains vlm policies for long-horizon super mario land gameplay using adapted ppo, a lightweight turn-level critic, positive-advantage filtering, and multi-level curriculum sampling. the study evaluates in-game and cross-game generalization while checking retention of general-domain capabilities. distinct from the odyssey company world-model releases and agentodyssey. classified by its learned vlm policy and rl method; the training framework is not a separate deployment-time harness contribution. the released training presets initialize from qwen3-vl-8b-instruct; the paper also studies supervised initialization. player" }, { "id": "arxiv-2604.08340", "name": "PokeGym", "title": "PokeGym: A Visually-Driven Long-Horizon Benchmark for Vision-Language Models", "date": "2026-04", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "3D RPG", "Pure Pixels", "Long-Horizon", "Automated Evaluation" ], "keywords_by_category": { "Benchmark": "3D RPG, Pure Pixels, Long-Horizon, Automated Evaluation" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2604.08340" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 284, "date": "2026-04", "title": "PokeGym: A Visually-Driven Long-Horizon Benchmark for Vision-Language Models", "venue": "arXiv", "keywords": "3D RPG, Pure Pixels, Long-Horizon, Automated Evaluation", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2604.08340" } ] } ], "sources": [ "https://arxiv.org/abs/2604.08340" ], "search_text": "pokegym: a visually-driven long-horizon benchmark for vision-language models pokegym 2026-04 arxiv benchmark 3d rpg pure pixels long-horizon automated evaluation player" }, { "id": "arxiv-2604.08995", "name": "Matrix-Game 3.0", "title": "Matrix-Game 3.0: Real-Time and Streaming Interactive World Model with Long-Horizon Memory", "date": "2026-04", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Long-Horizon Memory", "Distribution Matching Distillation", "720p Real-Time Generation" ], "keywords_by_category": { "Model": "Long-Horizon Memory, Distribution Matching Distillation, 720p Real-Time Generation" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2604.08995" }, { "label": "Code", "url": "https://github.com/SkyworkAI/Matrix-Game" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 107, "date": "2026-04", "title": "Matrix-Game 3.0: Real-Time and Streaming Interactive World Model with Long-Horizon Memory", "venue": "arXiv", "keywords": "Long-Horizon Memory, Distribution Matching Distillation, 720p Real-Time Generation", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2604.08995" }, { "label": "Code", "url": "https://github.com/SkyworkAI/Matrix-Game" } ] } ], "sources": [ "https://arxiv.org/abs/2604.08995", "https://github.com/SkyworkAI/Matrix-Game" ], "search_text": "matrix-game 3.0: real-time and streaming interactive world model with long-horizon memory matrix-game 3.0 2026-04 arxiv model long-horizon memory distribution matching distillation 720p real-time generation supporting environment" }, { "id": "arxiv-2604.07429", "name": "GameWorld", "title": "GameWorld: Towards Standardized and Verifiable Evaluation of Multimodal Game Agents", "date": "2026-04", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Browser Games", "Multimodal Agents", "Standardized Evaluation", "Verifiable Metrics" ], "keywords_by_category": { "Benchmark": "Browser Games, Multimodal Agents, Standardized Evaluation, Verifiable Metrics" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2604.07429v1" }, { "label": "Code", "url": "https://github.com/gameworld-project/gameworld" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 283, "date": "2026-04", "title": "GameWorld: Towards Standardized and Verifiable Evaluation of Multimodal Game Agents", "venue": "arXiv", "keywords": "Browser Games, Multimodal Agents, Standardized Evaluation, Verifiable Metrics", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2604.07429v1" }, { "label": "Code", "url": "https://github.com/gameworld-project/gameworld" } ] } ], "sources": [ "https://arxiv.org/abs/2604.07429v1", "https://github.com/gameworld-project/gameworld" ], "search_text": "gameworld: towards standardized and verifiable evaluation of multimodal game agents gameworld 2026-04 arxiv benchmark browser games multimodal agents standardized evaluation verifiable metrics player" }, { "id": "arxiv-2603.16871", "name": "WorldCam", "title": "WorldCam: Interactive Autoregressive 3D Gaming Worlds with Camera Pose as a Unifying Geometric Representation", "date": "2026-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Camera Pose Control", "Long-Horizon 3D Consistency", "Autoregressive World Model" ], "keywords_by_category": { "Model": "Camera Pose Control, Long-Horizon 3D Consistency, Autoregressive World Model" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.16871" }, { "label": "Code", "url": "https://github.com/cvlab-kaist/WorldCam" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 108, "date": "2026-03", "title": "WorldCam: Interactive Autoregressive 3D Gaming Worlds with Camera Pose as a Unifying Geometric Representation", "venue": "arXiv", "keywords": "Camera Pose Control, Long-Horizon 3D Consistency, Autoregressive World Model", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.16871" }, { "label": "Code", "url": "https://github.com/cvlab-kaist/WorldCam" } ] } ], "sources": [ "https://arxiv.org/abs/2603.16871", "https://github.com/cvlab-kaist/WorldCam" ], "search_text": "worldcam: interactive autoregressive 3d gaming worlds with camera pose as a unifying geometric representation worldcam 2026-03 arxiv model camera pose control long-horizon 3d consistency autoregressive world model supporting environment" }, { "id": "arxiv-2603.15563", "name": "The PokeAgent Challenge", "title": "The PokeAgent Challenge: Competitive and Long-Context Learning at Scale", "date": "2026-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NeurIPS 2025 Competition Track / NeurIPS 2025 competition", "categories": [ "Harness", "Benchmark" ], "primary_category": null, "keywords": [ "Textual Environment", "Battle Data Augmentation", "Reasoning Mode", "Time Adaption", "Opponent Modeling", "Reinforcement Learning", "LLMs", "Game AI" ], "keywords_by_category": { "Harness": "Textual Environment, Battle Data Augmentation, Reasoning Mode, Time Adaption", "Benchmark": "Opponent Modeling, Reinforcement Learning, LLMs, Game AI" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.15563" }, { "label": "Code", "url": "https://github.com/sethkarten/pokeagent-speedrun" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 190, "date": "2026-03", "title": "The PokeAgent Challenge: Competitive and Long-Context Learning at Scale", "venue": "NeurIPS 2025 Competition Track", "keywords": "Textual Environment, Battle Data Augmentation, Reasoning Mode, Time Adaption", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.15563" }, { "label": "Code", "url": "https://github.com/sethkarten/pokeagent-speedrun" } ] }, { "section": "Benchmark", "line": 286, "date": "2026-03", "title": "The PokeAgent Challenge: Competitive and Long-Context Learning at Scale", "venue": "NeurIPS 2025 competition", "keywords": "Opponent Modeling, Reinforcement Learning, LLMs, Game AI", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.15563" }, { "label": "Code", "url": "https://github.com/sethkarten/pokeagent-speedrun" } ] } ], "sources": [ "https://arxiv.org/abs/2603.15563", "https://github.com/sethkarten/pokeagent-speedrun" ], "search_text": "the pokeagent challenge: competitive and long-context learning at scale the pokeagent challenge 2026-03 neurips 2025 competition track / neurips 2025 competition harness benchmark textual environment battle data augmentation reasoning mode time adaption opponent modeling reinforcement learning llms game ai player" }, { "id": "arxiv-2603.13131", "name": "Steve-Evolving", "title": "Steve-Evolving: Open-World Embodied Self-Evolution via Fine-Grained Diagnosis and Dual-Track Knowledge Distillation", "date": "2026-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "Reflection", "Skill Library" ], "keywords_by_category": { "Harness": "Reflection, Skill Library" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.13131v1" }, { "label": "Code", "url": "https://github.com/xzw-ustc/Steve-Evolving" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 192, "date": "2026-03", "title": "Steve-Evolving: Open-World Embodied Self-Evolution via Fine-Grained Diagnosis and Dual-Track Knowledge Distillation", "venue": "arXiv", "keywords": "Reflection, Skill Library", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.13131v1" }, { "label": "Code", "url": "https://github.com/xzw-ustc/Steve-Evolving" } ] } ], "sources": [ "https://arxiv.org/abs/2603.13131v1", "https://github.com/xzw-ustc/Steve-Evolving" ], "search_text": "steve-evolving: open-world embodied self-evolution via fine-grained diagnosis and dual-track knowledge distillation steve-evolving 2026-03 arxiv harness reflection skill library player" }, { "id": "arxiv-2603.10512", "name": "Resource-constrained Amazons chess decision framework integrating large language models and graph attention", "title": "Resource-constrained Amazons chess decision framework integrating large language models and graph attention", "date": "2026-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "LLM Distillation", "Specialized Model" ], "keywords_by_category": { "Harness": "LLM Distillation, Specialized Model" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.10512" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 191, "date": "2026-03", "title": "Resource-constrained Amazons chess decision framework integrating large language models and graph attention", "venue": "arXiv", "keywords": "LLM Distillation, Specialized Model", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.10512" } ] } ], "sources": [ "https://arxiv.org/abs/2603.10512" ], "search_text": "resource-constrained amazons chess decision framework integrating large language models and graph attention resource-constrained amazons chess decision framework integrating large language models and graph attention 2026-03 arxiv harness llm distillation specialized model player" }, { "id": "arxiv-2603.03447", "name": "Proact-VL", "title": "Proact-VL: A Proactive VideoLLM for Real-Time AI Companions", "date": "2026-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Dataset", "Model" ], "primary_category": null, "keywords": [ "Live Gaming Benchmark", "Commentary & Guidance Scenarios", "Proactive VideoLLM", "Real-Time Streaming", "SFT" ], "keywords_by_category": { "Dataset": "Live Gaming Benchmark, Commentary & Guidance Scenarios", "Model": "Proactive VideoLLM, Real-Time Streaming, SFT" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.03447" }, { "label": "Code", "url": "https://github.com/microsoft/AnthropomorphicIntelligence/tree/main/Proact-VL" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 109, "date": "2026-03", "title": "Proact-VL: A Proactive VideoLLM for Real-Time AI Companions", "venue": "arXiv", "keywords": "Proactive VideoLLM, Real-Time Streaming, SFT", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.03447" }, { "label": "Code", "url": "https://github.com/microsoft/AnthropomorphicIntelligence/tree/main/Proact-VL" } ] }, { "section": "Dataset", "line": 244, "date": "2026-03", "title": "Proact-VL: A Proactive VideoLLM for Real-Time AI Companions", "venue": "arXiv", "keywords": "Live Gaming Benchmark, Commentary & Guidance Scenarios", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.03447" }, { "label": "Code", "url": "https://github.com/microsoft/AnthropomorphicIntelligence/tree/main/Proact-VL" } ] } ], "sources": [ "https://arxiv.org/abs/2603.03447", "https://github.com/microsoft/AnthropomorphicIntelligence/tree/main/Proact-VL" ], "search_text": "proact-vl: a proactive videollm for real-time ai companions proact-vl 2026-03 arxiv dataset model live gaming benchmark commentary & guidance scenarios proactive videollm real-time streaming sft player" }, { "id": "arxiv-2603.23660", "name": "GTO Wizard Benchmark", "title": "GTO Wizard Benchmark", "date": "2026-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Poker", "HUNL", "Nash Equilibrium", "Superhuman Anchor" ], "keywords_by_category": { "Benchmark": "Poker, HUNL, Nash Equilibrium, Superhuman Anchor" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.23660" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 287, "date": "2026-03", "title": "GTO Wizard Benchmark", "venue": "arXiv", "keywords": "Poker, HUNL, Nash Equilibrium, Superhuman Anchor", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.23660" } ] } ], "sources": [ "https://arxiv.org/abs/2603.23660" ], "search_text": "gto wizard benchmark gto wizard benchmark 2026-03 arxiv benchmark poker hunl nash equilibrium superhuman anchor player" }, { "id": "arxiv-2603.06656", "name": "GameVerse", "title": "GameVerse: Can Vision-Language Models Learn from Video-based Reflection?", "date": "2026-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Harness", "Benchmark" ], "primary_category": null, "keywords": [ "VLM", "Video Reflection", "15 Games" ], "keywords_by_category": { "Harness": "VLM, Video Reflection, 15 Games", "Benchmark": "VLM, Video Reflection, 15 Games" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.06656" }, { "label": "Code", "url": "https://github.com/THUSI-Lab/GameVerse" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 189, "date": "2026-03", "title": "GameVerse: Can Vision-Language Models Learn from Video-based Reflection?", "venue": "arXiv", "keywords": "VLM, Video Reflection, 15 Games", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.06656" }, { "label": "Code", "url": "https://github.com/THUSI-Lab/GameVerse" } ] }, { "section": "Benchmark", "line": 290, "date": "2026-03", "title": "GameVerse: Can Vision-Language Models Learn from Video-based Reflection?", "venue": "arXiv", "keywords": "VLM, Video Reflection, 15 Games", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.06656" }, { "label": "Code", "url": "https://github.com/THUSI-Lab/GameVerse" } ] } ], "sources": [ "https://arxiv.org/abs/2603.06656", "https://github.com/THUSI-Lab/GameVerse" ], "search_text": "gameverse: can vision-language models learn from video-based reflection? gameverse 2026-03 arxiv harness benchmark vlm video reflection 15 games player" }, { "id": "arxiv-2603.24329", "name": "GameplayQA", "title": "GameplayQA: A Benchmarking Framework for Decision-Dense POV-Synced Multi-Video Understanding of 3D Virtual Agents", "date": "2026-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ACL 2026", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "POV-Synced Multi-Video Understanding", "3D Virtual Agents", "Self-Other-World Taxonomy", "2.4K QA Pairs", "15 Task Categories" ], "keywords_by_category": { "Benchmark": "POV-Synced Multi-Video Understanding, 3D Virtual Agents, Self-Other-World Taxonomy, 2.4K QA Pairs, 15 Task Categories" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.24329" }, { "label": "Code", "url": "https://hats-ict.github.io/gameplayqa/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 285, "date": "2026-03", "title": "GameplayQA: A Benchmarking Framework for Decision-Dense POV-Synced Multi-Video Understanding of 3D Virtual Agents", "venue": "ACL 2026", "keywords": "POV-Synced Multi-Video Understanding, 3D Virtual Agents, Self-Other-World Taxonomy, 2.4K QA Pairs, 15 Task Categories", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.24329" }, { "label": "Code", "url": "https://hats-ict.github.io/gameplayqa/" } ] } ], "sources": [ "https://arxiv.org/abs/2603.24329", "https://hats-ict.github.io/gameplayqa/" ], "search_text": "gameplayqa: a benchmarking framework for decision-dense pov-synced multi-video understanding of 3d virtual agents gameplayqa 2026-03 acl 2026 benchmark pov-synced multi-video understanding 3d virtual agents self-other-world taxonomy 2.4k qa pairs 15 task categories player" }, { "id": "arxiv-2603.09337", "name": "Beyond Scaling", "title": "Beyond Scaling: Assessing Strategic Reasoning and Rapid Decision-Making Capability of LLMs in Zero-sum Environments", "date": "2026-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Strategic Reasoning", "Zero-sum Games", "Real-time Decision", "STAR Framework" ], "keywords_by_category": { "Benchmark": "Strategic Reasoning, Zero-sum Games, Real-time Decision, STAR Framework" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.09337" }, { "label": "Code", "url": "https://github.com/star-nexus/star" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 289, "date": "2026-03", "title": "Beyond Scaling: Assessing Strategic Reasoning and Rapid Decision-Making Capability of LLMs in Zero-sum Environments", "venue": "arXiv", "keywords": "Strategic Reasoning, Zero-sum Games, Real-time Decision, STAR Framework", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.09337" }, { "label": "Code", "url": "https://github.com/star-nexus/star" } ] } ], "sources": [ "https://arxiv.org/abs/2603.09337", "https://github.com/star-nexus/star" ], "search_text": "beyond scaling: assessing strategic reasoning and rapid decision-making capability of llms in zero-sum environments beyond scaling 2026-03 arxiv benchmark strategic reasoning zero-sum games real-time decision star framework player" }, { "id": "arxiv-2603.24621", "name": "ARC-AGI-3", "title": "ARC-AGI-3: A New Challenge for Frontier Agentic Intelligence", "date": "2026-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "ARC-AGI-3", "Agentic Intelligence", "Interactive Reasoning", "Human Calibration" ], "keywords_by_category": { "Benchmark": "ARC-AGI-3, Agentic Intelligence, Interactive Reasoning, Human Calibration" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.24621v1" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 288, "date": "2026-03", "title": "ARC-AGI-3: A New Challenge for Frontier Agentic Intelligence", "venue": "arXiv", "keywords": "ARC-AGI-3, Agentic Intelligence, Interactive Reasoning, Human Calibration", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2603.24621v1" } ] } ], "sources": [ "https://arxiv.org/abs/2603.24621v1" ], "search_text": "arc-agi-3: a new challenge for frontier agentic intelligence arc-agi-3 2026-03 arxiv benchmark arc-agi-3 agentic intelligence interactive reasoning human calibration player" }, { "id": "arxiv-2602.16833", "name": "VAM", "title": "VAM: Verbalized Action Masking for Controllable Exploration in RL Post-Training -- A Chess Case Study", "date": "2026-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Action Masking", "RL Post-Training", "Chess" ], "keywords_by_category": { "Model": "Action Masking, RL Post-Training, Chess" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.16833" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 112, "date": "2026-02", "title": "VAM: Verbalized Action Masking for Controllable Exploration in RL Post-Training -- A Chess Case Study", "venue": "arXiv", "keywords": "Action Masking, RL Post-Training, Chess", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.16833" } ] } ], "sources": [ "https://arxiv.org/abs/2602.16833" ], "search_text": "vam: verbalized action masking for controllable exploration in rl post-training -- a chess case study vam 2026-02 arxiv model action masking rl post-training chess player" }, { "id": "arxiv-2602.22208", "name": "Solaris", "title": "Solaris: Building a Multiplayer Video World Model in Minecraft", "date": "2026-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Multiplayer World Model", "Cross-Player Consistency", "Self Forcing" ], "keywords_by_category": { "Model": "Multiplayer World Model, Cross-Player Consistency, Self Forcing" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.22208" }, { "label": "Code", "url": "https://github.com/solaris-wm/solaris" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 110, "date": "2026-02", "title": "Solaris: Building a Multiplayer Video World Model in Minecraft", "venue": "arXiv", "keywords": "Multiplayer World Model, Cross-Player Consistency, Self Forcing", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.22208" }, { "label": "Code", "url": "https://github.com/solaris-wm/solaris" } ] } ], "sources": [ "https://arxiv.org/abs/2602.22208", "https://github.com/solaris-wm/solaris" ], "search_text": "solaris: building a multiplayer video world model in minecraft solaris 2026-02 arxiv model multiplayer world model cross-player consistency self forcing supporting environment" }, { "id": "work-98122d9adddf", "name": "ProxyWar", "title": "ProxyWar: Dynamic Assessment of LLM Code Generation in Game Arenas", "date": "2026-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICSE 2026", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "LLM codes modules for game playing" ], "keywords_by_category": { "Harness": "LLM codes modules for game playing" }, "links": [ { "label": "Paper", "url": "https://conf.researchr.org/details/icse-2026/icse-2026-research-track/178/ProxyWar-Dynamic-Assessment-of-LLM-Code-Generation-in-Game-Arenas" }, { "label": "Code", "url": "https://github.com/xinke-wang/ProxyWar" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 193, "date": "2026-02", "title": "ProxyWar: Dynamic Assessment of LLM Code Generation in Game Arenas", "venue": "ICSE 2026", "keywords": "LLM codes modules for game playing", "links": [ { "label": "Paper", "url": "https://conf.researchr.org/details/icse-2026/icse-2026-research-track/178/ProxyWar-Dynamic-Assessment-of-LLM-Code-Generation-in-Game-Arenas" }, { "label": "Code", "url": "https://github.com/xinke-wang/ProxyWar" } ] } ], "sources": [ "https://conf.researchr.org/details/icse-2026/icse-2026-research-track/178/ProxyWar-Dynamic-Assessment-of-LLM-Code-Generation-in-Game-Arenas", "https://github.com/xinke-wang/ProxyWar" ], "search_text": "proxywar: dynamic assessment of llm code generation in game arenas proxywar 2026-02 icse 2026 harness llm codes modules for game playing player" }, { "id": "arxiv-2602.04447", "name": "Mixture of Masters", "title": "Mixture of Masters: Sparse Chess Language Models with Player Routing", "date": "2026-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "MoE", "RL Fine-tuning", "Chess" ], "keywords_by_category": { "Model": "MoE, RL Fine-tuning, Chess" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.04447" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 114, "date": "2026-02", "title": "Mixture of Masters: Sparse Chess Language Models with Player Routing", "venue": "arXiv", "keywords": "MoE, RL Fine-tuning, Chess", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.04447" } ] } ], "sources": [ "https://arxiv.org/abs/2602.04447" ], "search_text": "mixture of masters: sparse chess language models with player routing mixture of masters 2026-02 arxiv model moe rl fine-tuning chess player" }, { "id": "arxiv-2602.02212", "name": "MAIN-VLA", "title": "MAIN-VLA: Modeling Abstraction of Intention and eNvironment for Vision-Language-Action Models", "date": "2026-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "VLA", "Intention Modeling", "SFT" ], "keywords_by_category": { "Model": "VLA, Intention Modeling, SFT" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.02212" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 111, "date": "2026-02", "title": "MAIN-VLA: Modeling Abstraction of Intention and eNvironment for Vision-Language-Action Models", "venue": "arXiv", "keywords": "VLA, Intention Modeling, SFT", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.02212" } ] } ], "sources": [ "https://arxiv.org/abs/2602.02212" ], "search_text": "main-vla: modeling abstraction of intention and environment for vision-language-action models main-vla 2026-02 arxiv model vla intention modeling sft player" }, { "id": "arxiv-2505.17512", "name": "Is Your LLM Really Mastering the Concept? A Multi-Agent Benchmark", "title": "Is Your LLM Really Mastering the Concept? A Multi-Agent Benchmark", "date": "2026-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Conceptual Knowledge", "Multi-agent Interaction", "Undercover Game", "CK-Arena" ], "keywords_by_category": { "Benchmark": "Conceptual Knowledge, Multi-agent Interaction, Undercover Game, CK-Arena" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2505.17512v2" }, { "label": "Code", "url": "https://github.com/xushuhang1122/CK-Arena" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 292, "date": "2026-02", "title": "Is Your LLM Really Mastering the Concept? A Multi-Agent Benchmark", "venue": "arXiv", "keywords": "Conceptual Knowledge, Multi-agent Interaction, Undercover Game, CK-Arena", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2505.17512v2" }, { "label": "Code", "url": "https://github.com/xushuhang1122/CK-Arena" } ] } ], "sources": [ "https://arxiv.org/abs/2505.17512v2", "https://github.com/xushuhang1122/CK-Arena" ], "search_text": "is your llm really mastering the concept? a multi-agent benchmark is your llm really mastering the concept? a multi-agent benchmark 2026-02 arxiv benchmark conceptual knowledge multi-agent interaction undercover game ck-arena player" }, { "id": "arxiv-2602.08041", "name": "Implicit Strategic Optimization", "title": "Implicit Strategic Optimization: Rethinking Long-Horizon Decision-Making in Adversarial Poker Environments", "date": "2026-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "GRPO", "Long-Horizon RL", "Poker" ], "keywords_by_category": { "Model": "GRPO, Long-Horizon RL, Poker" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.08041" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 113, "date": "2026-02", "title": "Implicit Strategic Optimization: Rethinking Long-Horizon Decision-Making in Adversarial Poker Environments", "venue": "arXiv", "keywords": "GRPO, Long-Horizon RL, Poker", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.08041" } ] } ], "sources": [ "https://arxiv.org/abs/2602.08041" ], "search_text": "implicit strategic optimization: rethinking long-horizon decision-making in adversarial poker environments implicit strategic optimization 2026-02 arxiv model grpo long-horizon rl poker player" }, { "id": "arxiv-2602.17594", "name": "AI Gamestore", "title": "AI Gamestore: Scalable, Open-Ended Evaluation of Machine General Intelligence with Human Games", "date": "2026-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Open-Ended Evaluation", "Automated Game Pipeline", "Large Game Space" ], "keywords_by_category": { "Benchmark": "Open-Ended Evaluation, Automated Game Pipeline, Large Game Space" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.17594" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 291, "date": "2026-02", "title": "AI Gamestore: Scalable, Open-Ended Evaluation of Machine General Intelligence with Human Games", "venue": "arXiv", "keywords": "Open-Ended Evaluation, Automated Game Pipeline, Large Game Space", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.17594" } ] } ], "sources": [ "https://arxiv.org/abs/2602.17594" ], "search_text": "ai gamestore: scalable, open-ended evaluation of machine general intelligence with human games ai gamestore 2026-02 arxiv benchmark open-ended evaluation automated game pipeline large game space player" }, { "id": "work-2d2791f5d562", "name": "TowerMind", "title": "TowerMind: A Tower Defence Game Learning Environment and Benchmark for LLM as Agents", "date": "2026-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "AAAI 2026", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Tower Defense", "Real-time Strategy", "Multimodal", "Planning", "Hallucination Eval" ], "keywords_by_category": { "Benchmark": "Tower Defense, Real-time Strategy, Multimodal, Planning, Hallucination Eval" }, "links": [ { "label": "Paper", "url": "https://ojs.aaai.org/index.php/AAAI/article/view/39818" }, { "label": "Code", "url": "https://github.com/tb6147877/TowerMind" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 298, "date": "2026-01", "title": "TowerMind: A Tower Defence Game Learning Environment and Benchmark for LLM as Agents", "venue": "AAAI 2026", "keywords": "Tower Defense, Real-time Strategy, Multimodal, Planning, Hallucination Eval", "links": [ { "label": "Paper", "url": "https://ojs.aaai.org/index.php/AAAI/article/view/39818" }, { "label": "Code", "url": "https://github.com/tb6147877/TowerMind" } ] } ], "sources": [ "https://ojs.aaai.org/index.php/AAAI/article/view/39818", "https://github.com/tb6147877/TowerMind" ], "search_text": "towermind: a tower defence game learning environment and benchmark for llm as agents towermind 2026-01 aaai 2026 benchmark tower defense real-time strategy multimodal planning hallucination eval player" }, { "id": "arxiv-2601.18077", "name": "Sparks of Cooperative Reasoning", "title": "Sparks of Cooperative Reasoning: LLMs as Strategic Hanabi Agents", "date": "2026-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Hanabi", "Cooperative Reasoning", "Imperfect Information", "Strategic Communication" ], "keywords_by_category": { "Benchmark": "Hanabi, Cooperative Reasoning, Imperfect Information, Strategic Communication" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.18077" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 294, "date": "2026-01", "title": "Sparks of