{ "added_at": "2026-10-17", "arxiv_id": "2405.15383", "arxiv_version_reviewed": null, "authors": [ "Nicola Dainese", "Matteo Merler", "Minttu Alakuijala", "Pekka Marttinen" ], "bibtex_url": "https://awesomerea.org/papers/citations/paper-code-world-models.bib", "citation": { "container_title": "Advances in Neural Information Processing Systems", "pages": "60429-60474", "sources": [ "https://proceedings.neurips.cc/paper_files/paper/2024/hash/6f479ea488e0908ac8b1b37b27fd134c-Abstract-Conference.html", "https://api.crossref.org/works/10.52202/079017-1933" ], "volume": "37" }, "code_url": "https://github.com/nicoladainese96/code-world-models", "date_basis": "The final NeurIPS proceedings and publisher-deposited Crossref record identify 2024. No publication month or day is asserted; conference event dates are not substituted for publication dates.", "date_basis_zh": "正式 NeurIPS 论文集及出版方提交的 Crossref 记录给出 2024 年。不推定出版月份或日期，也不把会议举办日期当作出版日期。", "date_precision": "year", "description_en": "GIF-MCTS generates executable Python transition and reward models from natural-language environment descriptions and offline state-transition trajectories, using prediction feedback to generate, improve and repair code. CWMB evaluates 18 reinforcement-learning environments and downstream planning; a separate experiment uses a simplified Read to Fight Monsters grid world.", "description_zh": "GIF-MCTS 根据环境的自然语言描述与离线状态转移轨迹生成可执行的 Python 状态转移和奖励模型，并利用预测反馈生成、改进和修复代码。CWMB 评估 18 个强化学习环境及后续规划；另一个实验使用简化的 Read to Fight Monsters 网格世界。", "doi": "10.52202/079017-1933", "entry_url": "https://awesomerea.org/papers/paper-code-world-models/", "entry_url_zh": "https://awesomerea.org/zh/papers/paper-code-world-models/", "evidence_urls": [ "https://proceedings.neurips.cc/paper_files/paper/2024/hash/6f479ea488e0908ac8b1b37b27fd134c-Abstract-Conference.html", "https://proceedings.neurips.cc/paper_files/paper/2024/file/6f479ea488e0908ac8b1b37b27fd134c-Paper-Conference.pdf", "https://api.crossref.org/works/10.52202/079017-1933", "https://arxiv.org/abs/2405.15383", "https://sites.google.com/view/code-world-models/home", "https://github.com/nicoladainese96/code-world-models/blob/e17e787b41137793bcef9f860bdfe9e6fc07b768/README.md", "https://github.com/nicoladainese96/code-world-models/blob/e17e787b41137793bcef9f860bdfe9e6fc07b768/LICENSE", "https://github.com/nicoladainese96/code-world-models/blob/e17e787b41137793bcef9f860bdfe9e6fc07b768/RTFM/LICENSE", "https://github.com/nicoladainese96/code-world-models/blob/e17e787b41137793bcef9f860bdfe9e6fc07b768/environment.yml", "https://github.com/nicoladainese96/code-world-models/blob/e17e787b41137793bcef9f860bdfe9e6fc07b768/src/mcts/gif_mcts/simulators/simulator_cwm.py", "https://github.com/nicoladainese96/code-world-models/blob/e17e787b41137793bcef9f860bdfe9e6fc07b768/src/cwm_benchmark/generate_env.py", "https://github.com/nicoladainese96/code-world-models/blob/e17e787b41137793bcef9f860bdfe9e6fc07b768/src/experiments/eval_cwm_planning.py", "https://github.com/nicoladainese96/code-world-models/blob/e17e787b41137793bcef9f860bdfe9e6fc07b768/data/prompts/rtfm/generate.md" ], "first_arxiv_date": "2024-05-24", "id": "paper-code-world-models", "json_url": "https://awesomerea.org/papers/paper-code-world-models/record.json", "kind": "research", "markdown_url": "https://awesomerea.org/papers/paper-code-world-models/index.md", "markdown_url_zh": "https://awesomerea.org/zh/papers/paper-code-world-models/index.md", "metadata_notes": "Reviewed the final proceedings paper, pages 60429–60474. Its WorldCoder baseline follows the first arXiv version, not later revisions. The Apache-2.0 repository provides prompts, replay-buffer files and evaluation scripts; bundled RTFM carries CC BY-NC 4.0. The pinned evaluator balances reward classes, grants partial credit over next-state components, skips terminal next-state checks, and accepts nonempty valid-action subsets for RTFM; continuous states use 1e-5 relative/absolute tolerance. The planning script defaults to two episodes, while the paper reports ten. Reproduction needs the Linux-oriented Conda/Python 3.10 environment, Gymnasium/MuJoCo and model access.", "metadata_notes_zh": "核查的是正式论文集版本，页码 60429–60474。其中的 WorldCoder 基线依据首个 arXiv 版本，而非后续修订版。Apache-2.0 仓库提供提示词、回放缓冲区文件和评估脚本；随附 RTFM 采用 CC BY-NC 4.0。固定版本评估器对奖励类别作平衡加权，按下一状态分量给予部分分数，跳过终止转移的下一状态核对，并允许 RTFM 返回有效动作集合的非空子集；连续状态采用 1e-5 的相对和绝对容差。规划脚本默认运行两个回合，而论文报告十个回合。复现需要面向 Linux 的 Conda/Python 3.10 环境、Gymnasium/MuJoCo 及模型访问权限。", "pdf_url": "https://proceedings.neurips.cc/paper_files/paper/2024/file/6f479ea488e0908ac8b1b37b27fd134c-Paper-Conference.pdf", "project_url": "https://sites.google.com/view/code-world-models/home", "publication_date": null, "publication_status": "conference-paper", "record_type": "paper", "runtime_tested": false, "scope_limitations_en": "Assumes deterministic, fully observable dynamics and sufficiently detailed descriptions; this is not recovery from pixels or binaries alone. Accuracy uses the replay transitions supplied during synthesis, alongside planning-return evaluation, rather than proving correctness on all unseen states or full-engine equivalence. RTFM fixes the manual across episodes and keeps monsters stationary. CWMB also includes non-game control and physics tasks; some planning returns are unavailable and complex continuous-state models remain difficult.", "scope_limitations_zh": "假设环境动态具有确定性、状态完全可观测，并且描述足够详细；并非仅凭像素或二进制进行恢复。准确率使用合成过程中提供的回放转移，并结合规划回报评估，不能证明所有未见状态上的正确性或完整引擎等价。RTFM 在不同回合间固定手册内容，并让怪物保持静止。CWMB 也包含非游戏的控制与物理任务；部分规划回报未给出，复杂连续状态模型仍然困难。", "targets": [ "game" ], "title": "Generating Code World Models with Large Language Models Guided by Monte Carlo Tree Search", "title_zh": "通过蒙特卡洛树搜索引导大语言模型生成代码世界模型", "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/6f479ea488e0908ac8b1b37b27fd134c-Abstract-Conference.html", "venue": "NeurIPS 2024", "verification": "source-reviewed", "verified_at": "2026-10-17", "why_fits_en": "Shows how documented rules and observed transitions in game-like environments can become executable behavior models with test feedback. It complements game reconstruction without claiming recovery of original engines, graphics or assets.", "why_fits_zh": "展示如何借助测试反馈，将类游戏环境中的已记录规则和已观察转移转为可执行行为模型。它可补充游戏重建研究，但不声称恢复原始引擎、图形或资产。", "year": 2024 }