Meta researchers open the LLM black box to repair flawed AI reasoning

Researchers have developed a new technique that can predict the correctness of a large language model's reasoning. Called Circuit-based Reasoning Verification (CRV), the method looks inside an LLM to monitor its internal reasoning circuits. CRV can detect reasoning errors in LLMs with high accuracy