CWM: An Open-Weights LLM for Research on Code Generation with World Models
This provides a testbed for researchers to explore world modeling for code generation, though it appears incremental as it builds on existing LLM and world modeling concepts.
The researchers tackled the problem of improving code generation by developing CWM, a 32-billion-parameter open-weights LLM that incorporates world modeling through mid-training on observation-action trajectories from computational environments, achieving strong performance such as 65.8% pass@1 on SWE-bench Verified and 96.6% on Math-500.
We release Code World Model (CWM), a 32-billion-parameter open-weights LLM, to advance research on code generation with world models. To improve code understanding beyond what can be learned from training on static code alone, we mid-train CWM on a large amount of observation-action trajectories from Python interpreter and agentic Docker environments, and perform extensive multi-task reasoning RL in verifiable coding, math, and multi-turn software engineering environments. With CWM, we provide a strong testbed for researchers to explore the opportunities world modeling affords for improving code generation with reasoning and planning in computational environments. We present first steps of how world models can benefit agentic coding, enable step-by-step simulation of Python code execution, and show early results of how reasoning can benefit from the latter. CWM is a dense, decoder-only LLM trained with a context size of up to 131k tokens. Independent of its world modeling capabilities, CWM offers strong performance on general coding and math tasks: it reaches pass@1 scores of 65.8% on SWE-bench Verified (with test-time scaling), 68.6% on LiveCodeBench, 96.6% on Math-500, and 76.0% on AIME 2024. To support further research on code world modeling, we release model checkpoints after mid-training, SFT, and RL.