What Makes a Latent Thought Valid? Defining and Auditing Thought Representations in LLMs
The research introduces rigorous frameworks to audit latent thought representations from Large Language Models (LLMs). By using four functional axioms—Causality, Minimality, Separability, and Stability—the authors move beyond simple benchmark scores. The findings reveal that while representations can distinguish between different tasks, they fail structurally to separate individual questions within the same task. This identifies a fundamental flaw in how LLMs encode thought processes internally.