← All reports

AI-generated formal specifications do not guarantee the correctness of the original natural language proofs.

AIFormal VerificationMathematicsOct 8, 2026score 0.272 posts · 0 replies across 2 instances
The thread discusses a research paper regarding the limitations of using AI to autoformalize natural language proofs into Lean. It highlights that while AI can generate formal specifications, these may not accurately reflect the original natural language proofs and can sometimes 'cheat' during the translation process.

Claims

AI-generated formal specifications do not guarantee the correctness of the original natural language proofs.
Parent: Artificial IntelligenceEntity: AI autoformalisationSub-entity: Lean verificationImpact: negativeDate: Oct 8, 2026Target: AI autoformalisation

Source posts

@[email protected]
https://www.reddit.com/r/BetterOffline/comments/1x0b49i/navierstokes_lost_in_translation_why_lean/ Title: Navier-Stokes lost in translation: Why Lean verification of AI autoformalisation does not guarantee correct natural language proofs
0 boosts · 0 favs · 0 replies · Oct 8, 2026
@[email protected]
An interesting paper: "Navier-Stokes lost in translation: Why Lean verification of AI autoformalisation does not guarantee correct natural language proofs". Coincidentally, we're starting to work on autoformalisation ourselves, i.e., translating statements written in natural language (e.g., a theorem) into a formal specification (e.g., in Lean). And it seems that current models have a tendency to "cheat" 😅 https://arxiv.org/abs/2610.08144
0 boosts · 0 favs · 0 replies · Oct 8, 2026
@[email protected]
💡 Navier-Stokes lost in translation: Why Lean verification of AI autoformalisation does not guarantee correct natural language proofs https://arxiv.org/abs/2610.08144 #ai #math #science #lean #navierstokes #paper
7 boosts · 0 favs · 0 replies · Oct 8, 2026
#ai#math#science#lean#navierstokes#paper