MAMath-Bench: Process-Aware and Robust Evaluation of Multimodal Multi-Agent Mathematical Reasoning

DOI10.48550/ARXIV.2610.26943MaRDI QIDQ7400251arXivFDO

Authors Xinping Wang, Jianhao Deng, Zihao Zhou, Jie Ma, Rui Xing, Wensong Guo, Xiayi Zhang, Boqian Yang, Feijuan He

Publication date 18 October 2026

Published in arXiv

Tldr MAMath-Bench evaluates multimodal mathematical reasoning in multi-agent systems (MASs) with a novel process-aware and robust evaluation framework covering problem-solving correctness, reasoning-process reliability, and robustness to adversarial perturbations. It provides the first large-scale benchmark for MASs with fine-grained milestone annotations and multimodal problems, yielding insights into strengths and weaknesses of current systems.

Full work available at URL https://doi.org/10.48550/ARXIV.2610.26943

arXiv classification

cs.AI

This page was built for publication: MAMath-Bench: Process-Aware and Robust Evaluation of Multimodal Multi-Agent Mathematical Reasoning

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7400251)