AwesomeREA.org

2024 · Conference paper

WorldCoder, a Model-Based LLM Agent: Building World Models by Writing Code and Interacting with the Environment

Hao Tang; Darren Key; Kevin Ellis

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.

Scope and limitations

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.

Why it fits this collection

A foundation for agent-assisted reconstruction of observed game behavior as executable code, with explicit consistency checks instead of a visual-only similarity claim.

Publication and review details

Venue
NeurIPS 2024
Status
Conference paper
Publication / posting date
2024 (exact date not verified)
Date source
The NeurIPS proceedings and publisher-deposited Crossref record identify 2024; no publication month or day is asserted.
First arXiv submission
February 19, 2024
arXiv
2402.12275
DOI
10.52202/079017-2243

Reviewed 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. No installation or model calls were performed.

Sources reviewed

Source review only; experiments were not rerun for this collection. Cite the original authors using the linked paper or BibTeX.

Primary sources

  1. https://papers.nips.cc/paper_files/paper/2024/hash/820c61a0cd419163ccbd2c33b268816e-Abstract-Conference.html
  2. https://papers.neurips.cc/paper_files/paper/2024/file/820c61a0cd419163ccbd2c33b268816e-Paper-Conference.pdf
  3. https://api.crossref.org/works/10.52202/079017-2243
  4. https://arxiv.org/abs/2402.12275
  5. https://haotang1995.github.io/projects/worldcoder
  6. https://github.com/haotang1995/WorldCoder/blob/9d9fc8d55b7537e2cc1e667a13185a0f2c9d9404/README.md
  7. https://github.com/haotang1995/WorldCoder/blob/9d9fc8d55b7537e2cc1e667a13185a0f2c9d9404/LICENSE
  8. https://github.com/haotang1995/WorldCoder/blob/9d9fc8d55b7537e2cc1e667a13185a0f2c9d9404/requirements.txt
  9. https://github.com/haotang1995/WorldCoder/blob/9d9fc8d55b7537e2cc1e667a13185a0f2c9d9404/worldcoder/envs/__init__.py
  10. https://github.com/haotang1995/WorldCoder/blob/9d9fc8d55b7537e2cc1e667a13185a0f2c9d9404/worldcoder/agent/synthesizer/evaluator.py