Code-as-World: Representing Physical Worlds as Executable Code for Reasoning

The paper "Code as Worlds" represents physical environments as executable code, discovered through iterative simulation and verification from raw observations. This gives vision-language models quantitative reasoning ability and scalable supervision signals, improving physical scene understanding and manipulation.

2026-08-31 ~ 2026-08-31 · 2 related posts

• Code-as-World: Agentic Discovery of Executable World Representations for Physical Reasoning — MirroS-Lab · 2026-08-31
• Code as Worlds: Agentic Discovery of Executable World Representations — _akhaliq · 2026-08-31