Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wu, Lingyan, Zheng, Xiang, Zhai, Weiqi, Wang, Wei, Ren, Xuan, Zhang, Zifan, Wei, Hu, Zhao, Bing
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2604.17282
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915944457043968
author Wu, Lingyan
Zheng, Xiang
Zhai, Weiqi
Wang, Wei
Ren, Xuan
Zhang, Zifan
Wei, Hu
Zhao, Bing
author_facet Wu, Lingyan
Zheng, Xiang
Zhai, Weiqi
Wang, Wei
Ren, Xuan
Zhang, Zifan
Wei, Hu
Zhao, Bing
contents Process-Level Reward Models (PRMs) are essential for guiding complex reasoning in large language models, yet existing PRM benchmarks cover only general domains such as mathematics, failing to address medical reasoning -- which is uniquely characterized by safety criticality, knowledge intensity, and diverse error patterns. Without a reliable medical PRM evaluation framework, we cannot quantify models' error detection capabilities in clinical reasoning, leaving their safety in real-world healthcare applications unverified. We propose MedPRMBench, the first process-level reward model benchmark for the medical domain. Built through a three-phase pipeline based on Clinical Reasoning Blueprints (CRBs), MedPRMBench systematically generates high-quality evaluation data from seven medical QA sources, covering 14 fine-grained error types across three categories (Simplicity, Soundness, and Sensitivity) with the first 4-level severity grading system to quantify clinical impact. The benchmark comprises 6{,}500 questions with 13{,}000 reasoning chains and 113{,}910 step-level labels, plus 6{,}879 questions for training. Our medical PRM baseline achieves an 87.1\% overall PRMScore -- substantially surpassing all baselines -- and serves as a plug-and-play verifier that improves downstream medical QA accuracy by 3.2--6.7 percentage points. Systematic evaluation spanning proprietary frontier models, open-source reasoning models, and medical-specialized models reveals critical weaknesses in current models' medical reasoning error detection capabilities, providing clear directions for future PRM improvement.
format Preprint
id arxiv_https___arxiv_org_abs_2604_17282
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle MedPRMBench: A Fine-grained Benchmark for Process Reward Models in Medical Reasoning
Wu, Lingyan
Zheng, Xiang
Zhai, Weiqi
Wang, Wei
Ren, Xuan
Zhang, Zifan
Wei, Hu
Zhao, Bing
Computation and Language
Process-Level Reward Models (PRMs) are essential for guiding complex reasoning in large language models, yet existing PRM benchmarks cover only general domains such as mathematics, failing to address medical reasoning -- which is uniquely characterized by safety criticality, knowledge intensity, and diverse error patterns. Without a reliable medical PRM evaluation framework, we cannot quantify models' error detection capabilities in clinical reasoning, leaving their safety in real-world healthcare applications unverified. We propose MedPRMBench, the first process-level reward model benchmark for the medical domain. Built through a three-phase pipeline based on Clinical Reasoning Blueprints (CRBs), MedPRMBench systematically generates high-quality evaluation data from seven medical QA sources, covering 14 fine-grained error types across three categories (Simplicity, Soundness, and Sensitivity) with the first 4-level severity grading system to quantify clinical impact. The benchmark comprises 6{,}500 questions with 13{,}000 reasoning chains and 113{,}910 step-level labels, plus 6{,}879 questions for training. Our medical PRM baseline achieves an 87.1\% overall PRMScore -- substantially surpassing all baselines -- and serves as a plug-and-play verifier that improves downstream medical QA accuracy by 3.2--6.7 percentage points. Systematic evaluation spanning proprietary frontier models, open-source reasoning models, and medical-specialized models reveals critical weaknesses in current models' medical reasoning error detection capabilities, providing clear directions for future PRM improvement.
title MedPRMBench: A Fine-grained Benchmark for Process Reward Models in Medical Reasoning
topic Computation and Language
url https://arxiv.org/abs/2604.17282