AdvancedMathBench: Benchmark Suite for Advanced Mathematical Proof Generation and Verification
InternLM
Releases ProverBench (296 undergraduate-to-doctoral-qualifying-exam problems) and VerifierBench (888 model-generated proofs with expert annotations), paired with an automated proof-verification pipeline that detects fine-grained logical errors and aligns strongly with expert human judgment. Frontier models achieve only 75.8% and 66.1% on the respective splits, revealing large headroom in advanced mathematical reasoning.
Why it matters
Existing math benchmarks saturate at final-answer accuracy and underweight proof validity; this suite exposes that current reasoning models still make systematic logical errors in graduate-level proofs—a cleaner signal for next-generation reasoning research. Received 13 upvotes on HuggingFace Daily Papers.
Importance: 2/5
HuggingFace Daily Papers (13 upvotes); two-part math proof benchmark from InternLM/Shanghai AI Lab exposing frontier model gaps