Formalizes 'Introspection' in LLMs and proves they have privileged access to their own policy logic beyond mere self-simulation.
March 24, 2026
Original Paper
Me, Myself, and $π$ : Evaluating and Explaining LLM Introspection
arXiv · 2603.20276
The Takeaway
Identifies the mechanistic emergence of self-awareness in models, showing they can predict their own future behavior better than peer models can, which changes how we evaluate model reliability and 'hallucination' awareness.
From the abstract
A hallmark of human intelligence is Introspection-the ability to assess and reason about one's own cognitive processes. Introspection has emerged as a promising but contested capability in large language models (LLMs). However, current evaluations often fail to distinguish genuine meta-cognition from the mere application of general world knowledge or text-based self-simulation. In this work, we propose a principled taxonomy that formalizes introspection as the latent computation of specific oper