AI RESEARCH
Correct Chains, Wrong Answers: Dissociating Reasoning from Output in LLM Logic
arXiv CS.AI
•
ArXi:2604.13065v1 Announce Type: cross LLMs can execute every step of chain-of-thought reasoning correctly and still produce wrong final answers. We