Cooperative Reasoning: LLMs as Strategic Hanabi Agents", "venue": "arXiv", "keywords": "Hanabi, Cooperative Reasoning, Imperfect Information, Strategic Communication", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.18077" } ] } ], "sources": [ "https://arxiv.org/abs/2601.18077" ], "search_text": "sparks of cooperative reasoning: llms as strategic hanabi agents sparks of cooperative reasoning 2026-01 arxiv benchmark hanabi cooperative reasoning imperfect information strategic communication player" }, { "id": "arxiv-2601.04575", "name": "Scaling Behavior Cloning Improves Causal Reasoning", "title": "Scaling Behavior Cloning Improves Causal Reasoning: An Open Model for Real-Time Video Game Playing", "date": "2026-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Dataset", "Model" ], "primary_category": null, "keywords": [ "8300+ hrs Human-Annotated", "Open-Source", "3D Games", "Behavior Cloning", "Scaling Law", "Causal Reasoning" ], "keywords_by_category": { "Dataset": "8300+ hrs Human-Annotated, Open-Source, 3D Games", "Model": "Behavior Cloning, Scaling Law, Causal Reasoning" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.04575" }, { "label": "Code", "url": "https://github.com/elefant-ai/open-p2p" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 116, "date": "2026-01", "title": "Scaling Behavior Cloning Improves Causal Reasoning: An Open Model for Real-Time Video Game Playing", "venue": "arXiv", "keywords": "Behavior Cloning, Scaling Law, Causal Reasoning", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.04575" }, { "label": "Code", "url": "https://github.com/elefant-ai/open-p2p" } ] }, { "section": "Dataset", "line": 246, "date": "2026-01", "title": "Scaling Behavior Cloning Improves Causal Reasoning: An Open Model for Real-Time Video Game Playing", "venue": "arXiv", "keywords": "8300+ hrs Human-Annotated, Open-Source, 3D Games", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.04575" }, { "label": "Code", "url": "https://github.com/elefant-ai/open-p2p" } ] } ], "sources": [ "https://arxiv.org/abs/2601.04575", "https://github.com/elefant-ai/open-p2p" ], "search_text": "scaling behavior cloning improves causal reasoning: an open model for real-time video game playing scaling behavior cloning improves causal reasoning 2026-01 arxiv dataset model 8300+ hrs human-annotated open-source 3d games behavior cloning scaling law causal reasoning player" }, { "id": "arxiv-2602.00676", "name": "OpenGuanDan", "title": "OpenGuanDan: A Large-Scale Imperfect Information Game Benchmark", "date": "2026-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "GuanDan", "Imperfect Information", "Cooperation & Competition", "Long-Horizon Decision", "Multi-Agent" ], "keywords_by_category": { "Benchmark": "GuanDan, Imperfect Information, Cooperation & Competition, Long-Horizon Decision, Multi-Agent" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.00676" }, { "label": "Code", "url": "https://github.com/GameAI-NJUPT/OpenGuanDan" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 293, "date": "2026-01", "title": "OpenGuanDan: A Large-Scale Imperfect Information Game Benchmark", "venue": "arXiv", "keywords": "GuanDan, Imperfect Information, Cooperation & Competition, Long-Horizon Decision, Multi-Agent", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.00676" }, { "label": "Code", "url": "https://github.com/GameAI-NJUPT/OpenGuanDan" } ] } ], "sources": [ "https://arxiv.org/abs/2602.00676", "https://github.com/GameAI-NJUPT/OpenGuanDan" ], "search_text": "openguandan: a large-scale imperfect information game benchmark openguandan 2026-01 arxiv benchmark guandan imperfect information cooperation & competition long-horizon decision multi-agent player" }, { "id": "arxiv-2601.02427", "name": "NitroGen", "title": "NitroGen: An Open Foundation Model for Generalist Gaming Agents", "date": "2026-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Dataset", "Model" ], "primary_category": null, "keywords": [ "40K hrs Gameplay", "1000+ Games", "Auto-Extracted Actions", "Behavior Cloning", "Flow Matching", "DiT" ], "keywords_by_category": { "Dataset": "40K hrs Gameplay, 1000+ Games, Auto-Extracted Actions", "Model": "Behavior Cloning, Flow Matching, DiT" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.02427" }, { "label": "Code", "url": "https://github.com/MineDojo/NitroGen" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 115, "date": "2026-01", "title": "NitroGen: An Open Foundation Model for Generalist Gaming Agents", "venue": "arXiv", "keywords": "Behavior Cloning, Flow Matching, DiT", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.02427" }, { "label": "Code", "url": "https://github.com/MineDojo/NitroGen" } ] }, { "section": "Dataset", "line": 245, "date": "2026-01", "title": "NitroGen: An Open Foundation Model for Generalist Gaming Agents", "venue": "arXiv", "keywords": "40K hrs Gameplay, 1000+ Games, Auto-Extracted Actions", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.02427" }, { "label": "Code", "url": "https://github.com/MineDojo/NitroGen" } ] } ], "sources": [ "https://arxiv.org/abs/2601.02427", "https://github.com/MineDojo/NitroGen" ], "search_text": "nitrogen: an open foundation model for generalist gaming agents nitrogen 2026-01 arxiv dataset model 40k hrs gameplay 1000+ games auto-extracted actions behavior cloning flow matching dit player" }, { "id": "arxiv-2601.09017", "name": "Multicultural Spyfall", "title": "Multicultural Spyfall: Assessing LLMs through Dynamic Multilingual Social Deduction Game", "date": "2026-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Spyfall", "Social Deduction", "Multilingual", "Multicultural" ], "keywords_by_category": { "Benchmark": "Spyfall, Social Deduction, Multilingual, Multicultural" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.09017" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 297, "date": "2026-01", "title": "Multicultural Spyfall: Assessing LLMs through Dynamic Multilingual Social Deduction Game", "venue": "arXiv", "keywords": "Spyfall, Social Deduction, Multilingual, Multicultural", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.09017" } ] } ], "sources": [ "https://arxiv.org/abs/2601.09017" ], "search_text": "multicultural spyfall: assessing llms through dynamic multilingual social deduction game multicultural spyfall 2026-01 arxiv benchmark spyfall social deduction multilingual multicultural player" }, { "id": "arxiv-2601.05215", "name": "MineNPC-Task", "title": "MineNPC-Task: Task Suite for Memory-Aware Minecraft Agents", "date": "2026-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Minecraft", "Memory-Aware", "Mixed-Initiative", "Open-world" ], "keywords_by_category": { "Benchmark": "Minecraft, Memory-Aware, Mixed-Initiative, Open-world" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.05215" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 299, "date": "2026-01", "title": "MineNPC-Task: Task Suite for Memory-Aware Minecraft Agents", "venue": "arXiv", "keywords": "Minecraft, Memory-Aware, Mixed-Initiative, Open-world", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.05215" } ] } ], "sources": [ "https://arxiv.org/abs/2601.05215" ], "search_text": "minenpc-task: task suite for memory-aware minecraft agents minenpc-task 2026-01 arxiv benchmark minecraft memory-aware mixed-initiative open-world player" }, { "id": "work-2b7ff9dfd404", "name": "LLMs as Rules Oracles", "title": "LLMs as Rules Oracles: Exploring Real-World Multimodal Reasoning in Tabletop Strategy Game Environments", "date": "2026-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2026", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Tabletop Games", "Rulebook Grounding", "LudoBench", "638 QA Examples" ], "keywords_by_category": { "Benchmark": "Tabletop Games, Rulebook Grounding, LudoBench, 638 QA Examples" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=TOgQ00DEek" }, { "label": "Code", "url": "https://github.com/jpeper/LudoBench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 296, "date": "2026-01", "title": "LLMs as Rules Oracles: Exploring Real-World Multimodal Reasoning in Tabletop Strategy Game Environments", "venue": "ICLR 2026", "keywords": "Tabletop Games, Rulebook Grounding, LudoBench, 638 QA Examples", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=TOgQ00DEek" }, { "label": "Code", "url": "https://github.com/jpeper/LudoBench" } ] } ], "sources": [ "https://openreview.net/forum?id=TOgQ00DEek", "https://github.com/jpeper/LudoBench" ], "search_text": "llms as rules oracles: exploring real-world multimodal reasoning in tabletop strategy game environments llms as rules oracles 2026-01 iclr 2026 benchmark tabletop games rulebook grounding ludobench 638 qa examples player" }, { "id": "arxiv-2601.16276", "name": "GameTalk", "title": "GameTalk: Training LLMs for Strategic Conversation", "date": "2026-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "GRPO/DPO", "Strategic Dialogue" ], "keywords_by_category": { "Model": "GRPO/DPO, Strategic Dialogue" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.16276" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 117, "date": "2026-01", "title": "GameTalk: Training LLMs for Strategic Conversation", "venue": "arXiv", "keywords": "GRPO/DPO, Strategic Dialogue", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.16276" } ] } ], "sources": [ "https://arxiv.org/abs/2601.16276" ], "search_text": "gametalk: training llms for strategic conversation gametalk 2026-01 arxiv model grpo/dpo strategic dialogue player" }, { "id": "arxiv-2601.15047", "name": "Game-Theoretic Lens on LLM-based Multi-Agent Systems", "title": "Game-Theoretic Lens on LLM-based Multi-Agent Systems", "date": "2026-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Survey" ], "primary_category": null, "keywords": [], "keywords_by_category": { "Survey": "" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.15047" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "survey", "source_rows": [ { "section": "Survey", "line": 373, "date": "2026-01", "title": "Game-Theoretic Lens on LLM-based Multi-Agent Systems", "venue": "arXiv", "keywords": "", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.15047" } ] } ], "sources": [ "https://arxiv.org/abs/2601.15047" ], "search_text": "game-theoretic lens on llm-based multi-agent systems game-theoretic lens on llm-based multi-agent systems 2026-01 arxiv survey survey" }, { "id": "arxiv-2601.16690", "name": "EMemBench", "title": "EMemBench: Interactive Benchmarking of Episodic Memory for VLM Agents", "date": "2026-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Episodic Memory", "VLM Agents", "Interactive Games", "Text+Visual" ], "keywords_by_category": { "Benchmark": "Episodic Memory, VLM Agents, Interactive Games, Text+Visual" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.16690" }, { "label": "Code", "url": "https://github.com/InternLM/EMemBench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 295, "date": "2026-01", "title": "EMemBench: Interactive Benchmarking of Episodic Memory for VLM Agents", "venue": "arXiv", "keywords": "Episodic Memory, VLM Agents, Interactive Games, Text+Visual", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2601.16690" }, { "label": "Code", "url": "https://github.com/InternLM/EMemBench" } ] } ], "sources": [ "https://arxiv.org/abs/2601.16690", "https://github.com/InternLM/EMemBench" ], "search_text": "emembench: interactive benchmarking of episodic memory for vlm agents emembench 2026-01 arxiv benchmark episodic memory vlm agents interactive games text+visual player" }, { "id": "arxiv-2602.13214", "name": "BotzoneBench", "title": "BotzoneBench: Scalable LLM Evaluation via Graded AI Anchors", "date": "2026-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Harness", "Benchmark" ], "primary_category": null, "keywords": [ "Textual Environment", "AI Anchors", "Scalable Evaluation", "Strategic Reasoning", "8 Games", "Elo Rating" ], "keywords_by_category": { "Harness": "Textual Environment", "Benchmark": "AI Anchors, Scalable Evaluation, Strategic Reasoning, 8 Games, Elo Rating" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.13214" }, { "label": "Code", "url": "https://github.com/AMysteriousBeing/BotzoneBench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 194, "date": "2026-01", "title": "BotzoneBench: Scalable LLM Evaluation via Graded AI Anchors", "venue": "arXiv", "keywords": "Textual Environment", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.13214" }, { "label": "Code", "url": "https://github.com/AMysteriousBeing/BotzoneBench" } ] }, { "section": "Benchmark", "line": 300, "date": "2026-01", "title": "BotzoneBench: Scalable LLM Evaluation via Graded AI Anchors", "venue": "arXiv", "keywords": "AI Anchors, Scalable Evaluation, Strategic Reasoning, 8 Games, Elo Rating", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2602.13214" } ] } ], "sources": [ "https://arxiv.org/abs/2602.13214", "https://github.com/AMysteriousBeing/BotzoneBench" ], "search_text": "botzonebench: scalable llm evaluation via graded ai anchors botzonebench 2026-01 arxiv harness benchmark textual environment ai anchors scalable evaluation strategic reasoning 8 games elo rating player" }, { "id": "arxiv-2512.09187", "name": "WOLF", "title": "WOLF: Werewolf-based Observations for LLM Deception and Falsehoods", "date": "2025-12", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Werewolf", "Deception Detection", "Multi-agent", "Social Deduction", "Longitudinal Eval" ], "keywords_by_category": { "Benchmark": "Werewolf, Deception Detection, Multi-agent, Social Deduction, Longitudinal Eval" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2512.09187" }, { "label": "Code", "url": "https://github.com/MrinalA2009/WOLF-Werewolf-based-Observations-for-LLM-Deception-and-Falsehoods" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 302, "date": "2025-12", "title": "WOLF: Werewolf-based Observations for LLM Deception and Falsehoods", "venue": "arXiv", "keywords": "Werewolf, Deception Detection, Multi-agent, Social Deduction, Longitudinal Eval", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2512.09187" }, { "label": "Code", "url": "https://github.com/MrinalA2009/WOLF-Werewolf-based-Observations-for-LLM-Deception-and-Falsehoods" } ] } ], "sources": [ "https://arxiv.org/abs/2512.09187", "https://github.com/MrinalA2009/WOLF-Werewolf-based-Observations-for-LLM-Deception-and-Falsehoods" ], "search_text": "wolf: werewolf-based observations for llm deception and falsehoods wolf 2025-12 arxiv benchmark werewolf deception detection multi-agent social deduction longitudinal eval player" }, { "id": "arxiv-2512.09706", "name": "Training One Model to Master Cross-Level Agentic Actions via Reinforcement Learning", "title": "Training One Model to Master Cross-Level Agentic Actions via Reinforcement Learning", "date": "2025-12", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "SFT + GRPO", "Cross-Level Action Space" ], "keywords_by_category": { "Model": "SFT + GRPO, Cross-Level Action Space" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2512.09706" }, { "label": "Code", "url": "https://github.com/CraftJarvis/OpenHA" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 118, "date": "2025-12", "title": "Training One Model to Master Cross-Level Agentic Actions via Reinforcement Learning", "venue": "arXiv", "keywords": "SFT + GRPO, Cross-Level Action Space", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2512.09706" }, { "label": "Code", "url": "https://github.com/CraftJarvis/OpenHA" } ] } ], "sources": [ "https://arxiv.org/abs/2512.09706", "https://github.com/CraftJarvis/OpenHA" ], "search_text": "training one model to master cross-level agentic actions via reinforcement learning training one model to master cross-level agentic actions via reinforcement learning 2025-12 arxiv model sft + grpo cross-level action space player" }, { "id": "arxiv-2512.04797", "name": "SIMA 2", "title": "SIMA 2: A Generalist Embodied Agent for Virtual Worlds", "date": "2025-12", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Gemini Fine-tuning", "Cross-World Transfer" ], "keywords_by_category": { "Model": "Gemini Fine-tuning, Cross-World Transfer" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2512.04797" }, { "label": "Project", "url": "https://deepmind.google/discover/blog/sima-2-an-agent-that-plays-reasons-and-learns-with-you-in-virtual-3d-worlds/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 119, "date": "2025-12", "title": "SIMA 2: A Generalist Embodied Agent for Virtual Worlds", "venue": "arXiv", "keywords": "Gemini Fine-tuning, Cross-World Transfer", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2512.04797" }, { "label": "Project", "url": "https://deepmind.google/discover/blog/sima-2-an-agent-that-plays-reasons-and-learns-with-you-in-virtual-3d-worlds/" } ] } ], "sources": [ "https://arxiv.org/abs/2512.04797", "https://deepmind.google/discover/blog/sima-2-an-agent-that-plays-reasons-and-learns-with-you-in-virtual-3d-worlds/" ], "search_text": "sima 2: a generalist embodied agent for virtual worlds sima 2 2025-12 arxiv model gemini fine-tuning cross-world transfer player" }, { "id": "arxiv-2512.01992", "name": "LLM CHESS", "title": "LLM CHESS: Benchmarking Reasoning and Instruction-Following in LLMs through Chess", "date": "2025-12", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Chess", "Reasoning", "Instruction-following", "Agentic Evaluation" ], "keywords_by_category": { "Benchmark": "Chess, Reasoning, Instruction-following, Agentic Evaluation" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2512.01992" }, { "label": "Code", "url": "https://github.com/LLM-CHESS/llm_chess" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 301, "date": "2025-12", "title": "LLM CHESS: Benchmarking Reasoning and Instruction-Following in LLMs through Chess", "venue": "arXiv", "keywords": "Chess, Reasoning, Instruction-following, Agentic Evaluation", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2512.01992" }, { "label": "Code", "url": "https://github.com/LLM-CHESS/llm_chess" } ] } ], "sources": [ "https://arxiv.org/abs/2512.01992", "https://github.com/LLM-CHESS/llm_chess" ], "search_text": "llm chess: benchmarking reasoning and instruction-following in llms through chess llm chess 2025-12 arxiv benchmark chess reasoning instruction-following agentic evaluation player" }, { "id": "work-4c743594e0f0", "name": "Do Persona-Infused LLMs Affect Performance in a Strategic Reasoning Game?", "title": "Do Persona-Infused LLMs Affect Performance in a Strategic Reasoning Game?", "date": "2025-12", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "IJCNLP-AACL 2025", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "Persona Prompting" ], "keywords_by_category": { "Harness": "Persona Prompting" }, "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.ijcnlp-long.186/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 195, "date": "2025-12", "title": "Do Persona-Infused LLMs Affect Performance in a Strategic Reasoning Game?", "venue": "IJCNLP-AACL 2025", "keywords": "Persona Prompting", "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.ijcnlp-long.186/" } ] } ], "sources": [ "https://aclanthology.org/2025.ijcnlp-long.186/" ], "search_text": "do persona-infused llms affect performance in a strategic reasoning game? do persona-infused llms affect performance in a strategic reasoning game? 2025-12 ijcnlp-aacl 2025 harness persona prompting player" }, { "id": "arxiv-2512.11362", "name": "An Anatomy of Vision-Language-Action Models", "title": "An Anatomy of Vision-Language-Action Models: From Modules to Milestones and Challenges", "date": "2025-12", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Survey" ], "primary_category": null, "keywords": [], "keywords_by_category": { "Survey": "" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2512.11362" }, { "label": "Project", "url": "https://suyuz1.github.io/VLA-Survey-Anatomy/" }, { "label": "GitHub", "url": "https://github.com/SuyuZ1/VLA-Survey-Anatomy" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "survey", "source_rows": [ { "section": "Survey", "line": 374, "date": "2025-12", "title": "An Anatomy of Vision-Language-Action Models: From Modules to Milestones and Challenges", "venue": "arXiv", "keywords": "", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2512.11362" }, { "label": "Project", "url": "https://suyuz1.github.io/VLA-Survey-Anatomy/" }, { "label": "GitHub", "url": "https://github.com/SuyuZ1/VLA-Survey-Anatomy" } ] } ], "sources": [ "https://arxiv.org/abs/2512.11362", "https://suyuz1.github.io/VLA-Survey-Anatomy/", "https://github.com/SuyuZ1/VLA-Survey-Anatomy" ], "search_text": "an anatomy of vision-language-action models: from modules to milestones and challenges an anatomy of vision-language-action models 2025-12 arxiv survey survey" }, { "id": "arxiv-2511.15065", "name": "Reasoning via Video", "title": "Reasoning via Video: The First Evaluation of Video Models' Reasoning Abilities through Maze-Solving Tasks", "date": "2025-11", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Maze Reasoning", "Spatial Planning", "Video Models" ], "keywords_by_category": { "Benchmark": "Maze Reasoning, Spatial Planning, Video Models" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2511.15065" }, { "label": "Project", "url": "https://imyangc7.github.io/VRBench_Web/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 303, "date": "2025-11", "title": "Reasoning via Video: The First Evaluation of Video Models' Reasoning Abilities through Maze-Solving Tasks", "venue": "arXiv", "keywords": "Maze Reasoning, Spatial Planning, Video Models", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2511.15065" }, { "label": "Project", "url": "https://imyangc7.github.io/VRBench_Web/" } ] } ], "sources": [ "https://arxiv.org/abs/2511.15065", "https://imyangc7.github.io/VRBench_Web/" ], "search_text": "reasoning via video: the first evaluation of video models' reasoning abilities through maze-solving tasks reasoning via video 2025-11 arxiv benchmark maze reasoning spatial planning video models player" }, { "id": "arxiv-2511.04898", "name": "Real-Time Reasoning Agents in Evolving Environments", "title": "Real-Time Reasoning Agents in Evolving Environments", "date": "2025-11", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "Real-Time Reasoning", "Multi-Agent" ], "keywords_by_category": { "Harness": "Real-Time Reasoning, Multi-Agent" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2511.04898" }, { "label": "Code", "url": "https://github.com/SALT-NLP/RealtimeGym" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 196, "date": "2025-11", "title": "Real-Time Reasoning Agents in Evolving Environments", "venue": "arXiv", "keywords": "Real-Time Reasoning, Multi-Agent", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2511.04898" }, { "label": "Code", "url": "https://github.com/SALT-NLP/RealtimeGym" } ] } ], "sources": [ "https://arxiv.org/abs/2511.04898", "https://github.com/SALT-NLP/RealtimeGym" ], "search_text": "real-time reasoning agents in evolving environments real-time reasoning agents in evolving environments 2025-11 arxiv harness real-time reasoning multi-agent player" }, { "id": "arxiv-2511.09057", "name": "PAN", "title": "PAN: A World Model for General, Interactable, and Long-Horizon World Simulation", "date": "2025-11", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "General World Model", "Language-Conditioned Actions", "Long-Horizon Simulation" ], "keywords_by_category": { "Model": "General World Model, Language-Conditioned Actions, Long-Horizon Simulation" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2511.09057" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 121, "date": "2025-11", "title": "PAN: A World Model for General, Interactable, and Long-Horizon World Simulation", "venue": "arXiv", "keywords": "General World Model, Language-Conditioned Actions, Long-Horizon Simulation", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2511.09057" } ] } ], "sources": [ "https://arxiv.org/abs/2511.09057" ], "search_text": "pan: a world model for general, interactable, and long-horizon world simulation pan 2025-11 arxiv model general world model language-conditioned actions long-horizon simulation supporting environment" }, { "id": "arxiv-2511.08892", "name": "Lumine", "title": "Lumine: An Open Recipe for Building Generalist Agents in 3D Open Worlds", "date": "2025-11", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "SFT", "Cross-Game Transfer", "Genshin Impact" ], "keywords_by_category": { "Model": "SFT, Cross-Game Transfer, Genshin Impact" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2511.08892" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 122, "date": "2025-11", "title": "Lumine: An Open Recipe for Building Generalist Agents in 3D Open Worlds", "venue": "arXiv", "keywords": "SFT, Cross-Game Transfer, Genshin Impact", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2511.08892" } ] } ], "sources": [ "https://arxiv.org/abs/2511.08892" ], "search_text": "lumine: an open recipe for building generalist agents in 3d open worlds lumine 2025-11 arxiv model sft cross-game transfer genshin impact player" }, { "id": "arxiv-2511.15407", "name": "IPR-1", "title": "IPR-1: Interactive Physical Reasoner", "date": "2025-11", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Prediction-Reinforced Reasoning", "PhysCode", "Game-to-Unseen Benchmark" ], "keywords_by_category": { "Model": "Prediction-Reinforced Reasoning, PhysCode, Game-to-Unseen Benchmark" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2511.15407" }, { "label": "Code", "url": "https://github.com/mybearyZhang/ipr-1" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 120, "date": "2025-11", "title": "IPR-1: Interactive Physical Reasoner", "venue": "arXiv", "keywords": "Prediction-Reinforced Reasoning, PhysCode, Game-to-Unseen Benchmark", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2511.15407" }, { "label": "Code", "url": "https://github.com/mybearyZhang/ipr-1" } ] } ], "sources": [ "https://arxiv.org/abs/2511.15407", "https://github.com/mybearyZhang/ipr-1" ], "search_text": "ipr-1: interactive physical reasoner ipr-1 2025-11 arxiv model prediction-reinforced reasoning physcode game-to-unseen benchmark player" }, { "id": "work-63329b47e5f4", "name": "Stronger-MAS", "title": "Stronger-MAS: Multi-Agent Reinforcement Learning for Collaborative LLMs", "date": "2025-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Multi-Agent GRPO", "Collaborative RL" ], "keywords_by_category": { "Model": "Multi-Agent GRPO, Collaborative RL" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=IdF6JqXWzx" }, { "label": "Code", "url": "https://github.com/pettingllms-ai/PettingLLMs" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 125, "date": "2025-10", "title": "Stronger-MAS: Multi-Agent Reinforcement Learning for Collaborative LLMs", "venue": "arXiv", "keywords": "Multi-Agent GRPO, Collaborative RL", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=IdF6JqXWzx" }, { "label": "Code", "url": "https://github.com/pettingllms-ai/PettingLLMs" } ] } ], "sources": [ "https://openreview.net/forum?id=IdF6JqXWzx", "https://github.com/pettingllms-ai/PettingLLMs" ], "search_text": "stronger-mas: multi-agent reinforcement learning for collaborative llms stronger-mas 2025-10 arxiv model multi-agent grpo collaborative rl player" }, { "id": "arxiv-2510.18483", "name": "StarBench", "title": "StarBench: A Turn-Based RPG Benchmark for Agentic Multimodal Decision-Making and Information Seeking", "date": "2025-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "VLMs", "Multimodal Benchmark", "Decision-making", "UI grounding" ], "keywords_by_category": { "Benchmark": "VLMs, Multimodal Benchmark, Decision-making, UI grounding" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2510.18483" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 304, "date": "2025-10", "title": "StarBench: A Turn-Based RPG Benchmark for Agentic Multimodal Decision-Making and Information Seeking", "venue": "arXiv", "keywords": "VLMs, Multimodal Benchmark, Decision-making, UI grounding", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2510.18483" } ] } ], "sources": [ "https://arxiv.org/abs/2510.18483" ], "search_text": "starbench: a turn-based rpg benchmark for agentic multimodal decision-making and information seeking starbench 2025-10 arxiv benchmark vlms multimodal benchmark decision-making ui grounding player" }, { "id": "arxiv-2510.06475", "name": "PuzzlePlex", "title": "PuzzlePlex: Benchmarking Foundation Models on Reasoning and Planning with Puzzles", "date": "2025-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Puzzle Games", "15 Types", "Deterministic & Stochastic", "Single & Two-player" ], "keywords_by_category": { "Benchmark": "Puzzle Games, 15 Types, Deterministic & Stochastic, Single & Two-player" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2510.06475v1" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 308, "date": "2025-10", "title": "PuzzlePlex: Benchmarking Foundation Models on Reasoning and Planning with Puzzles", "venue": "arXiv", "keywords": "Puzzle Games, 15 