Investigating LLM's Problem Solving Capability -- a Study on Statics Questions
Researchers at arXiv found that large language models (LLMs) perform well on text-only statics problems but their accuracy drops when diagrams are added and multi-step reasoning is required. The perfo…