Exploring LLMs for Ranking Mathematical Truth Probabilities

PTenigma · x · 2026-08-21

The author suggests that heavily used LLMs like Claude Opus could generate a sorted list of p(correct) for sets of papers, potentially creating an interesting branch of mathematical epistemology. This is linked to the author's 1998 paper on how expert mathematicians estimate the probability of correctness based on analogy, experience, and logical proof.

Original post →

More from Research

Research channel →