Types, Deterministic & Stochastic, Single & Two-player", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2510.06475v1" } ] } ], "sources": [ "https://arxiv.org/abs/2510.06475v1" ], "search_text": "puzzleplex: benchmarking foundation models on reasoning and planning with puzzles puzzleplex 2025-10 arxiv benchmark puzzle games 15 types deterministic & stochastic single & two-player player" }, { "id": "work-cba44458bed9", "name": "Mixing Expert Knowledge", "title": "Mixing Expert Knowledge: Bring Human Thoughts Back To the Game of Go", "date": "2025-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NeurIPS 2025", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Go", "Expert-Level", "Mixed Fine-tuning", "SFT+RL", "Human-AI Alignment" ], "keywords_by_category": { "Benchmark": "Go, Expert-Level, Mixed Fine-tuning, SFT+RL, Human-AI Alignment" }, "links": [ { "label": "Paper", "url": "https://neurips.cc/virtual/2025/loc/san-diego/poster/117166" }, { "label": "Code", "url": "https://github.com/Entarochuan/InternGo-LoGos" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 309, "date": "2025-10", "title": "Mixing Expert Knowledge: Bring Human Thoughts Back To the Game of Go", "venue": "NeurIPS 2025", "keywords": "Go, Expert-Level, Mixed Fine-tuning, SFT+RL, Human-AI Alignment", "links": [ { "label": "Paper", "url": "https://neurips.cc/virtual/2025/loc/san-diego/poster/117166" }, { "label": "Code", "url": "https://github.com/Entarochuan/InternGo-LoGos" } ] } ], "sources": [ "https://neurips.cc/virtual/2025/loc/san-diego/poster/117166", "https://github.com/Entarochuan/InternGo-LoGos" ], "search_text": "mixing expert knowledge: bring human thoughts back to the game of go mixing expert knowledge 2025-10 neurips 2025 benchmark go expert-level mixed fine-tuning sft+rl human-ai alignment player" }, { "id": "arxiv-2510.15414", "name": "MARSHAL", "title": "MARSHAL: Incentivizing Multi-Agent Reasoning via Self-Play with Strategic LLMs", "date": "2025-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Self-Play RL", "Strategic Transfer" ], "keywords_by_category": { "Model": "Self-Play RL, Strategic Transfer" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2510.15414" }, { "label": "Code", "url": "https://github.com/thu-nics/MARSHAL" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 126, "date": "2025-10", "title": "MARSHAL: Incentivizing Multi-Agent Reasoning via Self-Play with Strategic LLMs", "venue": "arXiv", "keywords": "Self-Play RL, Strategic Transfer", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2510.15414" }, { "label": "Code", "url": "https://github.com/thu-nics/MARSHAL" } ] } ], "sources": [ "https://arxiv.org/abs/2510.15414", "https://github.com/thu-nics/MARSHAL" ], "search_text": "marshal: incentivizing multi-agent reasoning via self-play with strategic llms marshal 2025-10 arxiv model self-play rl strategic transfer player" }, { "id": "arxiv-2510.04980", "name": "LLM-Hanabi", "title": "LLM-Hanabi: Evaluating Multi-Agent Gameplays with Theory-of-Mind and Rationale Inference in Imperfect Information Collaboration Game", "date": "2025-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "EMNLP 2025 workshop", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Hanabi", "Theory-of-Mind", "Rationale Inference", "Cooperative", "Imperfect Information" ], "keywords_by_category": { "Benchmark": "Hanabi, Theory-of-Mind, Rationale Inference, Cooperative, Imperfect Information" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2510.04980" }, { "label": "Code", "url": "https://github.com/HKUST-KnowComp/LLM-Hanabi" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 305, "date": "2025-10", "title": "LLM-Hanabi: Evaluating Multi-Agent Gameplays with Theory-of-Mind and Rationale Inference in Imperfect Information Collaboration Game", "venue": "EMNLP 2025 workshop", "keywords": "Hanabi, Theory-of-Mind, Rationale Inference, Cooperative, Imperfect Information", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2510.04980" }, { "label": "Code", "url": "https://github.com/HKUST-KnowComp/LLM-Hanabi" } ] } ], "sources": [ "https://arxiv.org/abs/2510.04980", "https://github.com/HKUST-KnowComp/LLM-Hanabi" ], "search_text": "llm-hanabi: evaluating multi-agent gameplays with theory-of-mind and rationale inference in imperfect information collaboration game llm-hanabi 2025-10 emnlp 2025 workshop benchmark hanabi theory-of-mind rationale inference cooperative imperfect information player" }, { "id": "arxiv-2510.16774", "name": "Learning to Play", "title": "Learning to Play: A Multimodal Agent for 3D Game-Play", "date": "2025-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "VLA", "Behavior Cloning", "3D Game-Play" ], "keywords_by_category": { "Model": "VLA, Behavior Cloning, 3D Game-Play" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2510.16774" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 124, "date": "2025-10", "title": "Learning to Play: A Multimodal Agent for 3D Game-Play", "venue": "arXiv", "keywords": "VLA, Behavior Cloning, 3D Game-Play", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2510.16774" } ] } ], "sources": [ "https://arxiv.org/abs/2510.16774" ], "search_text": "learning to play: a multimodal agent for 3d game-play learning to play 2025-10 arxiv model vla behavior cloning 3d game-play player" }, { "id": "arxiv-2510.23691", "name": "Game-TARS", "title": "Game-TARS: Pretrained Foundation Models for Scalable Generalist Multimodal Game Agents", "date": "2025-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Pre-training", "Unified Action Space" ], "keywords_by_category": { "Model": "Pre-training, Unified Action Space" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2510.23691" }, { "label": "Project", "url": "https://seed-tars.com/game-tars/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 123, "date": "2025-10", "title": "Game-TARS: Pretrained Foundation Models for Scalable Generalist Multimodal Game Agents", "venue": "arXiv", "keywords": "Pre-training, Unified Action Space", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2510.23691" }, { "label": "Project", "url": "https://seed-tars.com/game-tars/" } ] } ], "sources": [ "https://arxiv.org/abs/2510.23691", "https://seed-tars.com/game-tars/" ], "search_text": "game-tars: pretrained foundation models for scalable generalist multimodal game agents game-tars 2025-10 arxiv model pre-training unified action space player" }, { "id": "work-d2dce667282a", "name": "Code World Models for General Game Playing", "title": "Code World Models for General Game Playing", "date": "2025-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2016", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "LLM codes modules for game playing" ], "keywords_by_category": { "Harness": "LLM codes modules for game playing" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=1UoB7IWiku" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 197, "date": "2025-10", "title": "Code World Models for General Game Playing", "venue": "ICLR 2016", "keywords": "LLM codes modules for game playing", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=1UoB7IWiku" } ] } ], "sources": [ "https://openreview.net/forum?id=1UoB7IWiku" ], "search_text": "code world models for general game playing code world models for general game playing 2025-10 iclr 2016 harness llm codes modules for game playing player" }, { "id": "arxiv-2510.26852", "name": "CATArena", "title": "CATArena: Evaluation of LLM Agents through Iterative Tournament Competitions", "date": "2025-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Tournament Evaluation", "Peer Learning", "Board Games", "Card Games" ], "keywords_by_category": { "Benchmark": "Tournament Evaluation, Peer Learning, Board Games, Card Games" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2510.26852v1" }, { "label": "Code", "url": "https://github.com/AGI-Eval-Official/CATArena" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 307, "date": "2025-10", "title": "CATArena: Evaluation of LLM Agents through Iterative Tournament Competitions", "venue": "arXiv", "keywords": "Tournament Evaluation, Peer Learning, Board Games, Card Games", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2510.26852v1" }, { "label": "Code", "url": "https://github.com/AGI-Eval-Official/CATArena" } ] } ], "sources": [ "https://arxiv.org/abs/2510.26852v1", "https://github.com/AGI-Eval-Official/CATArena" ], "search_text": "catarena: evaluation of llm agents through iterative tournament competitions catarena 2025-10 arxiv benchmark tournament evaluation peer learning board games card games player" }, { "id": "arxiv-2510.11389", "name": "Beyond Survival", "title": "Beyond Survival: Evaluating LLMs in Social Deduction Games with Human-Aligned Strategies", "date": "2025-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Werewolf", "Human-Aligned Evaluation", "Strategy Alignment", "100+ Hours Video" ], "keywords_by_category": { "Benchmark": "Werewolf, Human-Aligned Evaluation, Strategy Alignment, 100+ Hours Video" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2510.11389" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 306, "date": "2025-10", "title": "Beyond Survival: Evaluating LLMs in Social Deduction Games with Human-Aligned Strategies", "venue": "arXiv", "keywords": "Werewolf, Human-Aligned Evaluation, Strategy Alignment, 100+ Hours Video", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2510.11389" } ] } ], "sources": [ "https://arxiv.org/abs/2510.11389" ], "search_text": "beyond survival: evaluating llms in social deduction games with human-aligned strategies beyond survival 2025-10 arxiv benchmark werewolf human-aligned evaluation strategy alignment 100+ hours video player" }, { "id": "arxiv-2509.02544", "name": "UI-TARS-2 Technical Report", "title": "UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning", "date": "2025-09", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Multi-Turn RL", "GUI Agent" ], "keywords_by_category": { "Model": "Multi-Turn RL, GUI Agent" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2509.02544" }, { "label": "Code", "url": "https://github.com/bytedance/UI-TARS" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 129, "date": "2025-09", "title": "UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning", "venue": "arXiv", "keywords": "Multi-Turn RL, GUI Agent", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2509.02544" }, { "label": "Code", "url": "https://github.com/bytedance/UI-TARS" } ] } ], "sources": [ "https://arxiv.org/abs/2509.02544", "https://github.com/bytedance/UI-TARS" ], "search_text": "ui-tars-2 technical report: advancing gui agent with multi-turn reinforcement learning ui-tars-2 technical report 2025-09 arxiv model multi-turn rl gui agent player" }, { "id": "arxiv-2509.22387", "name": "SpinGPT", "title": "SpinGPT: A Large-Language-Model Approach to Playing Poker Correctly", "date": "2025-09", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ACG 2025", "categories": [ "Model" ], "primary_category": null, "keywords": [ "SFT + RL", "Poker" ], "keywords_by_category": { "Model": "SFT + RL, Poker" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2509.22387" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 128, "date": "2025-09", "title": "SpinGPT: A Large-Language-Model Approach to Playing Poker Correctly", "venue": "ACG 2025", "keywords": "SFT + RL, Poker", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2509.22387" } ] } ], "sources": [ "https://arxiv.org/abs/2509.22387" ], "search_text": "spingpt: a large-language-model approach to playing poker correctly spingpt 2025-09 acg 2025 model sft + rl poker player" }, { "id": "work-b49843a0df70", "name": "PillagerBench", "title": "PillagerBench: Benchmarking LLM-Based Agents in Competitive Minecraft Team Environments", "date": "2025-09", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "CoG 2025", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Minecraft", "Multi-agent Systems", "Competitive Team Environment" ], "keywords_by_category": { "Benchmark": "Minecraft, Multi-agent Systems, Competitive Team Environment" }, "links": [ { "label": "Paper", "url": "https://ieeexplore.ieee.org/document/11114387" }, { "label": "Code", "url": "https://github.com/aialt/PillagerBench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 310, "date": "2025-09", "title": "PillagerBench: Benchmarking LLM-Based Agents in Competitive Minecraft Team Environments", "venue": "CoG 2025", "keywords": "Minecraft, Multi-agent Systems, Competitive Team Environment", "links": [ { "label": "Paper", "url": "https://ieeexplore.ieee.org/document/11114387" }, { "label": "Code", "url": "https://github.com/aialt/PillagerBench" } ] } ], "sources": [ "https://ieeexplore.ieee.org/document/11114387", "https://github.com/aialt/PillagerBench" ], "search_text": "pillagerbench: benchmarking llm-based agents in competitive minecraft team environments pillagerbench 2025-09 cog 2025 benchmark minecraft multi-agent systems competitive team environment player" }, { "id": "arxiv-2509.13347", "name": "OpenHA", "title": "OpenHA: A Series of Open-Source Hierarchical Agentic Models in Minecraft", "date": "2025-09", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Dataset", "Model" ], "primary_category": null, "keywords": [ "~4.2B Tokens", "5 Hierarchical Action Datasets", "800+ Tasks", "SFT", "Chain of Action", "Hierarchical VLA" ], "keywords_by_category": { "Dataset": "~4.2B Tokens, 5 Hierarchical Action Datasets, 800+ Tasks", "Model": "SFT, Chain of Action, Hierarchical VLA" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2509.13347" }, { "label": "Code", "url": "https://github.com/CraftJarvis/OpenHA" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 127, "date": "2025-09", "title": "OpenHA: A Series of Open-Source Hierarchical Agentic Models in Minecraft", "venue": "arXiv", "keywords": "SFT, Chain of Action, Hierarchical VLA", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2509.13347" }, { "label": "Code", "url": "https://github.com/CraftJarvis/OpenHA" } ] }, { "section": "Dataset", "line": 247, "date": "2025-09", "title": "OpenHA: A Series of Open-Source Hierarchical Agentic Models in Minecraft", "venue": "arXiv", "keywords": "~4.2B Tokens, 5 Hierarchical Action Datasets, 800+ Tasks", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2509.13347" }, { "label": "Code", "url": "https://github.com/CraftJarvis/OpenHA" } ] } ], "sources": [ "https://arxiv.org/abs/2509.13347", "https://github.com/CraftJarvis/OpenHA" ], "search_text": "openha: a series of open-source hierarchical agentic models in minecraft openha 2025-09 arxiv dataset model ~4.2b tokens 5 hierarchical action datasets 800+ tasks sft chain of action hierarchical vla player" }, { "id": "arxiv-2509.22170", "name": "Leveraging LLM Agents for Automated Video Game Testing", "title": "Leveraging LLM Agents for Automated Video Game Testing", "date": "2025-09", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "Textual environment" ], "keywords_by_category": { "Harness": "Textual environment" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2509.22170" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 198, "date": "2025-09", "title": "Leveraging LLM Agents for Automated Video Game Testing", "venue": "arXiv", "keywords": "Textual environment", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2509.22170" } ] } ], "sources": [ "https://arxiv.org/abs/2509.22170" ], "search_text": "leveraging llm agents for automated video game testing leveraging llm agents for automated video game testing 2025-09 arxiv harness textual environment player" }, { "id": "work-4de219360365", "name": "FlashAdventure", "title": "FlashAdventure: A Benchmark for GUI Agents Solving Full Story Arcs in Diverse Adventure Games", "date": "2025-09", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "EMNLP 2025", "categories": [ "Harness", "Benchmark" ], "primary_category": null, "keywords": [ "Memory", "Subtask", "GUI Agents", "Adventure Games", "Full Story Arcs", "34 Flash Games" ], "keywords_by_category": { "Harness": "Memory, Subtask", "Benchmark": "GUI Agents, Adventure Games, Full Story Arcs, 34 Flash Games" }, "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.emnlp-main.1192/" }, { "label": "Code", "url": "https://github.com/ahnjaewoo/FlashAdventure" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 199, "date": "2025-09", "title": "FlashAdventure: A Benchmark for GUI Agents Solving Full Story Arcs in Diverse Adventure Games", "venue": "EMNLP 2025", "keywords": "Memory, Subtask", "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.emnlp-main.1192/" }, { "label": "Code", "url": "https://github.com/ahnjaewoo/FlashAdventure" } ] }, { "section": "Benchmark", "line": 312, "date": "2025-09", "title": "FlashAdventure: A Benchmark for GUI Agents Solving Full Story Arcs in Diverse Adventure Games", "venue": "EMNLP 2025", "keywords": "GUI Agents, Adventure Games, Full Story Arcs, 34 Flash Games", "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.emnlp-main.1192/" }, { "label": "Code", "url": "https://github.com/ahnjaewoo/FlashAdventure" } ] } ], "sources": [ "https://aclanthology.org/2025.emnlp-main.1192/", "https://github.com/ahnjaewoo/FlashAdventure" ], "search_text": "flashadventure: a benchmark for gui agents solving full story arcs in diverse adventure games flashadventure 2025-09 emnlp 2025 harness benchmark memory subtask gui agents adventure games full story arcs 34 flash games player" }, { "id": "work-bd8ae4a59073", "name": "EvoEmpirBench", "title": "EvoEmpirBench: Dynamic Spatial Reasoning with Agent-ExpVer", "date": "2025-09", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "AAAI 2026", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Dynamic Spatial Reasoning", "Maze Navigation", "Agent-ExpVer" ], "keywords_by_category": { "Benchmark": "Dynamic Spatial Reasoning, Maze Navigation, Agent-ExpVer" }, "links": [ { "label": "Paper", "url": "https://ojs.aaai.org/index.php/AAAI/article/view/40979" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 313, "date": "2025-09", "title": "EvoEmpirBench: Dynamic Spatial Reasoning with Agent-ExpVer", "venue": "AAAI 2026", "keywords": "Dynamic Spatial Reasoning, Maze Navigation, Agent-ExpVer", "links": [ { "label": "Paper", "url": "https://ojs.aaai.org/index.php/AAAI/article/view/40979" } ] } ], "sources": [ "https://ojs.aaai.org/index.php/AAAI/article/view/40979" ], "search_text": "evoempirbench: dynamic spatial reasoning with agent-expver evoempirbench 2025-09 aaai 2026 benchmark dynamic spatial reasoning maze navigation agent-expver player" }, { "id": "work-8c8b518d8193", "name": "Can Large Language Models Master Complex Card Games?", "title": "Can Large Language Models Master Complex Card Games?", "date": "2025-09", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NeurIPS 2025", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Card Games", "Fine-tuning", "Multi-game Learning", "Knowledge Retention" ], "keywords_by_category": { "Benchmark": "Card Games, Fine-tuning, Multi-game Learning, Knowledge Retention" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=cmN8Wbvanr" }, { "label": "Code", "url": "https://github.com/THUDM/LLM4CardGame" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 311, "date": "2025-09", "title": "Can Large Language Models Master Complex Card Games?", "venue": "NeurIPS 2025", "keywords": "Card Games, Fine-tuning, Multi-game Learning, Knowledge Retention", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=cmN8Wbvanr" }, { "label": "Code", "url": "https://github.com/THUDM/LLM4CardGame" } ] } ], "sources": [ "https://openreview.net/forum?id=cmN8Wbvanr", "https://github.com/THUDM/LLM4CardGame" ], "search_text": "can large language models master complex card games? can large language models master complex card games? 2025-09 neurips 2025 benchmark card games fine-tuning multi-game learning knowledge retention player" }, { "id": "arxiv-2508.04720", "name": "Who is a Better Player", "title": "Who is a Better Player: LLM against LLM", "date": "2025-08", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Round-robin Tournament", "Elo Rating", "Adversarial Board Games" ], "keywords_by_category": { "Benchmark": "Round-robin Tournament, Elo Rating, Adversarial Board Games" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.04720v1" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 317, "date": "2025-08", "title": "Who is a Better Player: LLM against LLM", "venue": "arXiv", "keywords": "Round-robin Tournament, Elo Rating, Adversarial Board Games", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.04720v1" } ] } ], "sources": [ "https://arxiv.org/abs/2508.04720v1" ], "search_text": "who is a better player: llm against llm who is a better player 2025-08 arxiv benchmark round-robin tournament elo rating adversarial board games player" }, { "id": "work-e7c8b018cd93", "name": "VistaWise", "title": "VistaWise: Building Cost-Effective Agent with Cross-Modal Knowledge Graph for Minecraft", "date": "2025-08", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "EMNLP 2025", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "Cross-modal RAG", "Memory", "Skill Library" ], "keywords_by_category": { "Harness": "Cross-modal RAG, Memory, Skill Library" }, "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.emnlp-main.1111/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 201, "date": "2025-08", "title": "VistaWise: Building Cost-Effective Agent with Cross-Modal Knowledge Graph for Minecraft", "venue": "EMNLP 2025", "keywords": "Cross-modal RAG, Memory, Skill Library", "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.emnlp-main.1111/" } ] } ], "sources": [ "https://aclanthology.org/2025.emnlp-main.1111/" ], "search_text": "vistawise: building cost-effective agent with cross-modal knowledge graph for minecraft vistawise 2025-08 emnlp 2025 harness cross-modal rag memory skill library player" }, { "id": "arxiv-2508.21365", "name": "Think in Games", "title": "Think in Games: Learning to Reason in Games via Reinforcement Learning with Large Language Models", "date": "2025-08", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "RL Training", "Game Reasoning" ], "keywords_by_category": { "Model": "RL Training, Game Reasoning" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.21365" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 131, "date": "2025-08", "title": "Think in Games: Learning to Reason in Games via Reinforcement Learning with Large Language Models", "venue": "arXiv", "keywords": "RL Training, Game Reasoning", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.21365" } ] } ], "sources": [ "https://arxiv.org/abs/2508.21365" ], "search_text": "think in games: learning to reason in games via reinforcement learning with large language models think in games 2025-08 arxiv model rl training game reasoning player" }, { "id": "arxiv-2508.10839", "name": "Reinforced Language Models for Sequential Decision Making", "title": "Reinforced Language Models for Sequential Decision Making", "date": "2025-08", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Multi-step GRPO", "Credit Assignment" ], "keywords_by_category": { "Model": "Multi-step GRPO, Credit Assignment" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.10839" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 132, "date": "2025-08", "title": "Reinforced Language Models for Sequential Decision Making", "venue": "arXiv", "keywords": "Multi-step GRPO, Credit Assignment", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.10839" } ] } ], "sources": [ "https://arxiv.org/abs/2508.10839" ], "search_text": "reinforced language models for sequential decision making reinforced language models for sequential decision making 2025-08 arxiv model multi-step grpo credit assignment player" }, { "id": "arxiv-2508.16821", "name": "PuzzleJAX", "title": "PuzzleJAX: A Benchmark for Reasoning and Learning", "date": "2025-08", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "GPU-accelerated", "PuzzleScript DSL", "Tree Search", "RL", "Hundreds of Games" ], "keywords_by_category": { "Benchmark": "GPU-accelerated, PuzzleScript DSL, Tree Search, RL, Hundreds of Games" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.16821v1" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 318, "date": "2025-08", "title": "PuzzleJAX: A Benchmark for Reasoning and Learning", "venue": "arXiv", "keywords": "GPU-accelerated, PuzzleScript DSL, Tree Search, RL, Hundreds of Games", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.16821v1" } ] } ], "sources": [ "https://arxiv.org/abs/2508.16821v1" ], "search_text": "puzzlejax: a benchmark for reasoning and learning puzzlejax 2025-08 arxiv benchmark gpu-accelerated puzzlescript dsl tree search rl hundreds of games player" }, { "id": "arxiv-2508.13009", "name": "Matrix-game 2.0", "title": "Matrix-game 2.0: An open-source real-time and streaming interactive world model", "date": "2025-08", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Few-Step Autoregressive Diffusion", "Streaming Generation", "Action Injection" ], "keywords_by_category": { "Model": "Few-Step Autoregressive Diffusion, Streaming Generation, Action Injection" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.13009" }, { "label": "Code", "url": "https://github.com/SkyworkAI/Matrix-Game/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 130, "date": "2025-08", "title": "Matrix-game 2.0: An open-source real-time and streaming interactive world model", "venue": "arXiv", "keywords": "Few-Step Autoregressive Diffusion, Streaming Generation, Action Injection", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.13009" }, { "label": "Code", "url": "https://github.com/SkyworkAI/Matrix-Game/" } ] } ], "sources": [ "https://arxiv.org/abs/2508.13009", "https://github.com/SkyworkAI/Matrix-Game/" ], "search_text": "matrix-game 2.0: an open-source real-time and streaming interactive world model matrix-game 2.0 2025-08 arxiv model few-step autoregressive diffusion streaming generation action injection supporting environment" }, { "id": "arxiv-2508.08501", "name": "GVGAI-LLM", "title": "GVGAI-LLM: Evaluating Large Language Model Agents with Infinite Games", "date": "2025-08", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "GVGAI", "Infinite Games", "Rule Generalization", "Procedural Games" ], "keywords_by_category": { "Benchmark": "GVGAI, Infinite Games, Rule Generalization, Procedural Games" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.08501" }, { "label": "Code", "url": "https://github.com/doveliyuchen/GVGAI_GYM" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 314, "date": "2025-08", "title": "GVGAI-LLM: Evaluating Large Language Model Agents with Infinite Games", "venue": "arXiv", "keywords": "GVGAI, Infinite Games, Rule Generalization, Procedural Games", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.08501" }, { "label": "Code", "url": "https://github.com/doveliyuchen/GVGAI_GYM" } ] } ], "sources": [ "https://arxiv.org/abs/2508.08501", "https://github.com/doveliyuchen/GVGAI_GYM" ], "search_text": "gvgai-llm: evaluating large language model agents with infinite games gvgai-llm 2025-08 arxiv benchmark gvgai infinite games rule generalization procedural games player" }, { "id": "arxiv-2508.03368", "name": "Game Reasoning Arena", "title": "Game Reasoning Arena: A Framework and Benchmark for Assessing Reasoning Capabilities of Large Language Models via Game Play", "date": "2025-08", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Board Games", "OpenSpiel", "Strategic Reasoning", "Framework" ], "keywords_by_category": { "Benchmark": "Board Games, OpenSpiel, Strategic Reasoning, Framework" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.03368" }, { "label": "Code", "url": "https://github.com/SLAMPAI/game_reasoning_arena" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 315, "date": "2025-08", "title": "Game Reasoning Arena: A Framework and Benchmark for Assessing Reasoning Capabilities of Large Language Models via Game Play", "venue": "arXiv", "keywords": "Board Games, OpenSpiel, Strategic Reasoning, Framework", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.03368" }, { "label": "Code", "url": "https://github.com/SLAMPAI/game_reasoning_arena" } ] } ], "sources": [ "https://arxiv.org/abs/2508.03368", "https://github.com/SLAMPAI/game_reasoning_arena" ], "search_text": "game reasoning arena: a framework and benchmark for assessing reasoning capabilities of large language models via game play game reasoning arena 2025-08 arxiv benchmark board games openspiel strategic reasoning framework player" }, { "id": "arxiv-2508.04280", "name": "Enhancing Vision-Language Model Training with Reinforcement Learning in Synthetic Worlds for Real-World Success", "title": "Enhancing Vision-Language Model Training with Reinforcement Learning in Synthetic Worlds for Real-World Success", "date": "2025-08", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "VLM + RL", "Synthetic Environment" ], "keywords_by_category": { "Model": "VLM + RL, Synthetic Environment" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.04280" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 