{
  "added_at": "2026-10-17",
  "arxiv_id": "2402.12275",
  "arxiv_version_reviewed": null,
  "authors": [
    "Hao Tang",
    "Darren Key",
    "Kevin Ellis"
  ],
  "bibtex_url": "https://awesomerea.org/papers/citations/paper-worldcoder.bib",
  "citation": {
    "container_title": "Advances in Neural Information Processing Systems",
    "pages": "70148-70212",
    "sources": [
      "https://papers.nips.cc/paper_files/paper/2024/hash/820c61a0cd419163ccbd2c33b268816e-Abstract-Conference.html",
      "https://api.crossref.org/works/10.52202/079017-2243"
    ],
    "volume": "37"
  },
  "code_url": "https://github.com/haotang1995/WorldCoder",
  "date_basis": "The NeurIPS proceedings and publisher-deposited Crossref record identify 2024; no publication month or day is asserted.",
  "date_basis_zh": "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。",
  "date_precision": "year",
  "description_en": "WorldCoder learns executable Python transition and reward models from environment interactions and natural-language goals. It refines code against recorded experiences and an optimistic planning constraint, studying transfer and task solving in Sokoban, MiniGrid and ALFWorld.",
  "description_zh": "WorldCoder 根据环境交互和自然语言目标学习可执行的 Python 状态转移与奖励模型，利用已记录经验和乐观规划约束修正代码，并在 Sokoban、MiniGrid 与 ALFWorld 中研究迁移和任务求解。",
  "doi": "10.52202/079017-2243",
  "entry_url": "https://awesomerea.org/papers/paper-worldcoder/",
  "entry_url_zh": "https://awesomerea.org/zh/papers/paper-worldcoder/",
  "evidence_urls": [
    "https://papers.nips.cc/paper_files/paper/2024/hash/820c61a0cd419163ccbd2c33b268816e-Abstract-Conference.html",
    "https://papers.neurips.cc/paper_files/paper/2024/file/820c61a0cd419163ccbd2c33b268816e-Paper-Conference.pdf",
    "https://api.crossref.org/works/10.52202/079017-2243",
    "https://arxiv.org/abs/2402.12275",
    "https://haotang1995.github.io/projects/worldcoder",
    "https://github.com/haotang1995/WorldCoder/blob/9d9fc8d55b7537e2cc1e667a13185a0f2c9d9404/README.md",
    "https://github.com/haotang1995/WorldCoder/blob/9d9fc8d55b7537e2cc1e667a13185a0f2c9d9404/LICENSE",
    "https://github.com/haotang1995/WorldCoder/blob/9d9fc8d55b7537e2cc1e667a13185a0f2c9d9404/requirements.txt",
    "https://github.com/haotang1995/WorldCoder/blob/9d9fc8d55b7537e2cc1e667a13185a0f2c9d9404/worldcoder/envs/__init__.py",
    "https://github.com/haotang1995/WorldCoder/blob/9d9fc8d55b7537e2cc1e667a13185a0f2c9d9404/worldcoder/agent/synthesizer/evaluator.py"
  ],
  "first_arxiv_date": "2024-02-19",
  "id": "paper-worldcoder",
  "json_url": "https://awesomerea.org/papers/paper-worldcoder/record.json",
  "kind": "research",
  "markdown_url": "https://awesomerea.org/papers/paper-worldcoder/index.md",
  "markdown_url_zh": "https://awesomerea.org/zh/papers/paper-worldcoder/index.md",
  "metadata_notes": "The final NeurIPS paper, pages 70148–70212. The MIT author repository is a partial experiment artifact: its pinned environment registry exposes MiniGrid only, despite the README’s Sokoban example and broader paper experiments. The README warns about NumPy 2 incompatibility and older Gymnasium/MiniGrid dependencies.",
  "metadata_notes_zh": "正式 NeurIPS 论文，页码 70148–70212。作者仓库采用 MIT，但仅提供部分实验产物：固定版本的环境注册表只暴露 MiniGrid，尽管 README 给出 Sokoban 示例、论文也评估了更多环境。README 提醒 NumPy 2 不兼容及较旧的 Gymnasium/MiniGrid 依赖问题。",
  "pdf_url": "https://papers.neurips.cc/paper_files/paper/2024/file/820c61a0cd419163ccbd2c33b268816e-Paper-Conference.pdf",
  "project_url": "https://haotang1995.github.io/projects/worldcoder",
  "publication_date": null,
  "publication_status": "conference-paper",
  "record_type": "paper",
  "runtime_tested": false,
  "scope_limitations_en": "Assumes deterministic dynamics and symbolic, discrete states rather than recovering a game from arbitrary pixels or binaries. Consistency with observed transitions and successful planning do not establish complete engine equivalence or correctness on every unseen state. Evaluation concerns bounded environment models and planning, not original source or asset recovery.",
  "scope_limitations_zh": "假设环境动态具有确定性，状态为符号化的离散表示，并非从任意像素或二进制恢复游戏。与已观察转移一致及规划成功，不能证明完整引擎等价或所有未见状态上的正确性。评估针对有限范围的环境模型与规划，不涉及恢复原始源码或资产。",
  "targets": [
    "game"
  ],
  "title": "WorldCoder, a Model-Based LLM Agent: Building World Models by Writing Code and Interacting with the Environment",
  "title_zh": "WorldCoder：通过编写代码并与环境交互构建世界模型的基于模型的语言智能体",
  "url": "https://papers.nips.cc/paper_files/paper/2024/hash/820c61a0cd419163ccbd2c33b268816e-Abstract-Conference.html",
  "venue": "NeurIPS 2024",
  "verification": "source-reviewed",
  "verified_at": "2026-10-17",
  "why_fits_en": "A foundation for agent-assisted reconstruction of observed game behavior as executable code, with explicit consistency checks instead of a visual-only similarity claim.",
  "why_fits_zh": "这是将已观察游戏行为重建为可执行代码的智能体辅助基础研究，采用明确的一致性检查，而非仅以视觉相似度作为依据。",
  "year": 2024
}