Informal mathematical proofs produced directly by LLMs are notoriously prone to hallucination. Apparent elegant argumentations from AI frequently demand substantial manual verification work from human ...

Should we trust AI-generated formal proofs in Lean 4? – mathoverflow.net
cyc
Tags