133, "date": "2025-08", "title": "Enhancing Vision-Language Model Training with Reinforcement Learning in Synthetic Worlds for Real-World Success", "venue": "arXiv", "keywords": "VLM + RL, Synthetic Environment", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.04280" } ] } ], "sources": [ "https://arxiv.org/abs/2508.04280" ], "search_text": "enhancing vision-language model training with reinforcement learning in synthetic worlds for real-world success enhancing vision-language model training with reinforcement learning in synthetic worlds for real-world success 2025-08 arxiv model vlm + rl synthetic environment player" }, { "id": "arxiv-2508.05405", "name": "DeepPHY", "title": "DeepPHY: Benchmarking Agentic VLMs on Physical Reasoning", "date": "2025-08", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Harness", "Benchmark" ], "primary_category": null, "keywords": [ "Action Interface", "Observation Augmentation", "Physical Reasoning", "Agentic VLMs", "Predictive Control" ], "keywords_by_category": { "Harness": "Action Interface, Observation Augmentation", "Benchmark": "Physical Reasoning, Agentic VLMs, Predictive Control" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.05405" }, { "label": "Code", "url": "https://github.com/XinrunXu/DeepPHY" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 200, "date": "2025-08", "title": "DeepPHY: Benchmarking Agentic VLMs on Physical Reasoning", "venue": "arXiv", "keywords": "Action Interface, Observation Augmentation", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.05405" }, { "label": "Code", "url": "https://github.com/XinrunXu/DeepPHY" } ] }, { "section": "Benchmark", "line": 316, "date": "2025-08", "title": "DeepPHY: Benchmarking Agentic VLMs on Physical Reasoning", "venue": "arXiv", "keywords": "Physical Reasoning, Agentic VLMs, Predictive Control", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2508.05405" }, { "label": "Code", "url": "https://github.com/XinrunXu/DeepPHY" } ] } ], "sources": [ "https://arxiv.org/abs/2508.05405", "https://github.com/XinrunXu/DeepPHY" ], "search_text": "deepphy: benchmarking agentic vlms on physical reasoning deepphy 2025-08 arxiv harness benchmark action interface observation augmentation physical reasoning agentic vlms predictive control player" }, { "id": "arxiv-2507.23701", "name": "TextQuests", "title": "TextQuests: How Good are LLMs at Text-Based Video Games?", "date": "2025-07", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Text-based Adventure", "Interactive Fiction", "Long-context Reasoning" ], "keywords_by_category": { "Benchmark": "Text-based Adventure, Interactive Fiction, Long-context Reasoning" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2507.23701" }, { "label": "Code", "url": "https://github.com/centerforaisafety/textquests" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 319, "date": "2025-07", "title": "TextQuests: How Good are LLMs at Text-Based Video Games?", "venue": "arXiv", "keywords": "Text-based Adventure, Interactive Fiction, Long-context Reasoning", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2507.23701" }, { "label": "Code", "url": "https://github.com/centerforaisafety/textquests" } ] } ], "sources": [ "https://arxiv.org/abs/2507.23701", "https://github.com/centerforaisafety/textquests" ], "search_text": "textquests: how good are llms at text-based video games? textquests 2025-07 arxiv benchmark text-based adventure interactive fiction long-context reasoning player" }, { "id": "arxiv-2507.07445", "name": "StarDojo", "title": "StarDojo: Benchmarking Open-Ended Behaviors of Agentic Multimodal LLMs in Production-Living Simulations with Stardew Valley", "date": "2025-07", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Stardew Valley", "Open-Ended Tasks", "1000 Tasks", "5 Domains" ], "keywords_by_category": { "Benchmark": "Stardew Valley, Open-Ended Tasks, 1000 Tasks, 5 Domains" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2507.07445" }, { "label": "Code", "url": "https://github.com/StarDojo2025/stardojo" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 322, "date": "2025-07", "title": "StarDojo: Benchmarking Open-Ended Behaviors of Agentic Multimodal LLMs in Production-Living Simulations with Stardew Valley", "venue": "arXiv", "keywords": "Stardew Valley, Open-Ended Tasks, 1000 Tasks, 5 Domains", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2507.07445" }, { "label": "Code", "url": "https://github.com/StarDojo2025/stardojo" } ] } ], "sources": [ "https://arxiv.org/abs/2507.07445", "https://github.com/StarDojo2025/stardojo" ], "search_text": "stardojo: benchmarking open-ended behaviors of agentic multimodal llms in production-living simulations with stardew valley stardojo 2025-07 arxiv benchmark stardew valley open-ended tasks 1000 tasks 5 domains player" }, { "id": "arxiv-2507.20395", "name": "MazeEval", "title": "MazeEval: A Benchmark for Testing Sequential Decision-Making in Language Models", "date": "2025-07", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Maze Navigation", "Sequential Decision Making", "Spatial Reasoning" ], "keywords_by_category": { "Benchmark": "Maze Navigation, Sequential Decision Making, Spatial Reasoning" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2507.20395" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 320, "date": "2025-07", "title": "MazeEval: A Benchmark for Testing Sequential Decision-Making in Language Models", "venue": "arXiv", "keywords": "Maze Navigation, Sequential Decision Making, Spatial Reasoning", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2507.20395" } ] } ], "sources": [ "https://arxiv.org/abs/2507.20395" ], "search_text": "mazeeval: a benchmark for testing sequential decision-making in language models mazeeval 2025-07 arxiv benchmark maze navigation sequential decision making spatial reasoning player" }, { "id": "arxiv-2507.11633", "name": "General Modular Harness for LLM Agents in Multi-Turn Gaming Environments", "title": "General Modular Harness for LLM Agents in Multi-Turn Gaming Environments", "date": "2025-07", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICML MAS workshop", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "Memory", "Reflection", "Workflow" ], "keywords_by_category": { "Harness": "Memory, Reflection, Workflow" }, "links": [ { "label": "paper", "url": "https://arxiv.org/abs/2507.11633" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 202, "date": "2025-07", "title": "General Modular Harness for LLM Agents in Multi-Turn Gaming Environments", "venue": "ICML MAS workshop", "keywords": "Memory, Reflection, Workflow", "links": [ { "label": "paper", "url": "https://arxiv.org/abs/2507.11633" } ] } ], "sources": [ "https://arxiv.org/abs/2507.11633" ], "search_text": "general modular harness for llm agents in multi-turn gaming environments general modular harness for llm agents in multi-turn gaming environments 2025-07 icml mas workshop harness memory reflection workflow player" }, { "id": "arxiv-2507.08491", "name": "A Third Paradigm for LLM Evaluation", "title": "A Third Paradigm for LLM Evaluation: Dialogue Game-Based Evaluation using clembench", "date": "2025-07", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Dialogue-game Evaluation", "clembench", "Interactive Evaluation" ], "keywords_by_category": { "Benchmark": "Dialogue-game Evaluation, clembench, Interactive Evaluation" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2507.08491" }, { "label": "Code", "url": "https://github.com/clembench/clembench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 321, "date": "2025-07", "title": "A Third Paradigm for LLM Evaluation: Dialogue Game-Based Evaluation using clembench", "venue": "arXiv", "keywords": "Dialogue-game Evaluation, clembench, Interactive Evaluation", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2507.08491" }, { "label": "Code", "url": "https://github.com/clembench/clembench" } ] } ], "sources": [ "https://arxiv.org/abs/2507.08491", "https://github.com/clembench/clembench" ], "search_text": "a third paradigm for llm evaluation: dialogue game-based evaluation using clembench a third paradigm for llm evaluation 2025-07 arxiv benchmark dialogue-game evaluation clembench interactive evaluation player" }, { "id": "arxiv-2506.02387", "name": "VS-Bench", "title": "VS-Bench: Evaluating VLMs for Strategic Abilities in Multi-Agent Environments", "date": "2025-06", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "CoT" ], "keywords_by_category": { "Harness": "CoT" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2506.02387" }, { "label": "Code", "url": "https://github.com/zelaix/VS-Bench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 204, "date": "2025-06", "title": "VS-Bench: Evaluating VLMs for Strategic Abilities in Multi-Agent Environments", "venue": "arXiv", "keywords": "CoT", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2506.02387" }, { "label": "Code", "url": "https://github.com/zelaix/VS-Bench" } ] } ], "sources": [ "https://arxiv.org/abs/2506.02387", "https://github.com/zelaix/VS-Bench" ], "search_text": "vs-bench: evaluating vlms for strategic abilities in multi-agent environments vs-bench 2025-06 arxiv harness cot player" }, { "id": "arxiv-2506.04098", "name": "TextAtari", "title": "TextAtari: 100K Frames Game Playing with Language Agents", "date": "2025-06", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Text-Based Atari", "Long-Horizon", "100K Frames", "23 Games" ], "keywords_by_category": { "Benchmark": "Text-Based Atari, Long-Horizon, 100K Frames, 23 Games" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2506.04098" }, { "label": "Code", "url": "https://github.com/Lww007/Text-Atari-Agents" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 323, "date": "2025-06", "title": "TextAtari: 100K Frames Game Playing with Language Agents", "venue": "arXiv", "keywords": "Text-Based Atari, Long-Horizon, 100K Frames, 23 Games", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2506.04098" }, { "label": "Code", "url": "https://github.com/Lww007/Text-Atari-Agents" } ] } ], "sources": [ "https://arxiv.org/abs/2506.04098", "https://github.com/Lww007/Text-Atari-Agents" ], "search_text": "textatari: 100k frames game playing with language agents textatari 2025-06 arxiv benchmark text-based atari long-horizon 100k frames 23 games player" }, { "id": "arxiv-2506.08011", "name": "Play to Generalize", "title": "Play to Generalize: Learning to Reason Through Game Play", "date": "2025-06", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "RL", "Game→Reasoning Transfer" ], "keywords_by_category": { "Model": "RL, Game→Reasoning Transfer" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2506.08011" }, { "label": "Project", "url": "https://yunfeixie233.github.io/ViGaL/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 134, "date": "2025-06", "title": "Play to Generalize: Learning to Reason Through Game Play", "venue": "arXiv", "keywords": "RL, Game→Reasoning Transfer", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2506.08011" }, { "label": "Project", "url": "https://yunfeixie233.github.io/ViGaL/" } ] } ], "sources": [ "https://arxiv.org/abs/2506.08011", "https://yunfeixie233.github.io/ViGaL/" ], "search_text": "play to generalize: learning to reason through game play play to generalize 2025-06 arxiv model rl game→reasoning transfer player" }, { "id": "arxiv-2506.03610", "name": "Orak", "title": "Orak: A Foundational Benchmark for Training and Evaluating LLM Agents on Diverse Video Games", "date": "2025-06", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2026", "categories": [ "Dataset", "Harness", "Benchmark" ], "primary_category": null, "keywords": [ "12 Genres", "Multi-Task Evaluation", "MCP", "Reflection", "Training & Evaluation", "Real-World Games" ], "keywords_by_category": { "Dataset": "12 Genres, Multi-Task Evaluation", "Harness": "MCP, Reflection", "Benchmark": "12 Genres, Training & Evaluation, Real-World Games" }, "links": [ { "label": "Paper", "url": "https://iclr.cc/virtual/2026/poster/10010438" }, { "label": "Code", "url": "https://github.com/krafton-ai/Orak" }, { "label": "Paper", "url": "https://arxiv.org/abs/2506.03610" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 203, "date": "2025-06", "title": "Orak: A Foundational Benchmark for Training and Evaluating LLM Agents on Diverse Video Games", "venue": "ICLR 2026", "keywords": "MCP, Reflection", "links": [ { "label": "Paper", "url": "https://iclr.cc/virtual/2026/poster/10010438" }, { "label": "Code", "url": "https://github.com/krafton-ai/Orak" } ] }, { "section": "Dataset", "line": 248, "date": "2025-06", "title": "Orak: A Foundational Benchmark for Training and Evaluating LLM Agents on Diverse Video Games", "venue": "ICLR 2026", "keywords": "12 Genres, Multi-Task Evaluation", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2506.03610" }, { "label": "Code", "url": "https://github.com/krafton-ai/Orak" } ] }, { "section": "Benchmark", "line": 324, "date": "2025-06", "title": "Orak: A Foundational Benchmark for Training and Evaluating LLM Agents on Diverse Video Games", "venue": "ICLR 2026", "keywords": "12 Genres, Training & Evaluation, Real-World Games", "links": [ { "label": "Paper", "url": "https://iclr.cc/virtual/2026/poster/10010438" }, { "label": "Code", "url": "https://github.com/krafton-ai/Orak" } ] } ], "sources": [ "https://iclr.cc/virtual/2026/poster/10010438", "https://github.com/krafton-ai/Orak", "https://arxiv.org/abs/2506.03610" ], "search_text": "orak: a foundational benchmark for training and evaluating llm agents on diverse video games orak 2025-06 iclr 2026 dataset harness benchmark 12 genres multi-task evaluation mcp reflection training & evaluation real-world games player" }, { "id": "arxiv-2506.18701", "name": "Matrix-Game", "title": "Matrix-Game: Interactive World Foundation Model", "date": "2025-06", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Image-to-World Generation", "Large-Scale Minecraft Pretraining", "Action Controllability" ], "keywords_by_category": { "Model": "Image-to-World Generation, Large-Scale Minecraft Pretraining, Action Controllability" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2506.18701" }, { "label": "Code", "url": "https://github.com/SkyworkAI/Matrix-Game" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 135, "date": "2025-06", "title": "Matrix-Game: Interactive World Foundation Model", "venue": "arXiv", "keywords": "Image-to-World Generation, Large-Scale Minecraft Pretraining, Action Controllability", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2506.18701" }, { "label": "Code", "url": "https://github.com/SkyworkAI/Matrix-Game" } ] } ], "sources": [ "https://arxiv.org/abs/2506.18701", "https://github.com/SkyworkAI/Matrix-Game" ], "search_text": "matrix-game: interactive world foundation model matrix-game 2025-06 arxiv model image-to-world generation large-scale minecraft pretraining action controllability supporting environment" }, { "id": "arxiv-2505.18134", "name": "VideoGameBench", "title": "VideoGameBench: Can Vision-Language Models complete popular video games?", "date": "2025-05", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Harness", "Benchmark" ], "primary_category": null, "keywords": [ "Action Interface", "ReAct", "Pause", "Text Memory", "10 Classic Games", "Real-Time VLM", "Raw Visual Inputs" ], "keywords_by_category": { "Harness": "Action Interface, ReAct, Pause, Text Memory", "Benchmark": "10 Classic Games, Real-Time VLM, Raw Visual Inputs" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2505.18134" }, { "label": "Code", "url": "https://github.com/alexzhang13/videogamebench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 207, "date": "2025-05", "title": "VideoGameBench: Can Vision-Language Models complete popular video games?", "venue": "arXiv", "keywords": "Action Interface, ReAct, Pause, Text Memory", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2505.18134" }, { "label": "Code", "url": "https://github.com/alexzhang13/videogamebench" } ] }, { "section": "Benchmark", "line": 327, "date": "2025-05", "title": "VideoGameBench: Can Vision-Language Models complete popular video games?", "venue": "arXiv", "keywords": "10 Classic Games, Real-Time VLM, Raw Visual Inputs", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2505.18134" }, { "label": "Code", "url": "https://github.com/alexzhang13/videogamebench" } ] } ], "sources": [ "https://arxiv.org/abs/2505.18134", "https://github.com/alexzhang13/videogamebench" ], "search_text": "videogamebench: can vision-language models complete popular video games? videogamebench 2025-05 arxiv harness benchmark action interface react pause text memory 10 classic games real-time vlm raw visual inputs player" }, { "id": "work-ef0a557bd444", "name": "Towards Efficient Online Tuning of VLM Agents via Counterfactual Soft Reinforcement Learning", "title": "Towards Efficient Online Tuning of VLM Agents via Counterfactual Soft Reinforcement Learning", "date": "2025-05", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICML 2025", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Counterfactual Soft RL", "Online Tuning" ], "keywords_by_category": { "Model": "Counterfactual Soft RL, Online Tuning" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=H76PMm7hf2" }, { "label": "Code", "url": "https://github.com/langfengQ/CoSo" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 137, "date": "2025-05", "title": "Towards Efficient Online Tuning of VLM Agents via Counterfactual Soft Reinforcement Learning", "venue": "ICML 2025", "keywords": "Counterfactual Soft RL, Online Tuning", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=H76PMm7hf2" }, { "label": "Code", "url": "https://github.com/langfengQ/CoSo" } ] } ], "sources": [ "https://openreview.net/forum?id=H76PMm7hf2", "https://github.com/langfengQ/CoSo" ], "search_text": "towards efficient online tuning of vlm agents via counterfactual soft reinforcement learning towards efficient online tuning of vlm agents via counterfactual soft reinforcement learning 2025-05 icml 2025 model counterfactual soft rl online tuning player" }, { "id": "arxiv-2505.16135", "name": "Sudoku-Bench", "title": "Sudoku-Bench: Evaluating creative reasoning with Sudoku variants", "date": "2025-05", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Creative Reasoning", "Sudoku Variants", "Anti-Memorization", "Logical Breakthroughs" ], "keywords_by_category": { "Benchmark": "Creative Reasoning, Sudoku Variants, Anti-Memorization, Logical Breakthroughs" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2505.16135" }, { "label": "Project", "url": "https://pub.sakana.ai/sudoku/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 328, "date": "2025-05", "title": "Sudoku-Bench: Evaluating creative reasoning with Sudoku variants", "venue": "arXiv", "keywords": "Creative Reasoning, Sudoku Variants, Anti-Memorization, Logical Breakthroughs", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2505.16135" }, { "label": "Project", "url": "https://pub.sakana.ai/sudoku/" } ] } ], "sources": [ "https://arxiv.org/abs/2505.16135", "https://pub.sakana.ai/sudoku/" ], "search_text": "sudoku-bench: evaluating creative reasoning with sudoku variants sudoku-bench 2025-05 arxiv benchmark creative reasoning sudoku variants anti-memorization logical breakthroughs player" }, { "id": "arxiv-2505.12707", "name": "PLAICraft", "title": "PLAICraft: Large-Scale Time-Aligned Vision-Speech-Action Dataset for Embodied AI", "date": "2025-05", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Dataset" ], "primary_category": null, "keywords": [ "10K hrs", "5 Modalities", "10K+ Players" ], "keywords_by_category": { "Dataset": "10K hrs, 5 Modalities, 10K+ Players" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2505.12707" }, { "label": "Project", "url": "https://www.plaicraft.ai/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Dataset", "line": 249, "date": "2025-05", "title": "PLAICraft: Large-Scale Time-Aligned Vision-Speech-Action Dataset for Embodied AI", "venue": "arXiv", "keywords": "10K hrs, 5 Modalities, 10K+ Players", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2505.12707" }, { "label": "Project", "url": "https://www.plaicraft.ai/" } ] } ], "sources": [ "https://arxiv.org/abs/2505.12707", "https://www.plaicraft.ai/" ], "search_text": "plaicraft: large-scale time-aligned vision-speech-action dataset for embodied ai plaicraft 2025-05 arxiv dataset 10k hrs 5 modalities 10k+ players player" }, { "id": "work-741e3982a6dd", "name": "MCU", "title": "MCU: An Evaluation Framework for Open-Ended Game Agents", "date": "2025-05", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICML 2025", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Minecraft", "Open-Ended", "3", "452 Tasks", "Task Composition", "91.5% Human Alignment" ], "keywords_by_category": { "Benchmark": "Minecraft, Open-Ended, 3,452 Tasks, Task Composition, 91.5% Human Alignment" }, "links": [ { "label": "Paper", "url": "https://proceedings.mlr.press/v267/zheng25j.html" }, { "label": "Code", "url": "https://github.com/CraftJarvis/MCU" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 325, "date": "2025-05", "title": "MCU: An Evaluation Framework for Open-Ended Game Agents", "venue": "ICML 2025", "keywords": "Minecraft, Open-Ended, 3,452 Tasks, Task Composition, 91.5% Human Alignment", "links": [ { "label": "Paper", "url": "https://proceedings.mlr.press/v267/zheng25j.html" }, { "label": "Code", "url": "https://github.com/CraftJarvis/MCU" } ] } ], "sources": [ "https://proceedings.mlr.press/v267/zheng25j.html", "https://github.com/CraftJarvis/MCU" ], "search_text": "mcu: an evaluation framework for open-ended game agents mcu 2025-05 icml 2025 benchmark minecraft open-ended 3 452 tasks task composition 91.5% human alignment player" }, { "id": "work-d33623365a4f", "name": "lmgame-Bench", "title": "lmgame-Bench: How Good are LLMs at Playing Games?", "date": "2025-05", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2026", "categories": [ "Harness", "Benchmark" ], "primary_category": null, "keywords": [ "Textual Environment", "Memory", "Reflection", "CoT", "Action Interface", "Perception", "Interactive Decision-Making", "13 Models" ], "keywords_by_category": { "Harness": "Textual Environment, Memory, Reflection, CoT, Action Interface", "Benchmark": "Perception, Memory, Interactive Decision-Making, 13 Models" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=qeziG97WUZ" }, { "label": "Code", "url": "https://github.com/lmgame-org/GamingAgent" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 205, "date": "2025-05", "title": "lmgame-Bench: How Good are LLMs at Playing Games?", "venue": "ICLR 2026", "keywords": "Textual Environment, Memory, Reflection, CoT, Action Interface", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=qeziG97WUZ" }, { "label": "Code", "url": "https://github.com/lmgame-org/GamingAgent" } ] }, { "section": "Benchmark", "line": 326, "date": "2025-05", "title": "lmgame-Bench: How Good are LLMs at Playing Games?", "venue": "ICLR 2026", "keywords": "Perception, Memory, Interactive Decision-Making, 13 Models", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=qeziG97WUZ" }, { "label": "Code", "url": "https://github.com/lmgame-org/GamingAgent" } ] } ], "sources": [ "https://openreview.net/forum?id=qeziG97WUZ", "https://github.com/lmgame-org/GamingAgent" ], "search_text": "lmgame-bench: how good are llms at playing games? lmgame-bench 2025-05 iclr 2026 harness benchmark textual environment memory reflection cot action interface perception interactive decision-making 13 models player" }, { "id": "work-a54538a06726", "name": "Learning to Play Like Humans", "title": "Learning to Play Like Humans: A Framework for LLM Adaptation in Interactive Fiction Games", "date": "2025-05", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ACL Findings 2025", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "Interactive Fiction games", "RAG", "CoT" ], "keywords_by_category": { "Harness": "Interactive Fiction games, RAG, CoT" }, "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.findings-acl.531/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 206, "date": "2025-05", "title": "Learning to Play Like Humans: A Framework for LLM Adaptation in Interactive Fiction Games", "venue": "ACL Findings 2025", "keywords": "Interactive Fiction games, RAG, CoT", "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.findings-acl.531/" } ] } ], "sources": [ "https://aclanthology.org/2025.findings-acl.531/" ], "search_text": "learning to play like humans: a framework for llm adaptation in interactive fiction games learning to play like humans 2025-05 acl findings 2025 harness interactive fiction games rag cot player" }, { "id": "work-6a6782f31963", "name": "KORGym", "title": "KORGym: A Dynamic Game Platform for LLM Reasoning Evaluation", "date": "2025-05", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NeurIPS 2025", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "50+ Games", "Dynamic Evaluation", "Multi-turn", "RL Scenarios", "Kingdom Rush" ], "keywords_by_category": { "Benchmark": "50+ Games, Dynamic Evaluation, Multi-turn, RL Scenarios, Kingdom Rush" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=uAeqQePu4c" }, { "label": "Code", "url": "https://github.com/multimodal-art-projection/KORGym" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 329, "date": "2025-05", "title": "KORGym: A Dynamic Game Platform for LLM Reasoning Evaluation", "venue": "NeurIPS 2025", "keywords": "50+ Games, Dynamic Evaluation, Multi-turn, RL Scenarios, Kingdom Rush", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=uAeqQePu4c" }, { "label": "Code", "url": "https://github.com/multimodal-art-projection/KORGym" } ] } ], "sources": [ "https://openreview.net/forum?id=uAeqQePu4c", "https://github.com/multimodal-art-projection/KORGym" ], "search_text": "korgym: a dynamic game platform for llm reasoning evaluation korgym 2025-05 neurips 2025 benchmark 50+ games dynamic evaluation multi-turn rl scenarios kingdom rush player" }, { "id": "arxiv-2505.13886", "name": "Game-RL", "title": "Game-RL: Synthesizing Multimodal Verifiable Game Data to Boost VLMs' General Reasoning", "date": "2025-05", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv / ICLR 2026", "categories": [ "Dataset", "Model" ], "primary_category": null, "keywords": [ "30 Games", "158 Tasks", "Verifiable QA", "RL", "Code2Logic", "Verifiable Reward" ], "keywords_by_category": { "Dataset": "30 Games, 158 Tasks, Verifiable QA", "Model": "RL, Code2Logic, Verifiable Reward" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2505.13886" }, { "label": "Paper", "url": "https://openreview.net/forum?id=e4FqU4SyHL" }, { "label": "Code", "url": "https://github.com/tongjingqi/Game-RL" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 136, "date": "2025-05", "title": "Game-RL: Synthesizing Multimodal Verifiable Game Data to Boost VLMs' General Reasoning", "venue": "arXiv", "keywords": "RL, Code2Logic, Verifiable Reward", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2505.13886" } ] }, { "section": "Dataset", "line": 250, "date": "2025-05", "title": "Game-RL: Synthesizing Multimodal Verifiable Game Data to Boost VLMs' General Reasoning", "venue": "ICLR 2026", "keywords": "30 Games, 158 Tasks, Verifiable QA", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=e4FqU4SyHL" }, { "label": "Code", "url": "https://github.com/tongjingqi/Game-RL" } ] } ], "sources": [ "https://arxiv.org/abs/2505.13886", "https://openreview.net/forum?id=e4FqU4SyHL", "https://github.com/tongjingqi/Game-RL" ], "search_text": "game-rl: synthesizing multimodal verifiable game data to boost vlms' general reasoning game-rl 2025-05 arxiv / iclr 2026 dataset model 30 games 158 tasks verifiable qa rl code2logic verifiable reward player" }, { "id": "arxiv-2505.13426", "name": "G1", "title": "G1: Bootstrapping Perception and Reasoning Abilities of Vision-Language Model via Reinforcement Learning", "date": "2025-05", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Visual RL", "VLM-Gym" ], "keywords_by_category": { "Model": "Visual RL, VLM-Gym" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2505.13426" }, { "label": "Code", "url": "https://github.com/chenllliang/G1" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 138, "date": "2025-05", "title": "G1: Bootstrapping Perception and Reasoning Abilities of Vision-Language Model via Reinforcement Learning", "venue": "arXiv", "keywords": "Visual RL, VLM-Gym", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2505.13426" }, { "label": "Code", "url": "https://github.com/chenllliang/G1" } ] } ], "sources": [ "https://arxiv.org/abs/2505.13426", "https://github.com/chenllliang/G1" ], "search_text": "g1: bootstrapping perception and reasoning abilities of vision-language model via reinforcement learning g1 2025-05 arxiv model visual rl vlm-gym player" }, { "id": "work-d8e84ce37480", "name": "WHAMM! Real-time world modelling of interactive environments.", "title": "WHAMM! Real-time world modelling of interactive environments.", "date": "2025-04", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "Research blog", "categories": [ "Model" ], "primary_category": null, "keywords": [ "MaskGIT World Model", "Real-Time Gameplay", "Quake II Transfer" ], "keywords_by_category": { "Model": "MaskGIT World Model, Real-Time Gameplay, Quake II Transfer" }, "links": [ { "label": "Paper", "url": "https://www.microsoft.com/en-us/research/articles/whamm-real-time-world-modelling-of-interactive-environments/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 141, "date": "2025-04", "title": "WHAMM! Real-time world modelling of interactive environments.", "venue": "Research blog", "keywords": "MaskGIT World Model, Real-Time Gameplay, Quake II Transfer", "links": [ { "label": "Paper", "url": "https://www.microsoft.com/en-us/research/articles/whamm-real-time-world-modelling-of-interactive-environments/" } ] } ], "sources": [ "https://www.microsoft.com/en-us/research/articles/whamm-real-time-world-modelling-of-interactive-environments/" ], "search_text": "whamm! real-time world modelling of interactive environments. whamm! real-time world modelling of interactive environments. 