A step-level verification framework for LLMs on research-level proofs from the FirstProof benchmark outperforms global methods by enforcing per-step context and theorem constraints, shifting errors from hallucinations to pedantic rejections.
Boundedness of total Cartier indices for rational singularities in families
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
We show that the total Cartier index of varieties with rational singularities in a bounded family is bounded. This solves a problem of Han and Jiang. The overall structure of the proof, which treats the surface case and the higher-dimensional case separately, was originated by generative AI, particularly the Rethlas system, and was substantially corrected and elaborated by hand.
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Evaluating Research-Level Math Proofs via Strict Step-Level Verification
A step-level verification framework for LLMs on research-level proofs from the FirstProof benchmark outperforms global methods by enforcing per-step context and theorem constraints, shifting errors from hallucinations to pedantic rejections.