Paper: LLMs' latent beliefs can predict their behavior
A new paper by Anthropic researcher brwilder shows that modeling LLMs as holding latent beliefs about unknown states enables high-precision prediction of their behavior, with accuracy rising sharply as general capability increases.
2026-09-10 ~ 2026-09-10 · 2 related posts
- New paper formalizes the 'intentional stance': attributing beliefs to LLMs predicts behavior well — brwilder · 2026-09-10
- Modeling LLMs as holding latent beliefs yields high-accuracy behavior predictions — brwilder · 2026-09-10