2025-04 research blog model maskgit world model real-time gameplay quake ii transfer supporting environment" }, { "id": "arxiv-2504.06148", "name": "V-MAGE", "title": "V-MAGE: A Game Evaluation Framework for Assessing Vision-Centric Capabilities in Multimodal Large Language Models", "date": "2025-04", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ACL 2026", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Vision-Centric Reasoning", "Elo Ranking", "Human Baseline", "Dynamic Interaction" ], "keywords_by_category": { "Benchmark": "Vision-Centric Reasoning, Elo Ranking, Human Baseline, Dynamic Interaction" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2504.06148" }, { "label": "Code", "url": "https://github.com/CSU-JPG/V-MAGE" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 331, "date": "2025-04", "title": "V-MAGE: A Game Evaluation Framework for Assessing Vision-Centric Capabilities in Multimodal Large Language Models", "venue": "ACL 2026", "keywords": "Vision-Centric Reasoning, Elo Ranking, Human Baseline, Dynamic Interaction", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2504.06148" }, { "label": "Code", "url": "https://github.com/CSU-JPG/V-MAGE" } ] } ], "sources": [ "https://arxiv.org/abs/2504.06148", "https://github.com/CSU-JPG/V-MAGE" ], "search_text": "v-mage: a game evaluation framework for assessing vision-centric capabilities in multimodal large language models v-mage 2025-04 acl 2026 benchmark vision-centric reasoning elo ranking human baseline dynamic interaction player" }, { "id": "work-be6510e383ca", "name": "UI-TARS-1.5", "title": "UI-TARS-1.5", "date": "2025-04", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "Research blog", "categories": [ "Model" ], "primary_category": null, "keywords": [ "VLA", "GUI Agent", "Game Reasoning" ], "keywords_by_category": { "Model": "VLA, GUI Agent, Game Reasoning" }, "links": [ { "label": "Project", "url": "https://seed-tars.com/1.5" }, { "label": "Code", "url": "https://github.com/bytedance/UI-TARS" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 140, "date": "2025-04", "title": "UI-TARS-1.5", "venue": "Research blog", "keywords": "VLA, GUI Agent, Game Reasoning", "links": [ { "label": "Project", "url": "https://seed-tars.com/1.5" }, { "label": "Code", "url": "https://github.com/bytedance/UI-TARS" } ] } ], "sources": [ "https://seed-tars.com/1.5", "https://github.com/bytedance/UI-TARS" ], "search_text": "ui-tars-1.5 ui-tars-1.5 2025-04 research blog model vla gui agent game reasoning player" }, { "id": "arxiv-2504.11442", "name": "TextArena", "title": "TextArena", "date": "2025-04", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Competitive Text Games", "Multi-agent", "Online TrueSkill", "Social Skills" ], "keywords_by_category": { "Benchmark": "Competitive Text Games, Multi-agent, Online TrueSkill, Social Skills" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2504.11442" }, { "label": "Code", "url": "https://github.com/TextArena/TextArena" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 330, "date": "2025-04", "title": "TextArena", "venue": "arXiv", "keywords": "Competitive Text Games, Multi-agent, Online TrueSkill, Social Skills", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2504.11442" }, { "label": "Code", "url": "https://github.com/TextArena/TextArena" } ] } ], "sources": [ "https://arxiv.org/abs/2504.11442", "https://github.com/TextArena/TextArena" ], "search_text": "textarena textarena 2025-04 arxiv benchmark competitive text games multi-agent online trueskill social skills player" }, { "id": "arxiv-2504.08388", "name": "MineWorld", "title": "MineWorld: a Real-Time and Open-Source Interactive World Model on Minecraft", "date": "2025-04", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Visual-Action Autoregressive Transformer", "Parallel Decoding", "Action Following" ], "keywords_by_category": { "Model": "Visual-Action Autoregressive Transformer, Parallel Decoding, Action Following" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2504.08388" }, { "label": "Code", "url": "https://github.com/microsoft/mineworld" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 139, "date": "2025-04", "title": "MineWorld: a Real-Time and Open-Source Interactive World Model on Minecraft", "venue": "arXiv", "keywords": "Visual-Action Autoregressive Transformer, Parallel Decoding, Action Following", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2504.08388" }, { "label": "Code", "url": "https://github.com/microsoft/mineworld" } ] } ], "sources": [ "https://arxiv.org/abs/2504.08388", "https://github.com/microsoft/mineworld" ], "search_text": "mineworld: a real-time and open-source interactive world model on minecraft mineworld 2025-04 arxiv model visual-action autoregressive transformer parallel decoding action following supporting environment" }, { "id": "work-b74cc122548e", "name": "Human-Level Competitive Pokémon via Scalable Offline Reinforcement Learning with Transformers", "title": "Human-Level Competitive Pokémon via Scalable Offline Reinforcement Learning with Transformers", "date": "2025-04", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "RL Conf. 2025", "categories": [ "Dataset" ], "primary_category": null, "keywords": [ "Decade of Replays", "Competitive Pokemon" ], "keywords_by_category": { "Dataset": "Decade of Replays, Competitive Pokemon" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=b1BaJ1edFS" }, { "label": "Code", "url": "https://github.com/UT-Austin-RPL/metamon/tree/main" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Dataset", "line": 251, "date": "2025-04", "title": "Human-Level Competitive Pokémon via Scalable Offline Reinforcement Learning with Transformers", "venue": "RL Conf. 2025", "keywords": "Decade of Replays, Competitive Pokemon", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=b1BaJ1edFS" }, { "label": "Code", "url": "https://github.com/UT-Austin-RPL/metamon/tree/main" } ] } ], "sources": [ "https://openreview.net/forum?id=b1BaJ1edFS", "https://github.com/UT-Austin-RPL/metamon/tree/main" ], "search_text": "human-level competitive pokémon via scalable offline reinforcement learning with transformers human-level competitive pokémon via scalable offline reinforcement learning with transformers 2025-04 rl conf. 2025 dataset decade of replays competitive pokemon player" }, { "id": "arxiv-2503.23064", "name": "VGRP-Bench", "title": "VGRP-Bench: Visual Grid Reasoning Puzzle Benchmark for Large Vision-Language Models", "date": "2025-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Visual Grid Reasoning", "LVLM", "20 Puzzles", "Rule Comprehension", "Spatial Reasoning" ], "keywords_by_category": { "Benchmark": "Visual Grid Reasoning, LVLM, 20 Puzzles, Rule Comprehension, Spatial Reasoning" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2503.23064" }, { "label": "Code", "url": "https://github.com/ryf1123/VGRP-Bench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 332, "date": "2025-03", "title": "VGRP-Bench: Visual Grid Reasoning Puzzle Benchmark for Large Vision-Language Models", "venue": "arXiv", "keywords": "Visual Grid Reasoning, LVLM, 20 Puzzles, Rule Comprehension, Spatial Reasoning", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2503.23064" }, { "label": "Code", "url": "https://github.com/ryf1123/VGRP-Bench" } ] } ], "sources": [ "https://arxiv.org/abs/2503.23064", "https://github.com/ryf1123/VGRP-Bench" ], "search_text": "vgrp-bench: visual grid reasoning puzzle benchmark for large vision-language models vgrp-bench 2025-03 arxiv benchmark visual grid reasoning lvlm 20 puzzles rule comprehension spatial reasoning player" }, { "id": "arxiv-2503.04094", "name": "PokéChamp", "title": "PokéChamp: an Expert-level Minimax Language Agent", "date": "2025-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICML 2025", "categories": [ "Dataset", "Benchmark" ], "primary_category": null, "keywords": [ "3M+ Battles", "Competitive Logs", "Elo", "strategy game", "Pokémon" ], "keywords_by_category": { "Dataset": "3M+ Battles, Competitive Logs", "Benchmark": "Elo, strategy game, Pokémon" }, "links": [ { "label": "Paper", "url": "https://icml.cc/virtual/2025/poster/45207" }, { "label": "Code", "url": "https://github.com/sethkarten/pokechamp" }, { "label": "Paper", "url": "https://arxiv.org/abs/2503.04094" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Dataset", "line": 252, "date": "2025-03", "title": "PokéChamp: an Expert-level Minimax Language Agent", "venue": "ICML 2025", "keywords": "3M+ Battles, Competitive Logs", "links": [ { "label": "Paper", "url": "https://icml.cc/virtual/2025/poster/45207" }, { "label": "Code", "url": "https://github.com/sethkarten/pokechamp" } ] }, { "section": "Benchmark", "line": 334, "date": "2025-03", "title": "PokéChamp: an Expert-level Minimax Language Agent", "venue": "ICML 2025", "keywords": "Elo, strategy game, Pokémon", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2503.04094" }, { "label": "Code", "url": "https://github.com/sethkarten/pokechamp" } ] } ], "sources": [ "https://icml.cc/virtual/2025/poster/45207", "https://github.com/sethkarten/pokechamp", "https://arxiv.org/abs/2503.04094" ], "search_text": "pokéchamp: an expert-level minimax language agent pokéchamp 2025-03 icml 2025 dataset benchmark 3m+ battles competitive logs elo strategy game pokémon player" }, { "id": "arxiv-2503.21172", "name": "Model as a Game", "title": "Model as a Game: On Numerical and Spatial Consistency for Generative Games", "date": "2025-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICCV Workshop 2025", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Numerical Consistency", "Spatial Consistency", "LogicNet + Spatial Memory" ], "keywords_by_category": { "Model": "Numerical Consistency, Spatial Consistency, LogicNet + Spatial Memory" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2503.21172" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 144, "date": "2025-03", "title": "Model as a Game: On Numerical and Spatial Consistency for Generative Games", "venue": "ICCV Workshop 2025", "keywords": "Numerical Consistency, Spatial Consistency, LogicNet + Spatial Memory", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2503.21172" } ] } ], "sources": [ "https://arxiv.org/abs/2503.21172" ], "search_text": "model as a game: on numerical and spatial consistency for generative games model as a game 2025-03 iccv workshop 2025 model numerical consistency spatial consistency logicnet + spatial memory supporting environment" }, { "id": "work-d82df89de450", "name": "JARVIS-VLA", "title": "JARVIS-VLA: Post-Training Large-Scale Vision Language Models to Play Visual Games with Keyboards and Mouse", "date": "2025-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ACL Findings 2025 / ACL 2025", "categories": [ "Dataset", "Model" ], "primary_category": null, "keywords": [ "1000+ Atomic Tasks", "Minecraft", "VLA Post-Training" ], "keywords_by_category": { "Dataset": "1000+ Atomic Tasks, Minecraft", "Model": "VLA Post-Training, Minecraft" }, "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.findings-acl.920/" }, { "label": "Code", "url": "https://github.com/CraftJarvis/JarvisVLA" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 142, "date": "2025-03", "title": "JARVIS-VLA: Post-Training Large-Scale Vision Language Models to Play Visual Games with Keyboards and Mouse", "venue": "ACL Findings 2025", "keywords": "VLA Post-Training, Minecraft", "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.findings-acl.920/" }, { "label": "Code", "url": "https://github.com/CraftJarvis/JarvisVLA" } ] }, { "section": "Dataset", "line": 253, "date": "2025-03", "title": "JARVIS-VLA: Post-Training Large-Scale Vision Language Models to Play Visual Games with Keyboards and Mouse", "venue": "ACL 2025", "keywords": "1000+ Atomic Tasks, Minecraft", "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.findings-acl.920/" }, { "label": "Code", "url": "https://github.com/CraftJarvis/JarvisVLA" } ] } ], "sources": [ "https://aclanthology.org/2025.findings-acl.920/", "https://github.com/CraftJarvis/JarvisVLA" ], "search_text": "jarvis-vla: post-training large-scale vision language models to play visual games with keyboards and mouse jarvis-vla 2025-03 acl findings 2025 / acl 2025 dataset model 1000+ atomic tasks minecraft vla post-training player" }, { "id": "arxiv-2503.13980", "name": "Empowering LLMs in Decision Games through Algorithmic Data Synthesis", "title": "Empowering LLMs in Decision Games through Algorithmic Data Synthesis", "date": "2025-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2025 Workshop", "categories": [ "Dataset" ], "primary_category": null, "keywords": [ "Synthetic Data", "Strategic Games" ], "keywords_by_category": { "Dataset": "Synthetic Data, Strategic Games" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2503.13980" }, { "label": "Code", "url": "https://github.com/opendilab/Mastermind" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Dataset", "line": 254, "date": "2025-03", "title": "Empowering LLMs in Decision Games through Algorithmic Data Synthesis", "venue": "ICLR 2025 Workshop", "keywords": "Synthetic Data, Strategic Games", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2503.13980" }, { "label": "Code", "url": "https://github.com/opendilab/Mastermind" } ] } ], "sources": [ "https://arxiv.org/abs/2503.13980", "https://github.com/opendilab/Mastermind" ], "search_text": "empowering llms in decision games through algorithmic data synthesis empowering llms in decision games through algorithmic data synthesis 2025-03 iclr 2025 workshop dataset synthetic data strategic games player" }, { "id": "arxiv-2503.06047", "name": "DSGBench", "title": "DSGBench: A Diverse Strategic Game Benchmark for Evaluating LLM-based Agents in Complex Decision-Making Environments", "date": "2025-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Harness", "Benchmark" ], "primary_category": null, "keywords": [ "Textual Environment", "6 Strategic Games", "StarCraft II", "Werewolf", "Social Reasoning" ], "keywords_by_category": { "Harness": "Textual Environment", "Benchmark": "6 Strategic Games, StarCraft II, Werewolf, Social Reasoning" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2503.06047" }, { "label": "Code", "url": "https://github.com/DeciBrain-Group/DSGBench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 209, "date": "2025-03", "title": "DSGBench: A Diverse Strategic Game Benchmark for Evaluating LLM-based Agents in Complex Decision-Making Environments", "venue": "arXiv", "keywords": "Textual Environment", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2503.06047" }, { "label": "Code", "url": "https://github.com/DeciBrain-Group/DSGBench" } ] }, { "section": "Benchmark", "line": 335, "date": "2025-03", "title": "DSGBench: A Diverse Strategic Game Benchmark for Evaluating LLM-based Agents in Complex Decision-Making Environments", "venue": "arXiv", "keywords": "6 Strategic Games, StarCraft II, Werewolf, Social Reasoning", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2503.06047" }, { "label": "Code", "url": "https://github.com/DeciBrain-Group/DSGBench" } ] } ], "sources": [ "https://arxiv.org/abs/2503.06047", "https://github.com/DeciBrain-Group/DSGBench" ], "search_text": "dsgbench: a diverse strategic game benchmark for evaluating llm-based agents in complex decision-making environments dsgbench 2025-03 arxiv harness benchmark textual environment 6 strategic games starcraft ii werewolf social reasoning player" }, { "id": "work-3f5c7834925d", "name": "Cultivating Gaming Sense for Yourself", "title": "Cultivating Gaming Sense for Yourself: Making VLMs Gaming Experts", "date": "2025-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ACL 2025", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "VLM codes modules for game playing" ], "keywords_by_category": { "Harness": "VLM codes modules for game playing" }, "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.acl-long.643/" }, { "label": "Code", "url": "https://github.com/ipsss2/GameSense" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 208, "date": "2025-03", "title": "Cultivating Gaming Sense for Yourself: Making VLMs Gaming Experts", "venue": "ACL 2025", "keywords": "VLM codes modules for game playing", "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.acl-long.643/" }, { "label": "Code", "url": "https://github.com/ipsss2/GameSense" } ] } ], "sources": [ "https://aclanthology.org/2025.acl-long.643/", "https://github.com/ipsss2/GameSense" ], "search_text": "cultivating gaming sense for yourself: making vlms gaming experts cultivating gaming sense for yourself 2025-03 acl 2025 harness vlm codes modules for game playing player" }, { "id": "work-82c779e95f93", "name": "CrossWordBench", "title": "CrossWordBench: Evaluating the Reasoning Capabilities of LLMs and LVLMs with Controllable Puzzle Generation", "date": "2025-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "COLM 2025", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Crossword Puzzles", "LLM+LVLM", "Controllable Generation", "Grid Constraints" ], "keywords_by_category": { "Benchmark": "Crossword Puzzles, LLM+LVLM, Controllable Generation, Grid Constraints" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=bJCQMKwPVq" }, { "label": "Code", "url": "https://github.com/SeanLeng1/CrossWordBench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 333, "date": "2025-03", "title": "CrossWordBench: Evaluating the Reasoning Capabilities of LLMs and LVLMs with Controllable Puzzle Generation", "venue": "COLM 2025", "keywords": "Crossword Puzzles, LLM+LVLM, Controllable Generation, Grid Constraints", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=bJCQMKwPVq" }, { "label": "Code", "url": "https://github.com/SeanLeng1/CrossWordBench" } ] } ], "sources": [ "https://openreview.net/forum?id=bJCQMKwPVq", "https://github.com/SeanLeng1/CrossWordBench" ], "search_text": "crosswordbench: evaluating the reasoning capabilities of llms and lvlms with controllable puzzle generation crosswordbench 2025-03 colm 2025 benchmark crossword puzzles llm+lvlm controllable generation grid constraints player" }, { "id": "work-cad4a0aea846", "name": "CombatVLA", "title": "CombatVLA: An Efficient Vision-Language-Action Model for Combat Tasks in 3D Action Role-Playing Games", "date": "2025-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICCV 2025", "categories": [ "Model" ], "primary_category": null, "keywords": [ "VLA", "SFT", "Combat ARPG" ], "keywords_by_category": { "Model": "VLA, SFT, Combat ARPG" }, "links": [ { "label": "Paper", "url": "https://openaccess.thecvf.com/content/ICCV2025/html/Chen_CombatVLA_An_Efficient_Vision-Language-Action_Model_for_Combat_Tasks_in_3D_ICCV_2025_paper.html" }, { "label": "Code", "url": "https://github.com/ChenVoid/CombatVLA" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 143, "date": "2025-03", "title": "CombatVLA: An Efficient Vision-Language-Action Model for Combat Tasks in 3D Action Role-Playing Games", "venue": "ICCV 2025", "keywords": "VLA, SFT, Combat ARPG", "links": [ { "label": "Paper", "url": "https://openaccess.thecvf.com/content/ICCV2025/html/Chen_CombatVLA_An_Efficient_Vision-Language-Action_Model_for_Combat_Tasks_in_3D_ICCV_2025_paper.html" }, { "label": "Code", "url": "https://github.com/ChenVoid/CombatVLA" } ] } ], "sources": [ "https://openaccess.thecvf.com/content/ICCV2025/html/Chen_CombatVLA_An_Efficient_Vision-Language-Action_Model_for_Combat_Tasks_in_3D_ICCV_2025_paper.html", "https://github.com/ChenVoid/CombatVLA" ], "search_text": "combatvla: an efficient vision-language-action model for combat tasks in 3d action role-playing games combatvla 2025-03 iccv 2025 model vla sft combat arpg player" }, { "id": "arxiv-2503.05383", "name": "AVA", "title": "AVA: Attentive VLM Agent for Mastering StarCraft II", "date": "2025-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "StarCraft II", "Vision-language Model", "Multimodal Decision", "RAG" ], "keywords_by_category": { "Benchmark": "StarCraft II, Vision-language Model, Multimodal Decision, RAG" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2503.05383" }, { "label": "Code", "url": "https://github.com/camel-ai/VLM-Play-StarCraft2" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 336, "date": "2025-03", "title": "AVA: Attentive VLM Agent for Mastering StarCraft II", "venue": "arXiv", "keywords": "StarCraft II, Vision-language Model, Multimodal Decision, RAG", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2503.05383" }, { "label": "Code", "url": "https://github.com/camel-ai/VLM-Play-StarCraft2" } ] } ], "sources": [ "https://arxiv.org/abs/2503.05383", "https://github.com/camel-ai/VLM-Play-StarCraft2" ], "search_text": "ava: attentive vlm agent for mastering starcraft ii ava 2025-03 arxiv benchmark starcraft ii vision-language model multimodal decision rag player" }, { "id": "work-58c0744afe0a", "name": "Are Large Vision Language Models Good Game Players?", "title": "Are Large Vision Language Models Good Game Players?", "date": "2025-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2025", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "LVLM Evaluation", "Game-Based Assessment", "Cognitive Reasoning" ], "keywords_by_category": { "Benchmark": "LVLM Evaluation, Game-Based Assessment, Cognitive Reasoning" }, "links": [ { "label": "Paper", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/27881a19f100fdbf57f0ba1c3d499b08-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/xinke-wang/LVLM-Playground" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 337, "date": "2025-03", "title": "Are Large Vision Language Models Good Game Players?", "venue": "ICLR 2025", "keywords": "LVLM Evaluation, Game-Based Assessment, Cognitive Reasoning", "links": [ { "label": "Paper", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/27881a19f100fdbf57f0ba1c3d499b08-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/xinke-wang/LVLM-Playground" } ] } ], "sources": [ "https://proceedings.iclr.cc/paper_files/paper/2025/hash/27881a19f100fdbf57f0ba1c3d499b08-Abstract-Conference.html", "https://github.com/xinke-wang/LVLM-Playground" ], "search_text": "are large vision language models good game players? are large vision language models good game players? 2025-03 iclr 2025 benchmark lvlm evaluation game-based assessment cognitive reasoning player" }, { "id": "work-bb37e9217614", "name": "World and Human Action Models towards gameplay ideation", "title": "World and Human Action Models towards gameplay ideation", "date": "2025-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "Nature", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Gameplay Ideation", "Human Action Modeling", "Consistency/Diversity/Persistency" ], "keywords_by_category": { "Model": "Gameplay Ideation, Human Action Modeling, Consistency/Diversity/Persistency" }, "links": [ { "label": "Paper", "url": "https://www.nature.com/articles/s41586-025-08600-3" }, { "label": "Code", "url": "https://huggingface.co/microsoft/wham" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 145, "date": "2025-02", "title": "World and Human Action Models towards gameplay ideation", "venue": "Nature", "keywords": "Gameplay Ideation, Human Action Modeling, Consistency/Diversity/Persistency", "links": [ { "label": "Paper", "url": "https://www.nature.com/articles/s41586-025-08600-3" }, { "label": "Code", "url": "https://huggingface.co/microsoft/wham" } ] } ], "sources": [ "https://www.nature.com/articles/s41586-025-08600-3", "https://huggingface.co/microsoft/wham" ], "search_text": "world and human action models towards gameplay ideation world and human action models towards gameplay ideation 2025-02 nature model gameplay ideation human action modeling consistency/diversity/persistency supporting environment" }, { "id": "arxiv-2502.13388", "name": "Reflection of Episodes", "title": "Reflection of Episodes: Learning to Play Game from Expert and Self Experiences", "date": "2025-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "Reflection", "Key frame", "Text StarCraft II" ], "keywords_by_category": { "Harness": "Reflection, Key frame, Text StarCraft II" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2502.13388" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 210, "date": "2025-02", "title": "Reflection of Episodes: Learning to Play Game from Expert and Self Experiences", "venue": "arXiv", "keywords": "Reflection, Key frame, Text StarCraft II", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2502.13388" } ] } ], "sources": [ "https://arxiv.org/abs/2502.13388" ], "search_text": "reflection of episodes: learning to play game from expert and self experiences reflection of episodes 2025-02 arxiv harness reflection key frame text starcraft ii player" }, { "id": "work-f56dd205cdad", "name": "Learning Strategic Language Agents in the Werewolf Game with Iterative Latent Space Policy Optimization", "title": "Learning Strategic Language Agents in the Werewolf Game with Iterative Latent Space Policy Optimization", "date": "2025-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICML 2025", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Werewolf", "Latent Space Policy", "CFR", "DPO", "Strategic Communication" ], "keywords_by_category": { "Model": "Werewolf, Latent Space Policy, CFR, DPO, Strategic Communication" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=N2mOBiSqhc" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 146, "date": "2025-02", "title": "Learning Strategic Language Agents in the Werewolf Game with Iterative Latent Space Policy Optimization", "venue": "ICML 2025", "keywords": "Werewolf, Latent Space Policy, CFR, DPO, Strategic Communication", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=N2mOBiSqhc" } ] } ], "sources": [ "https://openreview.net/forum?id=N2mOBiSqhc" ], "search_text": "learning strategic language agents in the werewolf game with iterative latent space policy optimization learning strategic language agents in the werewolf game with iterative latent space policy optimization 2025-02 icml 2025 model werewolf latent space policy cfr dpo strategic communication player" }, { "id": "work-6f2f97070e98", "name": "Game Theory Meets Large Language Models", "title": "Game Theory Meets Large Language Models: A Systematic Survey", "date": "2025-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "IJCAI 2025", "categories": [ "Survey" ], "primary_category": null, "keywords": [], "keywords_by_category": { "Survey": "" }, "links": [ { "label": "Paper", "url": "https://www.ijcai.org/proceedings/2025/1184" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "survey", "source_rows": [ { "section": "Survey", "line": 375, "date": "2025-02", "title": "Game Theory Meets Large Language Models: A Systematic Survey", "venue": "IJCAI 2025", "keywords": "", "links": [ { "label": "Paper", "url": "https://www.ijcai.org/proceedings/2025/1184" } ] } ], "sources": [ "https://www.ijcai.org/proceedings/2025/1184" ], "search_text": "game theory meets large language models: a systematic survey game theory meets large language models 2025-02 ijcai 2025 survey survey" }, { "id": "work-cabec571eb6f", "name": "Collab-Overcooked", "title": "Collab-Overcooked: Benchmarking and Evaluating Large Language Models as Collaborative Agents", "date": "2025-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "EMNLP 2025", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Overcooked", "Cooperative Agents", "Collaboration", "Coordination" ], "keywords_by_category": { "Benchmark": "Overcooked, Cooperative Agents, Collaboration, Coordination" }, "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.emnlp-main.249/" }, { "label": "Code", "url": "https://github.com/YusaeMeow/Collab-Overcooked" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 338, "date": "2025-02", "title": "Collab-Overcooked: Benchmarking and Evaluating Large Language Models as Collaborative Agents", "venue": "EMNLP 2025", "keywords": "Overcooked, Cooperative Agents, Collaboration, Coordination", "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.emnlp-main.249/" }, { "label": "Code", "url": "https://github.com/YusaeMeow/Collab-Overcooked" } ] } ], "sources": [ "https://aclanthology.org/2025.emnlp-main.249/", "https://github.com/YusaeMeow/Collab-Overcooked" ], "search_text": "collab-overcooked: benchmarking and evaluating large language models as collaborative agents collab-overcooked 2025-02 emnlp 2025 benchmark overcooked cooperative agents collaboration coordination player" }, { "id": "arxiv-2501.12326", "name": "UI-TARS", "title": "UI-TARS: Pioneering Automated GUI Interaction with Native Agents", "date": "2025-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "SFT", "System-2 Reasoning", "GUI Agent" ], "keywords_by_category": { "Model": "SFT, System-2 Reasoning, GUI Agent" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2501.12326" }, { "label": "Code", "url": "https://github.com/bytedance/UI-TARS" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 148, "date": "2025-01", "title": "UI-TARS: Pioneering Automated GUI Interaction with Native Agents", "venue": "arXiv", "keywords": "SFT, System-2 Reasoning, GUI Agent", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2501.12326" }, { "label": "Code", "url": "https://github.com/bytedance/UI-TARS" } ] } ], "sources": [ "https://arxiv.org/abs/2501.12326", "https://github.com/bytedance/UI-TARS" ], "search_text": "ui-tars: pioneering automated gui interaction with native agents ui-tars 2025-01 arxiv model sft system-2 reasoning gui agent player" }, { "id": "work-90628d2632f3", "name": "PokerBench", "title": "PokerBench: Training Large Language Models to Become Professional Poker Players", "date": "2025-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "AAAI 2025", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Poker", "Imperfect Information", "Strategic Decision Making", "Game Theory" ], "keywords_by_category": { "Benchmark": "Poker, Imperfect Information, Strategic Decision Making, Game Theory" }, "links": [ { "label": "Paper", "url": "https://ojs.aaai.org/index.php/AAAI/article/view/34814" }, { "label": "Code", "url": "https://github.com/pokerllm/pokerbench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 340, "date": "2025-01", "title": "PokerBench: Training Large Language Models to Become Professional Poker Players", "venue": "AAAI 2025", "keywords": "Poker, Imperfect Information, Strategic Decision Making, Game Theory", "links": [ { "label": "Paper", "url": "https://ojs.aaai.org/index.php/AAAI/article/view/34814" }, { "label": "Code", "url": "https://github.com/pokerllm/pokerbench" } ] } ], "sources": [ "https://ojs.aaai.org/index.php/AAAI/article/view/34814", "https://github.com/pokerllm/pokerbench" ], "search_text": "pokerbench: training large language models to become professional poker players pokerbench 2025-01 aaai 2025 benchmark poker imperfect information strategic decision making game theory player" }, { "id": "work-58b0b8c7bfd7", "name": "GameFactory", "title": "GameFactory: Creating New Games with Generative Interactive Videos", "date": "2025-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICCV 2025", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Scene-Generalizable Control", "Open-Domain Game Generation", "Multi-Phase Training" ], "keywords_by_category": { "Model": "Scene-Generalizable Control, Open-Domain Game Generation, Multi-Phase Training" }, "links": [ { "label": "Paper", "url": "https://openaccess.thecvf.com/content/ICCV2025/html/Yu_GameFactory_Creating_New_Games_with_Generative_Interactive_Videos_ICCV_2025_paper.html" }, { "label": "Code", "url": "https://github.com/KlingAIResearch/GameFactory" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 147, "date": "2025-01", "title": "GameFactory: Creating New Games with Generative Interactive Videos", "venue": "ICCV 2025", "keywords": "Scene-Generalizable Control, Open-Domain Game Generation, Multi-Phase Training", "links": [ { "label": "Paper", "url": "https://openaccess.thecvf.com/content/ICCV2025/html/Yu_GameFactory_Creating_New_Games_with_Generative_Interactive_Videos_ICCV_2025_paper.html" }, { "label": "Code", "url": "https://github.com/KlingAIResearch/GameFactory" } ] } ], "sources": [ "https://openaccess.thecvf.com/content/ICCV2025/html/Yu_GameFactory_Creating_New_Games_with_Generative_Interactive_Videos_ICCV_2025_paper.html", "https://github.com/KlingAIResearch/GameFactory" ], "search_text": "gamefactory: creating new games with generative interactive videos gamefactory 2025-01 iccv 2025 model scene-generalizable control open-domain game generation multi-phase training supporting environment" }, { "id": "work-e0b51f652bcf", "name": "Complete Chess Games Enable LLM Become A Chess Master", "title": "Complete Chess Games Enable LLM Become A Chess Master", "date": "2025-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NAACL 2025", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "ChessLLM", "FEN Notation", "Complete Game Training", "Strategic Reasoning" ], "keywords_by_category": { "Benchmark": "ChessLLM, FEN Notation, Complete Game Training, Strategic Reasoning" }, "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.naacl-short.1/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 339, "date": "2025-01", "title": "Complete Chess Games Enable LLM Become A Chess Master", "venue": "NAACL 2025", "keywords": "ChessLLM, FEN Notation, Complete Game Training, Strategic Reasoning", "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.naacl-short.1/" } ] } ], "sources": [ "https://aclanthology.org/2025.naacl-short.1/" ], "search_text": "complete chess games enable llm become a chess master complete chess games enable llm become a chess master 2025-01 naacl 2025 benchmark chessllm fen notation complete game training strategic reasoning player" }, { "id": "arxiv-2412.05255", "name": "TeamCraft", "title": "TeamCraft: A Benchmark for Multi-Modal Multi-Agent Systems in Minecraft", "date": "2024-12", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Minecraft", "Multi-agent", "Multi-modal", "55", "000 Tasks", "Imitation Learning" ], "keywords_by_category": { "Benchmark": "Minecraft, Multi-agent, Multi-modal, 55,000 Tasks, Imitation Learning" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2412.05255v1" }, { "label": "Code", "url": "https://github.com/teamcraft-bench/teamcraft" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 342, "date": "2024-12", "title": "TeamCraft: A Benchmark for Multi-Modal Multi-Agent Systems in Minecraft", "venue": "arXiv", "keywords": "Minecraft, Multi-agent, Multi-modal, 55,000 Tasks, Imitation Learning", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2412.05255v1" }, { "label": "Code", "url": "https://github.com/teamcraft-bench/teamcraft" } ] } ], "sources": [ "https://arxiv.org/abs/2412.05255v1", "https://github.com/teamcraft-bench/teamcraft" ], "search_text": "teamcraft: a benchmark for multi-modal multi-agent systems in minecraft teamcraft 2024-12 arxiv benchmark minecraft multi-agent multi-modal 55 000 tasks imitation learning player" }, { "id": "arxiv-2412.00887", "name": "Playable Game Generation", "title": "Playable Game Generation", "date": "2024-12", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Autoregressive DiT", "Playability Evaluation", "Real-Time Mechanics Simulation" ], "keywords_by_category": { "Model": "Autoregressive DiT, Playability Evaluation, Real-Time Mechanics Simulation" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2412.00887" }, { "label": "Code", "url": "https://github.com/GreatX3/Playable-Game-Generation" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 149, "date": "2024-12", "title": "Playable Game Generation", "venue": "arXiv", "keywords": "Autoregressive DiT, Playability Evaluation, Real-Time Mechanics Simulation", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2412.00887" }, { "label": "Code", "url": "https://github.com/GreatX3/Playable-Game-Generation" } ] } ], "sources": [ "https://arxiv.org/abs/2412.00887", "https://github.com/GreatX3/Playable-Game-Generation" ], "search_text": "playable game generation playable game generation 2024-12 arxiv model autoregressive dit playability evaluation real-time mechanics simulation supporting environment" }, { "id": "work-677898f8219a", "name": "GAMEBoT", "title": "GAMEBoT: Transparent Assessment of LLM Reasoning in Games", "date": "2024-12", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ACL 2025", "categories": [ "Harness", "Benchmark" ], "primary_category": null, "keywords": [ "CoT", "Transparent Metrics", "Modular Subproblems", "17 LLMs", "8 Games" ], "keywords_by_category": { "Harness": "CoT", "Benchmark": "Transparent Metrics, Modular Subproblems, 17 LLMs, 8 Games" }, "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.acl-long.378/" }, { "label": "Code", "url": "https://github.com/Visual-AI/GAMEBoT" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 211, "date": "2024-12", "title": "GAMEBoT: Transparent Assessment of LLM Reasoning in Games", "venue": "ACL 2025", "keywords": "CoT", "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.acl-long.378/" }, { "label": "Code", "url": "https://github.com/Visual-AI/GAMEBoT" } ] }, { "section": "Benchmark", "line": 341, "date": "2024-12", "title": "GAMEBoT: Transparent Assessment of LLM Reasoning in Games", "venue": "ACL 2025", "keywords": "Transparent Metrics, Modular Subproblems, 17 LLMs, 8 Games", "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.acl-long.378/" }, { "label": "Code", "url": "https://github.com/Visual-AI/GAMEBoT" } ] } ], "sources": [ "https://aclanthology.org/2025.acl-long.378/", "https://github.com/Visual-AI/GAMEBoT" ], "search_text": "gamebot: transparent assessment of llm reasoning in games gamebot 2024-12 acl 2025 harness benchmark cot transparent metrics modular subproblems 17 llms 8 games player" }, { "id": "work-639877e9db2a", "name": "GameArena", "title": "GameArena: Evaluating LLM Reasoning through Live Computer Games", "date": "2024-12", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2025", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Live Computer Games", "Human-LLM Gameplay", "2000+ Sessions" ], "keywords_by_category": { "Benchmark": "Live Computer Games, Human-LLM Gameplay, 2000+ Sessions" }, "links": [ { "label": "Paper", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/520416e27d3b0cef3cd70a083e2991c7-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/lmgame-org/ai-space-escape-engine" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 343, "date": "2024-12", "title": "GameArena: Evaluating LLM Reasoning through Live Computer Games", "venue": "ICLR 2025", "keywords": "Live Computer Games, Human-LLM Gameplay, 2000+ Sessions", "links": [ { "label": "Paper", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/520416e27d3b0cef3cd70a083e2991c7-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/lmgame-org/ai-space-escape-engine" } ] } ], "sources": [ "https://proceedings.iclr.cc/paper_files/paper/2025/hash/520416e27d3b0cef3cd70a083e2991c7-Abstract-Conference.html", "https://github.com/lmgame-org/ai-space-escape-engine" ], "search_text": "gamearena: evaluating llm reasoning through live computer games gamearena 2024-12 iclr 2025 benchmark live computer games human-llm gameplay 2000+ sessions player" }, { "id": "work-8e1f44174fd3", "name": "A Survey on Large Language Model-Based Social Agents in Game-Theoretic Scenarios", "title": "A Survey on Large Language Model-Based Social Agents in Game-Theoretic Scenarios", "date": "2024-12", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "TMLR 2025", "categories": [ "Survey" ], "primary_category": null, "keywords": [], "keywords_by_category": { "Survey": "" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=CsoSWpR5xC" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "survey", "source_rows": [ { "section": "Survey", "line": 376, "date": "2024-12", "title": "A Survey on Large Language Model-Based Social Agents in Game-Theoretic Scenarios", "venue": "TMLR 2025", "keywords": "", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=CsoSWpR5xC" } ] } ], "sources": [ "https://openreview.net/forum?id=CsoSWpR5xC" ], "search_text": "a survey on large language model-based social agents in game-theoretic scenarios a survey on large language model-based social agents in game-theoretic scenarios 2024-12 tmlr 2025 survey survey" }, { "id": "work-ccdce69a675a", "name": "GameGen-X", "title": "GameGen-X: Interactive Open-world Game Video Generation", "date": "2024-11", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2025", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Open-World Game Video Generation", "Multi-Modal Control", "Instruction Tuning" ], "keywords_by_category": { "Model": "Open-World Game Video Generation, Multi-Modal Control, Instruction Tuning" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=8VG8tpPZhe" }, { "label": "Code", "url": "https://github.com/GameGen-X/GameGen-X" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 150, "date": "2024-11", "title": "GameGen-X: Interactive Open-world Game Video Generation", "venue": "ICLR 2025", "keywords": "Open-World Game Video Generation, Multi-Modal Control, Instruction Tuning", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=8VG8tpPZhe" }, { "label": "Code", "url": "https://github.com/GameGen-X/GameGen-X" } ] } ], "sources": [ "https://openreview.net/forum?id=8VG8tpPZhe", "https://github.com/GameGen-X/GameGen-X" ], "search_text": "gamegen-x: interactive open-world game video generation gamegen-x 2024-11 iclr 2025 model open-world game video generation multi-modal control instruction tuning supporting environment" }, { "id": "work-478c649f5ed7", "name": "BALROG", "title": "BALROG: Benchmarking Agentic LLM and VLM Reasoning On Games", "date": "2024-11", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2025", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Agentic Capabilities", "Long-Horizon RL", "Unified Interface" ], "keywords_by_category": { "Benchmark": "Agentic Capabilities, Long-Horizon RL, Unified Interface" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=fp6t3F669F" }, { "label": "Project", "url": "https://balrogai.com/" }, { "label": "Code", "url": "https://github.com/balrog-ai/BALROG" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 344, "date": "2024-11", "title": "BALROG: Benchmarking Agentic LLM and VLM Reasoning On Games", "venue": "ICLR 2025", "keywords": "Agentic Capabilities, Long-Horizon RL, Unified Interface", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=fp6t3F669F" }, { "label": "Project", "url": "https://balrogai.com/" }, { "label": "Code", "url": "https://github.com/balrog-ai/BALROG" } ] } ], "sources": [ "https://openreview.net/forum?id=fp6t3F669F", "https://balrogai.com/", "https://github.com/balrog-ai/BALROG" ], "search_text": "balrog: benchmarking agentic llm and vlm reasoning on games balrog 2024-11 iclr 2025 benchmark agentic capabilities long-horizon rl unified interface player" }, { "id": "arxiv-2410.10479", "name": "TMGBench", "title": "TMGBench: A Systematic Game Benchmark for Evaluating Strategic Reasoning Abilities of LLMs", "date": "2024-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Game Theory", "Comprehensive Coverage", "Diverse Scenarios" ], "keywords_by_category": { "Benchmark": "Game Theory, Comprehensive Coverage, Diverse Scenarios" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2410.10479" }, { "label": "Code", "url": "https://github.com/PinkEx/TMGBench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 347, "date": "2024-10", "title": "TMGBench: A Systematic Game Benchmark for Evaluating Strategic Reasoning Abilities of LLMs", "venue": "arXiv", "keywords": "Game Theory, Comprehensive Coverage, Diverse Scenarios", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2410.10479" }, { "label": "Code", "url": "https://github.com/PinkEx/TMGBench" } ] } ], "sources": [ "https://arxiv.org/abs/2410.10479", "https://github.com/PinkEx/TMGBench" ], "search_text": "tmgbench: a systematic game benchmark for evaluating strategic reasoning abilities of llms tmgbench 2024-10 arxiv benchmark game theory comprehensive coverage diverse scenarios player" }, { "id": "work-d3298349ed3f", "name": "Scaling Offline Model-Based RL via Jointly-Optimized World-Action Model Pretraining", "title": "Scaling Offline Model-Based RL via Jointly-Optimized World-Action Model Pretraining", "date": "2024-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2025", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Joint World-Action Pretraining", "Offline Model-Based RL", "Multi-Game Atari Transfer" ], "keywords_by_category": { "Model": "Joint World-Action Pretraining, Offline Model-Based RL, Multi-Game Atari Transfer" }, "links": [ { "label": "Paper", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/689cffc97600f9deb8374fc8fa918b8e-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/CJReinforce/JOWA" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 151, "date": "2024-10", "title": "Scaling Offline Model-Based RL via Jointly-Optimized World-Action Model Pretraining", "venue": "ICLR 2025", "keywords": "Joint World-Action Pretraining, Offline Model-Based RL, Multi-Game Atari Transfer", "links": [ { "label": "Paper", "url": "https://proceedings.iclr.cc/paper_files/paper/2025/hash/689cffc97600f9deb8374fc8fa918b8e-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/CJReinforce/JOWA" } ] } ], "sources": [ "https://proceedings.iclr.cc/paper_files/paper/2025/hash/689cffc97600f9deb8374fc8fa918b8e-Abstract-Conference.html", "https://github.com/CJReinforce/JOWA" ], "search_text": "scaling offline model-based rl via jointly-optimized world-action model pretraining scaling offline model-based rl via jointly-optimized world-action model pretraining 2024-10 iclr 2025 model joint world-action pretraining offline model-based rl multi-game atari transfer player" }, { "id": "arxiv-2410.17856", "name": "ROCKET-1", "title": "ROCKET-1: Mastering Open-World Interaction with Visual-Temporal Context Prompting", "date": "2024-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "CVPR 2025", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Hierarchical VLM", "Visual-Temporal Prompting" ], "keywords_by_category": { "Model": "Hierarchical VLM, Visual-Temporal Prompting" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2410.17856" }, { "label": "Code", "url": "https://github.com/CraftJarvis/ROCKET-1" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 152, "date": "2024-10", "title": "ROCKET-1: Mastering Open-World Interaction with Visual-Temporal Context Prompting", "venue": "CVPR 2025", "keywords": "Hierarchical VLM, Visual-Temporal Prompting", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2410.17856" }, { "label": "Code", "url": "https://github.com/CraftJarvis/ROCKET-1" } ] } ], "sources": [ "https://arxiv.org/abs/2410.17856", "https://github.com/CraftJarvis/ROCKET-1" ], "search_text": "rocket-1: mastering open-world interaction with visual-temporal context prompting rocket-1 2024-10 cvpr 2025 model hierarchical vlm visual-temporal prompting player" }, { "id": "work-f33dbb80396f", "name": "Procedural Content Generation in Games", "title": "Procedural Content Generation in Games: A Survey with Insights on Emerging LLM Integration", "date": "2024-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "AIIDE 2024", "categories": [ "Survey" ], "primary_category": null, "keywords": [], "keywords_by_category": { "Survey": "" }, "links": [ { "label": "Paper", "url": "https://ojs.aaai.org/index.php/AIIDE/article/view/31877" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "supporting environment", "source_rows": [ { "section": "Survey", "line": 377, "date": "2024-10", "title": "Procedural Content Generation in Games: A Survey with Insights on Emerging LLM Integration", "venue": "AIIDE 2024", "keywords": "", "links": [ { "label": "Paper", "url": "https://ojs.aaai.org/index.php/AIIDE/article/view/31877" } ] } ], "sources": [ "https://ojs.aaai.org/index.php/AIIDE/article/view/31877" ], "search_text": "procedural content generation in games: a survey with insights on emerging llm integration procedural content generation in games 2024-10 aiide 2024 survey supporting environment" }, { "id": "arxiv-2410.06555", "name": "ING-VP", "title": "ING-VP: MLLMs cannot Play Easy Vision-based Games Yet", "date": "2024-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Vision-based Games", "Spatial Reasoning", "Multi-step Planning", "MLLM Evaluation" ], "keywords_by_category": { "Benchmark": "Vision-based Games, Spatial Reasoning, Multi-step Planning, MLLM Evaluation" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2410.06555" }, { "label": "Code", "url": "https://github.com/Thisisus7/ING-VP" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 345, "date": "2024-10", "title": "ING-VP: MLLMs cannot Play Easy Vision-based Games Yet", "venue": "arXiv", "keywords": "Vision-based Games, Spatial Reasoning, Multi-step Planning, MLLM Evaluation", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2410.06555" }, { "label": "Code", "url": "https://github.com/Thisisus7/ING-VP" } ] } ], "sources": [ "https://arxiv.org/abs/2410.06555", "https://github.com/Thisisus7/ING-VP" ], "search_text": "ing-vp: mllms cannot play easy vision-based games yet ing-vp 2024-10 arxiv benchmark vision-based games spatial reasoning multi-step planning mllm evaluation player" }, { "id": "work-66155807f257", "name": "GameTraversalBenchmark", "title": "GameTraversalBenchmark: Evaluating Planning Abilities Of Large Language Models Through Traversing 2D Game Maps", "date": "2024-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NeurIPS 2024", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "2D Grid Maps", "Planning Benchmark", "Path Traversal", "Multi-step Reasoning" ], "keywords_by_category": { "Benchmark": "2D Grid Maps, Planning Benchmark, Path Traversal, Multi-step Reasoning" }, "links": [ { "label": "Paper", "url": "https://papers.nips.cc/paper_files/paper/2024/hash/3852c8254bc6d904c09db9921157f59b-Abstract-Datasets_and_Benchmarks_Track.html" }, { "label": "Code", "url": "https://github.com/umair-nasir14/Game-Traversal-Benchmark" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 346, "date": "2024-10", "title": "GameTraversalBenchmark: Evaluating Planning Abilities Of Large Language Models Through Traversing 2D Game Maps", "venue": "NeurIPS 2024", "keywords": "2D Grid Maps, Planning Benchmark, Path Traversal, Multi-step Reasoning", "links": [ { "label": "Paper", "url": "https://papers.nips.cc/paper_files/paper/2024/hash/3852c8254bc6d904c09db9921157f59b-Abstract-Datasets_and_Benchmarks_Track.html" }, { "label": "Code", "url": "https://github.com/umair-nasir14/Game-Traversal-Benchmark" } ] } ], "sources": [ "https://papers.nips.cc/paper_files/paper/2024/hash/3852c8254bc6d904c09db9921157f59b-Abstract-Datasets_and_Benchmarks_Track.html", "https://github.com/umair-nasir14/Game-Traversal-Benchmark" ], "search_text": "gametraversalbenchmark: evaluating planning abilities of large language models through traversing 2d game maps gametraversalbenchmark 2024-10 neurips 2024 benchmark 2d grid maps planning benchmark path traversal multi-step reasoning player" }, { "id": "work-ad0bd8561882", "name": "StarCraft II Arena", "title": "StarCraft II Arena: Evaluating LLMs in Strategic Planning, Real-Time Decision Making, and Adaptability", "date": "2024-09", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "StarCraft II", "Real-time Strategy", "Strategic Planning", "Adaptability" ], "keywords_by_category": { "Benchmark": "StarCraft II, Real-time Strategy, Strategic Planning, Adaptability" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=o3V7OuPxu4" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 349, "date": "2024-09", "title": "StarCraft II Arena: Evaluating LLMs in Strategic Planning, Real-Time Decision Making, and Adaptability", "venue": "arXiv", "keywords": "StarCraft II, Real-time Strategy, Strategic Planning, Adaptability", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=o3V7OuPxu4" } ] } ], "sources": [ "https://openreview.net/forum?id=o3V7OuPxu4" ], "search_text": "starcraft ii arena: evaluating llms in strategic planning, real-time decision making, and adaptability starcraft ii arena 2024-09 arxiv benchmark starcraft ii real-time strategy strategic planning adaptability player" }, { "id": "work-a47b32f282cb", "name": "Mars", "title": "Mars: Situated Inductive Reasoning in an Open-World Environment", "date": "2024-09", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NeurIPS 2024 (D&B)", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Counter-commonsense", "Reflection", "Open-world Environment" ], "keywords_by_category": { "Benchmark": "Counter-commonsense, Reflection, Open-world Environment" }, "links": [ { "label": "Paper", "url": "https://neurips.cc/virtual/2024/poster/97857" }, { "label": "Code", "url": "https://github.com/XiaojuanTang/Mars" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 350, "date": "2024-09", "title": "Mars: Situated Inductive Reasoning in an Open-World Environment", "venue": "NeurIPS 2024 (D&B)", "keywords": "Counter-commonsense, Reflection, Open-world Environment", "links": [ { "label": "Paper", "url": "https://neurips.cc/virtual/2024/poster/97857" }, { "label": "Code", "url": "https://github.com/XiaojuanTang/Mars" } ] } ], "sources": [ "https://neurips.cc/virtual/2024/poster/97857", "https://github.com/XiaojuanTang/Mars" ], "search_text": "mars: situated inductive reasoning in an open-world environment mars 2024-09 neurips 2024 (d&b) benchmark counter-commonsense reflection open-world environment player" }, { "id": "arxiv-2409.12889", "name": "Can VLMs Play Action Role-Playing Games? Take Black Myth Wukong as a Study Case", "title": "Can VLMs Play Action Role-Playing Games? Take Black Myth Wukong as a Study Case", "date": "2024-09", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv / NeurIPS 2024 Workshop", "categories": [ "Harness", "Benchmark" ], "primary_category": null, "keywords": [ "Reflection", "Workflow", "Human trajectory", "ARPG", "Non-API Interaction", "Visual-Centric Control", "Keyboard/Mouse Actions", "Success-Rate Evaluation" ], "keywords_by_category": { "Harness": "Reflection, Workflow, Human trajectory", "Benchmark": "ARPG, Non-API Interaction, Visual-Centric Control, Keyboard/Mouse Actions, Success-Rate Evaluation" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2409.12889" }, { "label": "Project", "url": "https://varp-agent.github.io/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 212, "date": "2024-09", "title": "Can VLMs Play Action Role-Playing Games? Take Black Myth Wukong as a Study Case", "venue": "arXiv", "keywords": "Reflection, Workflow, Human trajectory", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2409.12889" }, { "label": "Project", "url": "https://varp-agent.github.io/" } ] }, { "section": "Benchmark", "line": 348, "date": "2024-09", "title": "Can VLMs Play Action Role-Playing Games? Take Black Myth Wukong as a Study Case", "venue": "NeurIPS 2024 Workshop", "keywords": "ARPG, Non-API Interaction, Visual-Centric Control, Keyboard/Mouse Actions, Success-Rate Evaluation", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2409.12889" }, { "label": "Project", "url": "https://varp-agent.github.io/" } ] } ], "sources": [ "https://arxiv.org/abs/2409.12889", "https://varp-agent.github.io/" ], "search_text": "can vlms play action role-playing games? take black myth wukong as a study case can vlms play action role-playing games? take black myth wukong as a study case 2024-09 arxiv / neurips 2024 workshop harness benchmark reflection workflow human trajectory arpg non-api interaction visual-centric control keyboard/mouse actions success-rate evaluation player" }, { "id": "work-9726682ec030", "name": "Diffusion Models Are Real-Time Game Engines", "title": "Diffusion Models Are Real-Time Game Engines", "date": "2024-08", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2025", "categories": [ "Model" ], "primary_category": null, "keywords": [ "DOOM Neural Game Engine", "Diffusion World Model", "RL Gameplay Data" ], "keywords_by_category": { "Model": "DOOM Neural Game Engine, Diffusion World Model, RL Gameplay Data" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=P8pqeEkn1H" }, { "label": "Project", "url": "https://gamengen.github.io/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 153, "date": "2024-08", "title": "Diffusion Models Are Real-Time Game Engines", "venue": "ICLR 2025", "keywords": "DOOM Neural Game Engine, Diffusion World Model, RL Gameplay Data", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=P8pqeEkn1H" }, { "label": "Project", "url": "https://gamengen.github.io/" } ] } ], "sources": [ "https://openreview.net/forum?id=P8pqeEkn1H", "https://gamengen.github.io/" ], "search_text": "diffusion models are real-time game engines diffusion models are real-time game engines 2024-08 iclr 2025 model doom neural game engine diffusion world model rl gameplay data supporting environment" }, { "id": "arxiv-2408.15950", "name": "Atari-GPT", "title": "Atari-GPT: Benchmarking Multimodal Large Language Models as Low-Level Policies in Atari