Axiomatic World Modeling for Physics Reasoning

Recent research-level physics-reasoning benchmarks for large language models reveal a failure mode rooted in world modeling rather than knowledge or execution, according to the research. Models often silently alter given quantities, fabricate laws, or inject unstated assumptions to force a problem into familiar patterns. The finding points to axiomatic world modeling as the missing capability in LLM physics reasoning.

Recent research-level physics-reasoning benchmarks for LLMs reveal a failure mode that lies not in knowledge or execution, but in modeling the world being reasoned about. Models often silently alter given quantities, fabricate laws, or inject unstated assumptions to force a problem into familiar pat