English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

AdvancedMathBench: A Benchmark Suite for Evaluating Advanced Mathematical Reasoning in LLMs

Forum topic · 小凯 · 2026-07-15

Summary

AdvancedMathBench (arXiv:2607.11849) is a benchmark suite designed to evaluate large language models' advanced mathematical reasoning beyond high-school and olympiad-level problems. Its core proof-generation benchmark, ProverBench, contains 296 problems spanning undergraduate and doctoral qualifying-exam difficulty. Because final-answer checking is inadequate for proofs, the authors built an automatic verification pipeline trained on large-scale expert annotations that outputs correctness verdicts plus fine-grained assessments of proof errors, achieving strong agreement with human experts on held-out proof traces. A companion benchmark, VerifierBench, pairs 888 model-generated proof traces with expert ground truth to test whether models can correctly judge proof validity and give sound verification rationales. Results show the suite remains challenging for frontier models: the best proof-generation model, GPT-5.5-xhigh, scores only 75.8 and 66.1 on the UGD and QE splits, while the best model on proof verification reaches a Balanced F1 of just 65.1, with low true-negative rates—detecting critical errors in proofs remains the main bottleneck. The suite addresses gaps in disciplinary coverage and evaluation granularity in existing math benchmarks.

Overview

  • Field: NLP
  • Authors: Lingkai Kong, Zijian Wu, Yuzhe Gu, Haiteng Zhao, Wenyong Huang
  • Published: 2026-07-13
  • arXiv: 2607.11849
  • Summary

    Large language models (LLMs) have achieved remarkable performance on high-school and olympiad-style mathematics, yet their capabilities on advanced mathematics remain poorly understood. Existing benchmarks fall short in both scope and evaluation granularity: they offer limited disciplinary coverage and often rely on final-answer correctness or coarse judgments, leaving the validity of the reasoning process inadequately assessed.

    To bridge this gap, the authors introduce AdvancedMathBench, a benchmark suite for evaluating advanced mathematical reasoning:

  • ProverBench: the core proof-generation benchmark, containing 296 problems at undergraduate and doctoral qualifying-exam levels.
  • Automatic verification pipeline: trained on large-scale expert annotations, it outputs correctness verdicts and fine-grained assessments of proof errors, and shows high agreement with human experts on held-out proof traces.
  • VerifierBench: 888 model-generated proof traces paired with expert ground truth, testing whether models can correctly judge proof validity and provide sound verification rationales.
  • Key Findings

  • AdvancedMathBench remains challenging for frontier models.
  • Best proof-generation model GPT-5.5-xhigh scores only 75.8 (UGD split) and 66.1 (QE split).
  • On proof verification, the best model achieves a Balanced F1 of only 65.1, with a low true-negative rate.
  • Detecting critical errors in proofs is the main bottleneck for current models.

Original Abstract (excerpt)

> Large language models (LLMs) have achieved remarkable performance on high-school and olympiad-style mathematics, yet their capabilities on advanced mathematics remain poorly understood. Existing benchmarks, however, fall short in both scope and evaluation granularity: they provide limited disciplinary coverage and often rely on final-answer correctness or coarse judgments, leaving the validity of the reasoning process inadequately assessed. To bridge this gap, we introduce AdvancedMathBench, a benchmark suite designed to evaluate advanced mathematical reasoning capabilities. Its core proof-generation benchmark, ProverBench, contains 296 problems spanning undergraduate and doctoral qualifying-exam levels. To provide reliable evaluation of the proofs, we develop a dedicated automatic verific...

---

*Auto-collected on 2026-07-15.*

Tags

#llm#benchmark#mathematical-reasoning#proof-verification#nlp#arxiv#evaluation

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178395148