Games", "date": "2024-08", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Low-Level Policies", "Multimodal LLM", "Atari Control" ], "keywords_by_category": { "Benchmark": "Low-Level Policies, Multimodal LLM, Atari Control" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2408.15950" }, { "label": "Code", "url": "https://github.com/nwayt001/atari-gpt" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 351, "date": "2024-08", "title": "Atari-GPT: Benchmarking Multimodal Large Language Models as Low-Level Policies in Atari Games", "venue": "arXiv", "keywords": "Low-Level Policies, Multimodal LLM, Atari Control", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2408.15950" }, { "label": "Code", "url": "https://github.com/nwayt001/atari-gpt" } ] } ], "sources": [ "https://arxiv.org/abs/2408.15950", "https://github.com/nwayt001/atari-gpt" ], "search_text": "atari-gpt: benchmarking multimodal large language models as low-level policies in atari games atari-gpt 2024-08 arxiv benchmark low-level policies multimodal llm atari control player" }, { "id": "arxiv-2407.13943", "name": "Werewolf Arena", "title": "Werewolf Arena: A Case Study in LLM Evaluation via Social Deduction", "date": "2024-07", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Werewolf", "Social Deduction", "Bidding System", "LLM Arena", "Deception & Persuasion" ], "keywords_by_category": { "Benchmark": "Werewolf, Social Deduction, Bidding System, LLM Arena, Deception & Persuasion" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2407.13943" }, { "label": "Code", "url": "https://github.com/google/werewolf_arena" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 352, "date": "2024-07", "title": "Werewolf Arena: A Case Study in LLM Evaluation via Social Deduction", "venue": "arXiv", "keywords": "Werewolf, Social Deduction, Bidding System, LLM Arena, Deception & Persuasion", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2407.13943" }, { "label": "Code", "url": "https://github.com/google/werewolf_arena" } ] } ], "sources": [ "https://arxiv.org/abs/2407.13943", "https://github.com/google/werewolf_arena" ], "search_text": "werewolf arena: a case study in llm evaluation via social deduction werewolf arena 2024-07 arxiv benchmark werewolf social deduction bidding system llm arena deception & persuasion player" }, { "id": "work-383b7f4c3818", "name": "OmniJARVIS", "title": "OmniJARVIS: Unified Vision-Language-Action Tokenization Enables Open-World Instruction Following Agents", "date": "2024-07", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NeurIPS 2024", "categories": [ "Model" ], "primary_category": null, "keywords": [ "VLA Tokenization", "Imitation Learning" ], "keywords_by_category": { "Model": "VLA Tokenization, Imitation Learning" }, "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/85f1225db986e629289f402c46eff1a4-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/CraftJarvis/OmniJARVIS" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 154, "date": "2024-07", "title": "OmniJARVIS: Unified Vision-Language-Action Tokenization Enables Open-World Instruction Following Agents", "venue": "NeurIPS 2024", "keywords": "VLA Tokenization, Imitation Learning", "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/85f1225db986e629289f402c46eff1a4-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/CraftJarvis/OmniJARVIS" } ] } ], "sources": [ "https://proceedings.neurips.cc/paper_files/paper/2024/hash/85f1225db986e629289f402c46eff1a4-Abstract-Conference.html", "https://github.com/CraftJarvis/OmniJARVIS" ], "search_text": "omnijarvis: unified vision-language-action tokenization enables open-world instruction following agents omnijarvis 2024-07 neurips 2024 model vla tokenization imitation learning player" }, { "id": "arxiv-2407.07796", "name": "Evaluating Large Language Models with Grid-Based Game Competitions", "title": "Evaluating Large Language Models with Grid-Based Game Competitions: An Extensible LLM Benchmark and Leaderboard", "date": "2024-07", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "Journal of Cognitive Systems 2025", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Grid Games", "Tic-Tac-Toe", "Connect Four", "Gomoku", "Leaderboard" ], "keywords_by_category": { "Benchmark": "Grid Games, Tic-Tac-Toe, Connect Four, Gomoku, Leaderboard" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2407.07796" }, { "label": "Code", "url": "https://github.com/research-outcome/LLM-Game-Benchmark" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 353, "date": "2024-07", "title": "Evaluating Large Language Models with Grid-Based Game Competitions: An Extensible LLM Benchmark and Leaderboard", "venue": "Journal of Cognitive Systems 2025", "keywords": "Grid Games, Tic-Tac-Toe, Connect Four, Gomoku, Leaderboard", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2407.07796" }, { "label": "Code", "url": "https://github.com/research-outcome/LLM-Game-Benchmark" } ] } ], "sources": [ "https://arxiv.org/abs/2407.07796", "https://github.com/research-outcome/LLM-Game-Benchmark" ], "search_text": "evaluating large language models with grid-based game competitions: an extensible llm benchmark and leaderboard evaluating large language models with grid-based game competitions 2024-07 journal of cognitive systems 2025 benchmark grid games tic-tac-toe connect four gomoku leaderboard player" }, { "id": "arxiv-2406.06613", "name": "GameBench", "title": "GameBench: Evaluating Strategic Reasoning Abilities of LLM Agents", "date": "2024-06", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "9 Games", "Cross-Domain", "Strategic Reasoning", "GPT-4" ], "keywords_by_category": { "Benchmark": "9 Games, Cross-Domain, Strategic Reasoning, GPT-4" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2406.06613" }, { "label": "Code", "url": "https://github.com/Joshuaclymer/GameBench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 354, "date": "2024-06", "title": "GameBench: Evaluating Strategic Reasoning Abilities of LLM Agents", "venue": "arXiv", "keywords": "9 Games, Cross-Domain, Strategic Reasoning, GPT-4", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2406.06613" }, { "label": "Code", "url": "https://github.com/Joshuaclymer/GameBench" } ] } ], "sources": [ "https://arxiv.org/abs/2406.06613", "https://github.com/Joshuaclymer/GameBench" ], "search_text": "gamebench: evaluating strategic reasoning abilities of llm agents gamebench 2024-06 arxiv benchmark 9 games cross-domain strategic reasoning gpt-4 player" }, { "id": "work-6fba3426fe25", "name": "DigiRL", "title": "DigiRL: Training In-The-Wild Device-Control Agents with Autonomous Reinforcement Learning", "date": "2024-06", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NeurIPS 2024", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Autonomous RL", "VLM Fine-tuning" ], "keywords_by_category": { "Model": "Autonomous RL, VLM Fine-tuning" }, "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/1704ddd0bb89f159dfe609b32c889995-Abstract-Conference.html" }, { "label": "Project", "url": "https://digirl-agent.github.io/" }, { "label": "Code", "url": "https://github.com/DigiRL-agent/digirl" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 155, "date": "2024-06", "title": "DigiRL: Training In-The-Wild Device-Control Agents with Autonomous Reinforcement Learning", "venue": "NeurIPS 2024", "keywords": "Autonomous RL, VLM Fine-tuning", "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/1704ddd0bb89f159dfe609b32c889995-Abstract-Conference.html" }, { "label": "Project", "url": "https://digirl-agent.github.io/" }, { "label": "Code", "url": "https://github.com/DigiRL-agent/digirl" } ] } ], "sources": [ "https://proceedings.neurips.cc/paper_files/paper/2024/hash/1704ddd0bb89f159dfe609b32c889995-Abstract-Conference.html", "https://digirl-agent.github.io/", "https://github.com/DigiRL-agent/digirl" ], "search_text": "digirl: training in-the-wild device-control agents with autonomous reinforcement learning digirl 2024-06 neurips 2024 model autonomous rl vlm fine-tuning player" }, { "id": "arxiv-2406.04208", "name": "Aligning Agents like Large Language Models", "title": "Aligning Agents like Large Language Models", "date": "2024-06", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Agent Alignment", "RLHF-style" ], "keywords_by_category": { "Model": "Agent Alignment, RLHF-style" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2406.04208" }, { "label": "Code", "url": "https://adamjelley.github.io/aligning-agents-like-llms/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 156, "date": "2024-06", "title": "Aligning Agents like Large Language Models", "venue": "arXiv", "keywords": "Agent Alignment, RLHF-style", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2406.04208" }, { "label": "Code", "url": "https://adamjelley.github.io/aligning-agents-like-llms/" } ] } ], "sources": [ "https://arxiv.org/abs/2406.04208", "https://adamjelley.github.io/aligning-agents-like-llms/" ], "search_text": "aligning agents like large language models aligning agents like large language models 2024-06 arxiv model agent alignment rlhf-style player" }, { "id": "work-ef814caf62a1", "name": "Self-Reflection in Large Language Model Agents", "title": "Self-Reflection in Large Language Model Agents: Effects on Problem-Solving Performance", "date": "2024-05", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "FLLM 2024", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "Self-Reflection" ], "keywords_by_category": { "Harness": "Self-Reflection" }, "links": [ { "label": "Paper", "url": "https://ieeexplore.ieee.org/document/10852426" }, { "label": "Code", "url": "https://github.com/matthewrenze/self-reflection" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 213, "date": "2024-05", "title": "Self-Reflection in Large Language Model Agents: Effects on Problem-Solving Performance", "venue": "FLLM 2024", "keywords": "Self-Reflection", "links": [ { "label": "Paper", "url": "https://ieeexplore.ieee.org/document/10852426" }, { "label": "Code", "url": "https://github.com/matthewrenze/self-reflection" } ] } ], "sources": [ "https://ieeexplore.ieee.org/document/10852426", "https://github.com/matthewrenze/self-reflection" ], "search_text": "self-reflection in large language model agents: effects on problem-solving performance self-reflection in large language model agents 2024-05 fllm 2024 harness self-reflection player" }, { "id": "arxiv-2403.11807", "name": "How Far Are We on the Decision-Making of LLMs? Evaluating LLMs' Gaming Ability in Multi-Agent Environments", "title": "How Far Are We on the Decision-Making of LLMs? Evaluating LLMs' Gaming Ability in Multi-Agent Environments", "date": "2024-05", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2025", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "GAMA-Bench", "Game Theory", "Multi-agent Decision Making", "Chain-of-Thought" ], "keywords_by_category": { "Benchmark": "GAMA-Bench, Game Theory, Multi-agent Decision Making, Chain-of-Thought" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2403.11807" }, { "label": "Code", "url": "https://github.com/CUHK-ARISE/GAMABench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 356, "date": "2024-05", "title": "How Far Are We on the Decision-Making of LLMs? Evaluating LLMs' Gaming Ability in Multi-Agent Environments", "venue": "ICLR 2025", "keywords": "GAMA-Bench, Game Theory, Multi-agent Decision Making, Chain-of-Thought", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2403.11807" }, { "label": "Code", "url": "https://github.com/CUHK-ARISE/GAMABench" } ] } ], "sources": [ "https://arxiv.org/abs/2403.11807", "https://github.com/CUHK-ARISE/GAMABench" ], "search_text": "how far are we on the decision-making of llms? evaluating llms' gaming ability in multi-agent environments how far are we on the decision-making of llms? evaluating llms' gaming ability in multi-agent environments 2024-05 iclr 2025 benchmark gama-bench game theory multi-agent decision making chain-of-thought player" }, { "id": "arxiv-2405.20859", "name": "clembench-2024", "title": "clembench-2024: A Challenging, Dynamic, Complementary, Multilingual Benchmark and Underlying Flexible Framework for LLMs as Multi-Action Agents", "date": "2024-05", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Dialogue Games", "Multilingual", "Multi-Action Agents", "Dynamic Evaluation" ], "keywords_by_category": { "Benchmark": "Dialogue Games, Multilingual, Multi-Action Agents, Dynamic Evaluation" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2405.20859" }, { "label": "Code", "url": "https://github.com/clembench/clembench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 355, "date": "2024-05", "title": "clembench-2024: A Challenging, Dynamic, Complementary, Multilingual Benchmark and Underlying Flexible Framework for LLMs as Multi-Action Agents", "venue": "arXiv", "keywords": "Dialogue Games, Multilingual, Multi-Action Agents, Dynamic Evaluation", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2405.20859" }, { "label": "Code", "url": "https://github.com/clembench/clembench" } ] } ], "sources": [ "https://arxiv.org/abs/2405.20859", "https://github.com/clembench/clembench" ], "search_text": "clembench-2024: a challenging, dynamic, complementary, multilingual benchmark and underlying flexible framework for llms as multi-action agents clembench-2024 2024-05 arxiv benchmark dialogue games multilingual multi-action agents dynamic evaluation player" }, { "id": "arxiv-2404.02039", "name": "A Survey on Large Language Model-Based Game Agents", "title": "A Survey on Large Language Model-Based Game Agents", "date": "2024-04", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Survey" ], "primary_category": null, "keywords": [], "keywords_by_category": { "Survey": "" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2404.02039" }, { "label": "GitHub", "url": "https://github.com/git-disl/awesome-LLM-game-agent-papers" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "survey", "source_rows": [ { "section": "Survey", "line": 378, "date": "2024-04", "title": "A Survey on Large Language Model-Based Game Agents", "venue": "arXiv", "keywords": "", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2404.02039" }, { "label": "GitHub", "url": "https://github.com/git-disl/awesome-LLM-game-agent-papers" } ] } ], "sources": [ "https://arxiv.org/abs/2404.02039", "https://github.com/git-disl/awesome-LLM-game-agent-papers" ], "search_text": "a survey on large language model-based game agents a survey on large language model-based game agents 2024-04 arxiv survey survey" }, { "id": "arxiv-2404.10179", "name": "Scaling Instructable Agents Across Many Simulated Worlds", "title": "Scaling Instructable Agents Across Many Simulated Worlds", "date": "2024-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Instructable Agent", "Cross-World Transfer", "Keyboard-Mouse Actions" ], "keywords_by_category": { "Model": "Instructable Agent, Cross-World Transfer, Keyboard-Mouse Actions" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2404.10179" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 157, "date": "2024-03", "title": "Scaling Instructable Agents Across Many Simulated Worlds", "venue": "arXiv", "keywords": "Instructable Agent, Cross-World Transfer, Keyboard-Mouse Actions", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2404.10179" } ] } ], "sources": [ "https://arxiv.org/abs/2404.10179" ], "search_text": "scaling instructable agents across many simulated worlds scaling instructable agents across many simulated worlds 2024-03 arxiv model instructable agent cross-world transfer keyboard-mouse actions player" }, { "id": "work-3e1546e13c87", "name": "Cradle", "title": "Cradle: Empowering Foundation Agents towards General Computer Control", "date": "2024-03", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICML 2025", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "Reflection", "RAG", "Workflow" ], "keywords_by_category": { "Harness": "Reflection, RAG, Workflow" }, "links": [ { "label": "Paper", "url": "https://proceedings.mlr.press/v267/tan25h.html" }, { "label": "Code", "url": "https://github.com/BAAI-Agents/Cradle" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 214, "date": "2024-03", "title": "Cradle: Empowering Foundation Agents towards General Computer Control", "venue": "ICML 2025", "keywords": "Reflection, RAG, Workflow", "links": [ { "label": "Paper", "url": "https://proceedings.mlr.press/v267/tan25h.html" }, { "label": "Code", "url": "https://github.com/BAAI-Agents/Cradle" } ] } ], "sources": [ "https://proceedings.mlr.press/v267/tan25h.html", "https://github.com/BAAI-Agents/Cradle" ], "search_text": "cradle: empowering foundation agents towards general computer control cradle 2024-03 icml 2025 harness reflection rag workflow player" }, { "id": "work-6eb9e0670615", "name": "Large Language Models and Games", "title": "Large Language Models and Games: A Survey and Roadmap", "date": "2024-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "IEEE Trans. Games", "categories": [ "Survey" ], "primary_category": null, "keywords": [], "keywords_by_category": { "Survey": "" }, "links": [ { "label": "Paper", "url": "https://ieeexplore.ieee.org/document/10680313" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "survey", "source_rows": [ { "section": "Survey", "line": 379, "date": "2024-02", "title": "Large Language Models and Games: A Survey and Roadmap", "venue": "IEEE Trans. Games", "keywords": "", "links": [ { "label": "Paper", "url": "https://ieeexplore.ieee.org/document/10680313" } ] } ], "sources": [ "https://ieeexplore.ieee.org/document/10680313" ], "search_text": "large language models and games: a survey and roadmap large language models and games 2024-02 ieee trans. games survey survey" }, { "id": "work-1ac442b36841", "name": "GTBench", "title": "GTBench: Uncovering the Strategic Reasoning Limitations of LLMs via Game-Theoretic Evaluations", "date": "2024-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NeurIPS 2024", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "10 Game-Theoretic Tasks", "LLM-vs-LLM", "Complete/Incomplete Info" ], "keywords_by_category": { "Benchmark": "10 Game-Theoretic Tasks, LLM-vs-LLM, Complete/Incomplete Info" }, "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/3191170938b6102e5c203b036b7c16dd-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/jinhaoduan/GTBench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 357, "date": "2024-02", "title": "GTBench: Uncovering the Strategic Reasoning Limitations of LLMs via Game-Theoretic Evaluations", "venue": "NeurIPS 2024", "keywords": "10 Game-Theoretic Tasks, LLM-vs-LLM, Complete/Incomplete Info", "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/3191170938b6102e5c203b036b7c16dd-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/jinhaoduan/GTBench" } ] } ], "sources": [ "https://proceedings.neurips.cc/paper_files/paper/2024/hash/3191170938b6102e5c203b036b7c16dd-Abstract-Conference.html", "https://github.com/jinhaoduan/GTBench" ], "search_text": "gtbench: uncovering the strategic reasoning limitations of llms via game-theoretic evaluations gtbench 2024-02 neurips 2024 benchmark 10 game-theoretic tasks llm-vs-llm complete/incomplete info player" }, { "id": "arxiv-2402.15391", "name": "Genie", "title": "Genie: Generative Interactive Environments", "date": "2024-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICML 2024", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Latent Action Modeling", "Unsupervised Video Learning", "Foundation World Model" ], "keywords_by_category": { "Model": "Latent Action Modeling, Unsupervised Video Learning, Foundation World Model" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2402.15391" }, { "label": "Project", "url": "https://sites.google.com/view/genie-2024/" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "supporting environment", "source_rows": [ { "section": "Model", "line": 158, "date": "2024-02", "title": "Genie: Generative Interactive Environments", "venue": "ICML 2024", "keywords": "Latent Action Modeling, Unsupervised Video Learning, Foundation World Model", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2402.15391" }, { "label": "Project", "url": "https://sites.google.com/view/genie-2024/" } ] } ], "sources": [ "https://arxiv.org/abs/2402.15391", "https://sites.google.com/view/genie-2024/" ], "search_text": "genie: generative interactive environments genie 2024-02 icml 2024 model latent action modeling unsupervised video learning foundation world model supporting environment" }, { "id": "arxiv-2402.02330", "name": "Enhance Reasoning for Large Language Models in the Game Werewolf", "title": "Enhance Reasoning for Large Language Models in the Game Werewolf", "date": "2024-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Dataset" ], "primary_category": null, "keywords": [ "18.8K Sessions", "Social Deduction" ], "keywords_by_category": { "Dataset": "18.8K Sessions, Social Deduction" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2402.02330" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Dataset", "line": 255, "date": "2024-02", "title": "Enhance Reasoning for Large Language Models in the Game Werewolf", "venue": "arXiv", "keywords": "18.8K Sessions, Social Deduction", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2402.02330" } ] } ], "sources": [ "https://arxiv.org/abs/2402.02330" ], "search_text": "enhance reasoning for large language models in the game werewolf enhance reasoning for large language models in the game werewolf 2024-02 arxiv dataset 18.8k sessions social deduction player" }, { "id": "work-55d45b31df5b", "name": "StarCraftImage", "title": "StarCraftImage: A Dataset For Prototyping Spatial Reasoning Methods For Multi-Agent Environments", "date": "2024-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "CVPR 2023", "categories": [ "Dataset" ], "primary_category": null, "keywords": [ "3.6M Images", "60K Replays", "Spatial Reasoning" ], "keywords_by_category": { "Dataset": "3.6M Images, 60K Replays, Spatial Reasoning" }, "links": [ { "label": "Paper", "url": "https://cvpr.thecvf.com/virtual/2023/poster/21261" }, { "label": "Code", "url": "https://github.com/inouye-lab/starcraftimage" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Dataset", "line": 256, "date": "2024-01", "title": "StarCraftImage: A Dataset For Prototyping Spatial Reasoning Methods For Multi-Agent Environments", "venue": "CVPR 2023", "keywords": "3.6M Images, 60K Replays, Spatial Reasoning", "links": [ { "label": "Paper", "url": "https://cvpr.thecvf.com/virtual/2023/poster/21261" }, { "label": "Code", "url": "https://github.com/inouye-lab/starcraftimage" } ] } ], "sources": [ "https://cvpr.thecvf.com/virtual/2023/poster/21261", "https://github.com/inouye-lab/starcraftimage" ], "search_text": "starcraftimage: a dataset for prototyping spatial reasoning methods for multi-agent environments starcraftimage 2024-01 cvpr 2023 dataset 3.6m images 60k replays spatial reasoning player" }, { "id": "work-24fef84212e9", "name": "CivRealm", "title": "CivRealm: A Learning and Reasoning Odyssey in Civilization for Decision-Making Agents", "date": "2024-01", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2024", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Civilization", "Learning & Reasoning", "Decision-Making", "Generalization", "Long-Horizon" ], "keywords_by_category": { "Benchmark": "Civilization, Learning & Reasoning, Decision-Making, Generalization, Long-Horizon" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=UBVNwD3hPN" }, { "label": "Code", "url": "https://github.com/bigai-ai/civrealm" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 358, "date": "2024-01", "title": "CivRealm: A Learning and Reasoning Odyssey in Civilization for Decision-Making Agents", "venue": "ICLR 2024", "keywords": "Civilization, Learning & Reasoning, Decision-Making, Generalization, Long-Horizon", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=UBVNwD3hPN" }, { "label": "Code", "url": "https://github.com/bigai-ai/civrealm" } ] } ], "sources": [ "https://openreview.net/forum?id=UBVNwD3hPN", "https://github.com/bigai-ai/civrealm" ], "search_text": "civrealm: a learning and reasoning odyssey in civilization for decision-making agents civrealm 2024-01 iclr 2024 benchmark civilization learning & reasoning decision-making generalization long-horizon player" }, { "id": "work-9830ccd79478", "name": "Large Language Models Play StarCraft II", "title": "Large Language Models Play StarCraft II: Benchmarks and A Chain of Summarization Approach", "date": "2023-12", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NeurIPS 2024", "categories": [ "Harness", "Benchmark" ], "primary_category": null, "keywords": [ "Textual environment", "Chain-of-Summarization", "StarCraft II", "Chain of Summarization", "Text-based Environment", "Real-time Strategy" ], "keywords_by_category": { "Harness": "Textual environment, Chain-of-Summarization", "Benchmark": "StarCraft II, Chain of Summarization, Text-based Environment, Real-time Strategy" }, "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/f0ebc318e2df08360b2df559e81602e5-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/histmeisah/large-language-models-play-starcraftii" }, { "label": "Code", "url": "https://github.com/histmeisah/Large-Language-Models-play-StarCraftII" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 215, "date": "2023-12", "title": "Large Language Models Play StarCraft II: Benchmarks and A Chain of Summarization Approach", "venue": "NeurIPS 2024", "keywords": "Textual environment, Chain-of-Summarization", "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/f0ebc318e2df08360b2df559e81602e5-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/histmeisah/large-language-models-play-starcraftii" } ] }, { "section": "Benchmark", "line": 359, "date": "2023-12", "title": "Large Language Models Play StarCraft II: Benchmarks and A Chain of Summarization Approach", "venue": "NeurIPS 2024", "keywords": "StarCraft II, Chain of Summarization, Text-based Environment, Real-time Strategy", "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/f0ebc318e2df08360b2df559e81602e5-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/histmeisah/Large-Language-Models-play-StarCraftII" } ] } ], "sources": [ "https://proceedings.neurips.cc/paper_files/paper/2024/hash/f0ebc318e2df08360b2df559e81602e5-Abstract-Conference.html", "https://github.com/histmeisah/large-language-models-play-starcraftii", "https://github.com/histmeisah/Large-Language-Models-play-StarCraftII" ], "search_text": "large language models play starcraft ii: benchmarks and a chain of summarization approach large language models play starcraft ii 2023-12 neurips 2024 harness benchmark textual environment chain-of-summarization starcraft ii chain of summarization text-based environment real-time strategy player" }, { "id": "arxiv-2312.02519", "name": "Creative Agents", "title": "Creative Agents: Empowering Agents with Imagination for Creative Tasks", "date": "2023-12", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Model" ], "primary_category": null, "keywords": [ "VLM", "Goal Image Generation", "Minecraft" ], "keywords_by_category": { "Model": "VLM, Goal Image Generation, Minecraft" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2312.02519" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 159, "date": "2023-12", "title": "Creative Agents: Empowering Agents with Imagination for Creative Tasks", "venue": "arXiv", "keywords": "VLM, Goal Image Generation, Minecraft", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2312.02519" } ] } ], "sources": [ "https://arxiv.org/abs/2312.02519" ], "search_text": "creative agents: empowering agents with imagination for creative tasks creative agents 2023-12 arxiv model vlm goal image generation minecraft player" }, { "id": "arxiv-2311.05997", "name": "JARVIS-1", "title": "JARVIS-1: Open-World Multi-task Agents with Memory-Augmented Multimodal Language Models", "date": "2023-11", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "Memory-Augmented LLM", "Multimodal Planning", "Minecraft" ], "keywords_by_category": { "Harness": "Memory-Augmented LLM, Multimodal Planning, Minecraft" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2311.05997" }, { "label": "Code", "url": "https://github.com/CraftJarvis/JARVIS-1" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 216, "date": "2023-11", "title": "JARVIS-1: Open-World Multi-task Agents with Memory-Augmented Multimodal Language Models", "venue": "arXiv", "keywords": "Memory-Augmented LLM, Multimodal Planning, Minecraft", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2311.05997" }, { "label": "Code", "url": "https://github.com/CraftJarvis/JARVIS-1" } ] } ], "sources": [ "https://arxiv.org/abs/2311.05997", "https://github.com/CraftJarvis/JARVIS-1" ], "search_text": "jarvis-1: open-world multi-task agents with memory-augmented multimodal language models jarvis-1 2023-11 arxiv harness memory-augmented llm multimodal planning minecraft player" }, { "id": "work-08e543fb4030", "name": "SmartPlay ", "title": "SmartPlay : A Benchmark for LLMs as Intelligent