Not All Proofs Are Equal: Evaluating LLM Proof Quality Beyond Correctness
cs.CL
Submitted: 2026-05-11
Updated: 2026-11-03
Terminology
Sources
- Extremal descendant integrals on moduli spaces of curves: An inequality discovered and proved in collaboration with AI
- Lattice-Valued Bottleneck Duality
- From Euler to AI: Unifying Formulas for Mathematical Constants
- Proof or Bluff? Evaluating LLMs on 2025 USA Math Olympiad
- The Open Proof Corpus: A Large-Scale Study of LLM-Generated Mathematical Proofs
- IMProofBench: Benchmarking AI on Research-Level Mathematical Proof Generation
- Brains vs. Bytes: Evaluating LLM Proficiency in Olympiad Mathematics
- Towards Robust Mathematical Reasoning
- Mathematical methods and human thought in the age of AI
- Mathematical Proof as a Litmus Test: Revealing Failure Modes of Advanced Large Reasoning Models
- Putnam-like dataset summary: LLMs as mathematical competition contestants
- QEDBENCH: Quantifying the Alignment Gap in Automated Evaluation of University-Level Mathematical Proofs
- LemmaBench: A Live, Research-Level Benchmark to Evaluate LLM Capabilities in Mathematics
- BrokenMath: A Benchmark for Sycophancy in Theorem Proving with LLMs
- Reliable Fine-Grained Evaluation of Natural Language Math Proofs
- Solving Inequality Proofs with Large Language Models
- Hard2Verify: A Step-Level Verification Benchmark for Open-Ended Frontier Math
- RefGrader: Automated Grading of Mathematical Competition Proofs using Agentic Workflows
- Evaluating Language Models for Mathematics through Interactions
- ReliableMath: Benchmark of Reliable Mathematical Reasoning on Large Language Models
Related papers
- ALARM: Audio-Language Alignment for Reasoning Models
- ThinkSum: Probabilistic reasoning over sets using large language models
- Coherence boosting: When your pretrained language model is not paying enough attention
- From Facts to Insights: A Persona-Driven Dual Memory Framework and Dataset for Role-Playing Agents
- Improving General Role-Playing Agents via Psychology-Grounded Reasoning and Role-Aware Policy Optimization
- The Art of Asking: Multilingual Prompt Optimization for Synthetic Data