Agents", "date": "2023-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2024", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "6 Games", "20 Settings", "Intelligent Agents", "Multi-Tasking" ], "keywords_by_category": { "Benchmark": "6 Games, 20 Settings, Intelligent Agents, Multi-Tasking" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=S2oTVrlcp3" }, { "label": "Code", "url": "https://github.com/microsoft/SmartPlay" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 362, "date": "2023-10", "title": "SmartPlay : A Benchmark for LLMs as Intelligent Agents", "venue": "ICLR 2024", "keywords": "6 Games, 20 Settings, Intelligent Agents, Multi-Tasking", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=S2oTVrlcp3" }, { "label": "Code", "url": "https://github.com/microsoft/SmartPlay" } ] } ], "sources": [ "https://openreview.net/forum?id=S2oTVrlcp3", "https://github.com/microsoft/SmartPlay" ], "search_text": "smartplay : a benchmark for llms as intelligent agents smartplay 2023-10 iclr 2024 benchmark 6 games 20 settings intelligent agents multi-tasking player" }, { "id": "work-3e0ec0170e71", "name": "LLM-Coordination", "title": "LLM-Coordination: Evaluating and Analyzing Multi-agent Coordination Abilities in Large Language Models", "date": "2023-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "Findings of NAACL 2025", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Coordination Games", "Multi-agent", "Theory of Mind", "Pure Coordination" ], "keywords_by_category": { "Benchmark": "Coordination Games, Multi-agent, Theory of Mind, Pure Coordination" }, "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.findings-naacl.448/" }, { "label": "Code", "url": "https://github.com/eric-ai-lab/llm_coordination" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 361, "date": "2023-10", "title": "LLM-Coordination: Evaluating and Analyzing Multi-agent Coordination Abilities in Large Language Models", "venue": "Findings of NAACL 2025", "keywords": "Coordination Games, Multi-agent, Theory of Mind, Pure Coordination", "links": [ { "label": "Paper", "url": "https://aclanthology.org/2025.findings-naacl.448/" }, { "label": "Code", "url": "https://github.com/eric-ai-lab/llm_coordination" } ] } ], "sources": [ "https://aclanthology.org/2025.findings-naacl.448/", "https://github.com/eric-ai-lab/llm_coordination" ], "search_text": "llm-coordination: evaluating and analyzing multi-agent coordination abilities in large language models llm-coordination 2023-10 findings of naacl 2025 benchmark coordination games multi-agent theory of mind pure coordination player" }, { "id": "work-f68d35caddb8", "name": "LLaMA-Rider", "title": "LLaMA-Rider: Spurring Large Language Models to Explore the Open World", "date": "2023-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NAACL Findings 2024", "categories": [ "Model" ], "primary_category": null, "keywords": [ "SFT", "Exploration", "Minecraft" ], "keywords_by_category": { "Model": "SFT, Exploration, Minecraft" }, "links": [ { "label": "Paper", "url": "https://aclanthology.org/2024.findings-naacl.292/" }, { "label": "Code", "url": "https://github.com/PKU-RL/LLaMA-Rider" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 161, "date": "2023-10", "title": "LLaMA-Rider: Spurring Large Language Models to Explore the Open World", "venue": "NAACL Findings 2024", "keywords": "SFT, Exploration, Minecraft", "links": [ { "label": "Paper", "url": "https://aclanthology.org/2024.findings-naacl.292/" }, { "label": "Code", "url": "https://github.com/PKU-RL/LLaMA-Rider" } ] } ], "sources": [ "https://aclanthology.org/2024.findings-naacl.292/", "https://github.com/PKU-RL/LLaMA-Rider" ], "search_text": "llama-rider: spurring large language models to explore the open world llama-rider 2023-10 naacl findings 2024 model sft exploration minecraft player" }, { "id": "work-cb163b936fc7", "name": "Language Agents with Reinforcement Learning for Strategic Play in the Werewolf Game", "title": "Language Agents with Reinforcement Learning for Strategic Play in the Werewolf Game", "date": "2023-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICML 2024", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Werewolf", "RL Policy", "Social Deduction", "LLM + RL", "Strategic Play" ], "keywords_by_category": { "Model": "Werewolf, RL Policy, Social Deduction, LLM + RL, Strategic Play" }, "links": [ { "label": "Paper", "url": "https://icml.cc/virtual/2024/poster/32805" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 160, "date": "2023-10", "title": "Language Agents with Reinforcement Learning for Strategic Play in the Werewolf Game", "venue": "ICML 2024", "keywords": "Werewolf, RL Policy, Social Deduction, LLM + RL, Strategic Play", "links": [ { "label": "Paper", "url": "https://icml.cc/virtual/2024/poster/32805" } ] } ], "sources": [ "https://icml.cc/virtual/2024/poster/32805" ], "search_text": "language agents with reinforcement learning for strategic play in the werewolf game language agents with reinforcement learning for strategic play in the werewolf game 2023-10 icml 2024 model werewolf rl policy social deduction llm + rl strategic play player" }, { "id": "work-e2f0fbee0379", "name": "GROOT", "title": "GROOT: Learning to Follow Instructions by Watching Gameplay Videos", "date": "2023-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2024", "categories": [ "Dataset", "Model" ], "primary_category": null, "keywords": [ "SkillForge Benchmark", "30 Tasks", "6 Categories", "Video-Conditioned Imitation", "Causal Transformer" ], "keywords_by_category": { "Dataset": "SkillForge Benchmark, 30 Tasks, 6 Categories", "Model": "Video-Conditioned Imitation, Causal Transformer" }, "links": [ { "label": "Paper", "url": "https://proceedings.iclr.cc/paper_files/paper/2024/hash/16986b69068fbe6acf64eb6566519c74-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/CraftJarvis/GROOT" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 162, "date": "2023-10", "title": "GROOT: Learning to Follow Instructions by Watching Gameplay Videos", "venue": "ICLR 2024", "keywords": "Video-Conditioned Imitation, Causal Transformer", "links": [ { "label": "Paper", "url": "https://proceedings.iclr.cc/paper_files/paper/2024/hash/16986b69068fbe6acf64eb6566519c74-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/CraftJarvis/GROOT" } ] }, { "section": "Dataset", "line": 257, "date": "2023-10", "title": "GROOT: Learning to Follow Instructions by Watching Gameplay Videos", "venue": "ICLR 2024", "keywords": "SkillForge Benchmark, 30 Tasks, 6 Categories", "links": [ { "label": "Paper", "url": "https://proceedings.iclr.cc/paper_files/paper/2024/hash/16986b69068fbe6acf64eb6566519c74-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/CraftJarvis/GROOT" } ] } ], "sources": [ "https://proceedings.iclr.cc/paper_files/paper/2024/hash/16986b69068fbe6acf64eb6566519c74-Abstract-Conference.html", "https://github.com/CraftJarvis/GROOT" ], "search_text": "groot: learning to follow instructions by watching gameplay videos groot 2023-10 iclr 2024 dataset model skillforge benchmark 30 tasks 6 categories video-conditioned imitation causal transformer player" }, { "id": "arxiv-2310.05036", "name": "AvalonBench", "title": "AvalonBench: Evaluating LLMs Playing the Game of Avalon", "date": "2023-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "arXiv", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Avalon", "Social Deduction", "LLM Evaluation", "Multi-Agent", "Deception" ], "keywords_by_category": { "Benchmark": "Avalon, Social Deduction, LLM Evaluation, Multi-Agent, Deception" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2310.05036" }, { "label": "Code", "url": "https://github.com/jonathanmli/Avalon-LLM" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 360, "date": "2023-10", "title": "AvalonBench: Evaluating LLMs Playing the Game of Avalon", "venue": "arXiv", "keywords": "Avalon, Social Deduction, LLM Evaluation, Multi-Agent, Deception", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2310.05036" }, { "label": "Code", "url": "https://github.com/jonathanmli/Avalon-LLM" } ] } ], "sources": [ "https://arxiv.org/abs/2310.05036", "https://github.com/jonathanmli/Avalon-LLM" ], "search_text": "avalonbench: evaluating llms playing the game of avalon avalonbench 2023-10 arxiv benchmark avalon social deduction llm evaluation multi-agent deception player" }, { "id": "work-ee214e269321", "name": "AgentVerse", "title": "AgentVerse: Facilitating Multi-Agent Collaboration and Exploring Emergent Behaviors", "date": "2023-08", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2024", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "Multi-Agent" ], "keywords_by_category": { "Harness": "Multi-Agent" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=EHg5GDnyq1" }, { "label": "Code", "url": "https://github.com/OpenBMB/AgentVerse" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 217, "date": "2023-08", "title": "AgentVerse: Facilitating Multi-Agent Collaboration and Exploring Emergent Behaviors", "venue": "ICLR 2024", "keywords": "Multi-Agent", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=EHg5GDnyq1" }, { "label": "Code", "url": "https://github.com/OpenBMB/AgentVerse" } ] } ], "sources": [ "https://openreview.net/forum?id=EHg5GDnyq1", "https://github.com/OpenBMB/AgentVerse" ], "search_text": "agentverse: facilitating multi-agent collaboration and exploring emergent behaviors agentverse 2023-08 iclr 2024 harness multi-agent player" }, { "id": "arxiv-2306.00937", "name": "STEVE-1", "title": "STEVE-1: A Generative Model for Text-to-Behavior in Minecraft", "date": "2023-06", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NeurIPS 2023", "categories": [ "Model" ], "primary_category": null, "keywords": [ "Behavioral Cloning", "Hindsight Relabeling" ], "keywords_by_category": { "Model": "Behavioral Cloning, Hindsight Relabeling" }, "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2306.00937" }, { "label": "Code", "url": "https://github.com/Shalev-Lifshitz/STEVE-1" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 163, "date": "2023-06", "title": "STEVE-1: A Generative Model for Text-to-Behavior in Minecraft", "venue": "NeurIPS 2023", "keywords": "Behavioral Cloning, Hindsight Relabeling", "links": [ { "label": "Paper", "url": "https://arxiv.org/abs/2306.00937" }, { "label": "Code", "url": "https://github.com/Shalev-Lifshitz/STEVE-1" } ] } ], "sources": [ "https://arxiv.org/abs/2306.00937", "https://github.com/Shalev-Lifshitz/STEVE-1" ], "search_text": "steve-1: a generative model for text-to-behavior in minecraft steve-1 2023-06 neurips 2023 model behavioral cloning hindsight relabeling player" }, { "id": "work-9baef100d285", "name": "Voyager", "title": "Voyager: An Open-Ended Embodied Agent with Large Language Models", "date": "2023-05", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NeurIPS 2023 Workshop IMOL", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "Textual environment", "LLM coding", "RAG(skill storage)", "Reflection" ], "keywords_by_category": { "Harness": "Textual environment,LLM coding, RAG(skill storage), Reflection" }, "links": [ { "label": "Publication", "url": "https://nips.cc/virtual/2023/77597" }, { "label": "Code", "url": "https://github.com/MineDojo/Voyager" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 218, "date": "2023-05", "title": "Voyager: An Open-Ended Embodied Agent with Large Language Models", "venue": "NeurIPS 2023 Workshop IMOL", "keywords": "Textual environment,LLM coding, RAG(skill storage), Reflection", "links": [ { "label": "Publication", "url": "https://nips.cc/virtual/2023/77597" }, { "label": "Code", "url": "https://github.com/MineDojo/Voyager" } ] } ], "sources": [ "https://nips.cc/virtual/2023/77597", "https://github.com/MineDojo/Voyager" ], "search_text": "voyager: an open-ended embodied agent with large language models voyager 2023-05 neurips 2023 workshop imol harness textual environment llm coding rag(skill storage) reflection player" }, { "id": "work-2c0a5d3cab82", "name": "clembench", "title": "clembench: Using Game Play to Evaluate Chat-Optimized Language Models as Conversational Agents", "date": "2023-05", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "EMNLP 2023", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Dialogue Games", "Conversational Agents", "Interactive Evaluation" ], "keywords_by_category": { "Benchmark": "Dialogue Games, Conversational Agents, Interactive Evaluation" }, "links": [ { "label": "Paper", "url": "https://aclanthology.org/2023.emnlp-main.689/" }, { "label": "Code", "url": "https://github.com/clembench/clembench" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 363, "date": "2023-05", "title": "clembench: Using Game Play to Evaluate Chat-Optimized Language Models as Conversational Agents", "venue": "EMNLP 2023", "keywords": "Dialogue Games, Conversational Agents, Interactive Evaluation", "links": [ { "label": "Paper", "url": "https://aclanthology.org/2023.emnlp-main.689/" }, { "label": "Code", "url": "https://github.com/clembench/clembench" } ] } ], "sources": [ "https://aclanthology.org/2023.emnlp-main.689/", "https://github.com/clembench/clembench" ], "search_text": "clembench: using game play to evaluate chat-optimized language models as conversational agents clembench 2023-05 emnlp 2023 benchmark dialogue games conversational agents interactive evaluation player" }, { "id": "work-bdfbff559c9a", "name": "Describe, Explain, Plan and Select", "title": "Describe, Explain, Plan and Select: Interactive Planning with Large Language Models Enables Open-World Multi-Task Agents", "date": "2023-02", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NeurIPS 2023", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "Zero-Shot LLM Planning", "Interactive Refinement", "Minecraft" ], "keywords_by_category": { "Harness": "Zero-Shot LLM Planning, Interactive Refinement, Minecraft" }, "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2023/hash/6b8dfb8c0c12e6fafc6c256cb08a5ca7-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/CraftJarvis/MC-Planner" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 219, "date": "2023-02", "title": "Describe, Explain, Plan and Select: Interactive Planning with Large Language Models Enables Open-World Multi-Task Agents", "venue": "NeurIPS 2023", "keywords": "Zero-Shot LLM Planning, Interactive Refinement, Minecraft", "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2023/hash/6b8dfb8c0c12e6fafc6c256cb08a5ca7-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/CraftJarvis/MC-Planner" } ] } ], "sources": [ "https://proceedings.neurips.cc/paper_files/paper/2023/hash/6b8dfb8c0c12e6fafc6c256cb08a5ca7-Abstract-Conference.html", "https://github.com/CraftJarvis/MC-Planner" ], "search_text": "describe, explain, plan and select: interactive planning with large language models enables open-world multi-task agents describe, explain, plan and select 2023-02 neurips 2023 harness zero-shot llm planning interactive refinement minecraft player" }, { "id": "work-593791331cd6", "name": "Human-level play in the game of Diplomacy by combining language models with strategic reasoning", "title": "Human-level play in the game of Diplomacy by combining language models with strategic reasoning", "date": "2022-11", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "Science 2022", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Diplomacy", "Negotiation", "Strategic Reasoning", "Language+Strategy" ], "keywords_by_category": { "Benchmark": "Diplomacy, Negotiation, Strategic Reasoning, Language+Strategy" }, "links": [ { "label": "Paper", "url": "https://doi.org/10.1126/science.ade9097" }, { "label": "Code", "url": "https://github.com/facebookresearch/diplomacy_cicero" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 364, "date": "2022-11", "title": "Human-level play in the game of Diplomacy by combining language models with strategic reasoning", "venue": "Science 2022", "keywords": "Diplomacy, Negotiation, Strategic Reasoning, Language+Strategy", "links": [ { "label": "Paper", "url": "https://doi.org/10.1126/science.ade9097" }, { "label": "Code", "url": "https://github.com/facebookresearch/diplomacy_cicero" } ] } ], "sources": [ "https://doi.org/10.1126/science.ade9097", "https://github.com/facebookresearch/diplomacy_cicero" ], "search_text": "human-level play in the game of diplomacy by combining language models with strategic reasoning human-level play in the game of diplomacy by combining language models with strategic reasoning 2022-11 science 2022 benchmark diplomacy negotiation strategic reasoning language+strategy player" }, { "id": "work-00d3a7d0f9d5", "name": "ReAct", "title": "ReAct: Synergizing Reasoning and Acting in Language Models", "date": "2022-10", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2023", "categories": [ "Harness" ], "primary_category": null, "keywords": [ "ReAct", "Chain-of-Thought (CoT)" ], "keywords_by_category": { "Harness": "ReAct, Chain-of-Thought (CoT)" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=WE_vluYUL-X" }, { "label": "Code", "url": "https://github.com/ysymyth/ReAct" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Harness", "line": 220, "date": "2022-10", "title": "ReAct: Synergizing Reasoning and Acting in Language Models", "venue": "ICLR 2023", "keywords": "ReAct, Chain-of-Thought (CoT)", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=WE_vluYUL-X" }, { "label": "Code", "url": "https://github.com/ysymyth/ReAct" } ] } ], "sources": [ "https://openreview.net/forum?id=WE_vluYUL-X", "https://github.com/ysymyth/ReAct" ], "search_text": "react: synergizing reasoning and acting in language models react 2022-10 iclr 2023 harness react chain-of-thought (cot) player" }, { "id": "work-0fc98c650cb8", "name": "Video PreTraining (VPT)", "title": "Video PreTraining (VPT): Learning to Act by Watching Unlabeled Online Videos", "date": "2022-06", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NeurIPS 2022", "categories": [ "Dataset", "Model" ], "primary_category": null, "keywords": [ "~70K hrs Unlabeled Video", "Pseudo-Labeled Actions", "Semi-Supervised IL + RL", "Inverse Dynamics" ], "keywords_by_category": { "Dataset": "~70K hrs Unlabeled Video, Pseudo-Labeled Actions", "Model": "Semi-Supervised IL + RL, Inverse Dynamics" }, "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2022/hash/9c7008aff45b5d8f0973b23e1a22ada0-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/openai/Video-Pre-Training" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 164, "date": "2022-06", "title": "Video PreTraining (VPT): Learning to Act by Watching Unlabeled Online Videos", "venue": "NeurIPS 2022", "keywords": "Semi-Supervised IL + RL, Inverse Dynamics", "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2022/hash/9c7008aff45b5d8f0973b23e1a22ada0-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/openai/Video-Pre-Training" } ] }, { "section": "Dataset", "line": 258, "date": "2022-06", "title": "Video PreTraining (VPT): Learning to Act by Watching Unlabeled Online Videos", "venue": "NeurIPS 2022", "keywords": "~70K hrs Unlabeled Video, Pseudo-Labeled Actions", "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2022/hash/9c7008aff45b5d8f0973b23e1a22ada0-Abstract-Conference.html" }, { "label": "Code", "url": "https://github.com/openai/Video-Pre-Training" } ] } ], "sources": [ "https://proceedings.neurips.cc/paper_files/paper/2022/hash/9c7008aff45b5d8f0973b23e1a22ada0-Abstract-Conference.html", "https://github.com/openai/Video-Pre-Training" ], "search_text": "video pretraining (vpt): learning to act by watching unlabeled online videos video pretraining (vpt) 2022-06 neurips 2022 dataset model ~70k hrs unlabeled video pseudo-labeled actions semi-supervised il + rl inverse dynamics player" }, { "id": "work-48cce40f663c", "name": "MineDojo", "title": "MineDojo: Building Open-Ended Embodied Agents with Internet-Scale Knowledge", "date": "2022-06", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NeurIPS 2022", "categories": [ "Dataset", "Model" ], "primary_category": null, "keywords": [ "Internet-Scale", "Multi-Task", "Minecraft", "Pre-training", "Multi-Task RL" ], "keywords_by_category": { "Dataset": "Internet-Scale, Multi-Task, Minecraft", "Model": "Pre-training, Multi-Task RL" }, "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2022/hash/74a67268c5cc5910f64938cac4526a90-Abstract-Datasets_and_Benchmarks.html" }, { "label": "Code", "url": "https://github.com/MineDojo/MineDojo" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 165, "date": "2022-06", "title": "MineDojo: Building Open-Ended Embodied Agents with Internet-Scale Knowledge", "venue": "NeurIPS 2022", "keywords": "Pre-training, Multi-Task RL", "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2022/hash/74a67268c5cc5910f64938cac4526a90-Abstract-Datasets_and_Benchmarks.html" }, { "label": "Code", "url": "https://github.com/MineDojo/MineDojo" } ] }, { "section": "Dataset", "line": 259, "date": "2022-06", "title": "MineDojo: Building Open-Ended Embodied Agents with Internet-Scale Knowledge", "venue": "NeurIPS 2022", "keywords": "Internet-Scale, Multi-Task, Minecraft", "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper_files/paper/2022/hash/74a67268c5cc5910f64938cac4526a90-Abstract-Datasets_and_Benchmarks.html" }, { "label": "Code", "url": "https://github.com/MineDojo/MineDojo" } ] } ], "sources": [ "https://proceedings.neurips.cc/paper_files/paper/2022/hash/74a67268c5cc5910f64938cac4526a90-Abstract-Datasets_and_Benchmarks.html", "https://github.com/MineDojo/MineDojo" ], "search_text": "minedojo: building open-ended embodied agents with internet-scale knowledge minedojo 2022-06 neurips 2022 dataset model internet-scale multi-task minecraft pre-training multi-task rl player" }, { "id": "work-69dfcf04ba07", "name": "Benchmarking the Spectrum of Agent Capabilities", "title": "Benchmarking the Spectrum of Agent Capabilities", "date": "2021-09", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "ICLR 2022", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Crafter", "Open-world Survival", "General Capabilities", "Single-Environment Benchmark" ], "keywords_by_category": { "Benchmark": "Crafter, Open-world Survival, General Capabilities, Single-Environment Benchmark" }, "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=1W0z96MFEoH" }, { "label": "Code", "url": "https://github.com/danijar/crafter" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 365, "date": "2021-09", "title": "Benchmarking the Spectrum of Agent Capabilities", "venue": "ICLR 2022", "keywords": "Crafter, Open-world Survival, General Capabilities, Single-Environment Benchmark", "links": [ { "label": "Paper", "url": "https://openreview.net/forum?id=1W0z96MFEoH" }, { "label": "Code", "url": "https://github.com/danijar/crafter" } ] } ], "sources": [ "https://openreview.net/forum?id=1W0z96MFEoH", "https://github.com/danijar/crafter" ], "search_text": "benchmarking the spectrum of agent capabilities benchmarking the spectrum of agent capabilities 2021-09 iclr 2022 benchmark crafter open-world survival general capabilities single-environment benchmark player" }, { "id": "work-73243e595ae7", "name": "Counter-Strike Deathmatch with Large-Scale Behavioural Cloning", "title": "Counter-Strike Deathmatch with Large-Scale Behavioural Cloning", "date": "2021-04", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "CoG 2022", "categories": [ "Dataset", "Model" ], "primary_category": null, "keywords": [ "4M Frames", "Online Human Gameplay", "FPS", "Large-Scale Behavioral Cloning", "Pixel-to-Action" ], "keywords_by_category": { "Dataset": "4M Frames, Online Human Gameplay, FPS", "Model": "Large-Scale Behavioral Cloning, FPS, Pixel-to-Action" }, "links": [ { "label": "Paper", "url": "https://ieeexplore.ieee.org/document/9893617" }, { "label": "Code", "url": "https://github.com/TeaPearce/Counter-Strike_Behavioural_Cloning" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Model", "line": 166, "date": "2021-04", "title": "Counter-Strike Deathmatch with Large-Scale Behavioural Cloning", "venue": "CoG 2022", "keywords": "Large-Scale Behavioral Cloning, FPS, Pixel-to-Action", "links": [ { "label": "Paper", "url": "https://ieeexplore.ieee.org/document/9893617" }, { "label": "Code", "url": "https://github.com/TeaPearce/Counter-Strike_Behavioural_Cloning" } ] }, { "section": "Dataset", "line": 260, "date": "2021-04", "title": "Counter-Strike Deathmatch with Large-Scale Behavioural Cloning", "venue": "CoG 2022", "keywords": "4M Frames, Online Human Gameplay, FPS", "links": [ { "label": "Paper", "url": "https://ieeexplore.ieee.org/document/9893617" }, { "label": "Code", "url": "https://github.com/TeaPearce/Counter-Strike_Behavioural_Cloning" } ] } ], "sources": [ "https://ieeexplore.ieee.org/document/9893617", "https://github.com/TeaPearce/Counter-Strike_Behavioural_Cloning" ], "search_text": "counter-strike deathmatch with large-scale behavioural cloning counter-strike deathmatch with large-scale behavioural cloning 2021-04 cog 2022 dataset model 4m frames online human gameplay fps large-scale behavioral cloning pixel-to-action player" }, { "id": "work-117f0e382a12", "name": "The NetHack Learning Environment", "title": "The NetHack Learning Environment", "date": "2020-06", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "NeurIPS 2020", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "NetHack", "Long-Horizon", "Partial Observability", "Difficult Environment" ], "keywords_by_category": { "Benchmark": "NetHack, Long-Horizon, Partial Observability, Difficult Environment" }, "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper/2020/hash/569ff987c643b4bedf504efda8f786c2-Abstract.html" }, { "label": "Code", "url": "https://github.com/facebookresearch/nle" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 366, "date": "2020-06", "title": "The NetHack Learning Environment", "venue": "NeurIPS 2020", "keywords": "NetHack, Long-Horizon, Partial Observability, Difficult Environment", "links": [ { "label": "Paper", "url": "https://proceedings.neurips.cc/paper/2020/hash/569ff987c643b4bedf504efda8f786c2-Abstract.html" }, { "label": "Code", "url": "https://github.com/facebookresearch/nle" } ] } ], "sources": [ "https://proceedings.neurips.cc/paper/2020/hash/569ff987c643b4bedf504efda8f786c2-Abstract.html", "https://github.com/facebookresearch/nle" ], "search_text": "the nethack learning environment the nethack learning environment 2020-06 neurips 2020 benchmark nethack long-horizon partial observability difficult environment player" }, { "id": "work-fa8dadce84bd", "name": "Interactive Fiction Games", "title": "Interactive Fiction Games: A Colossal Adventure", "date": "2019-09", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "AAAI 2020", "categories": [ "Benchmark" ], "primary_category": null, "keywords": [ "Jericho", "Interactive Fiction", "Text-based Games", "Transfer" ], "keywords_by_category": { "Benchmark": "Jericho, Interactive Fiction, Text-based Games, Transfer" }, "links": [ { "label": "Paper", "url": "https://doi.org/10.1609/aaai.v34i05.6297" }, { "label": "Code", "url": "https://github.com/microsoft/jericho" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Benchmark", "line": 367, "date": "2019-09", "title": "Interactive Fiction Games: A Colossal Adventure", "venue": "AAAI 2020", "keywords": "Jericho, Interactive Fiction, Text-based Games, Transfer", "links": [ { "label": "Paper", "url": "https://doi.org/10.1609/aaai.v34i05.6297" }, { "label": "Code", "url": "https://github.com/microsoft/jericho" } ] } ], "sources": [ "https://doi.org/10.1609/aaai.v34i05.6297", "https://github.com/microsoft/jericho" ], "search_text": "interactive fiction games: a colossal adventure interactive fiction games 2019-09 aaai 2020 benchmark jericho interactive fiction text-based games transfer player" }, { "id": "work-c8e58b8c9e64", "name": "MineRL", "title": "MineRL: A Large-Scale Dataset of Minecraft Demonstrations", "date": "2019-07", "date_precision": "month", "date_basis": "Month transcribed from the existing README; not independently reverified by this build.", "venue": "IJCAI 2019", "categories": [ "Dataset" ], "primary_category": null, "keywords": [ "60M+ State-Action Pairs", "Imitation Learning" ], "keywords_by_category": { "Dataset": "60M+ State-Action Pairs, Imitation Learning" }, "links": [ { "label": "Paper", "url": "https://www.ijcai.org/Proceedings/2019/339" }, { "label": "Code", "url": "https://github.com/minerllabs/minerl" } ], "summary": null, "release_status": null, "notes": null, "source": "baseline", "focus": "player", "source_rows": [ { "section": "Dataset", "line": 261, "date": "2019-07", "title": "MineRL: A Large-Scale Dataset of Minecraft Demonstrations", "venue": "IJCAI 2019", "keywords": "60M+ State-Action Pairs, Imitation Learning", "links": [ { "label": "Paper", "url": "https://www.ijcai.org/Proceedings/2019/339" }, { "label": "Code", "url": "https://github.com/minerllabs/minerl" } ] } ], "sources": [ "https://www.ijcai.org/Proceedings/2019/339", "https://github.com/minerllabs/minerl" ], "search_text": "minerl: a large-scale dataset of minecraft demonstrations minerl 2019-07 ijcai 2019 dataset 60m+ state-action pairs imitation learning player" } ] }