Beyond the First Error: Process Reward Models for Reflective Mathematical Reasoning

TL;DR

This paper introduces a novel annotation method for process reward models that better captures long chain-of-thought reasoning, including self-correction, leading to improved performance in mathematical reasoning tasks.

Contribution

The paper proposes a new data annotation technique for process reward models that accounts for self-correction and reflection in long reasoning chains, trained on 1.7 million samples.

Findings

PRM outperforms existing models on multiple metrics

The annotation method improves data efficiency and performance

The approach demonstrates stability and generalizability

Abstract

Many studies focus on data annotation techniques for training effective PRMs. However, current methods encounter a significant issue when applied to long CoT reasoning processes: they tend to focus solely on the first incorrect step and all preceding steps, assuming that all subsequent steps are incorrect. These methods overlook the unique self-correction and reflection mechanisms inherent in long CoT, where correct reasoning steps may still occur after initial reasoning mistakes. To address this issue, we propose a novel data annotation method for PRMs specifically designed to score the long CoT reasoning process. Given that under the reflection pattern, correct and incorrect steps often alternate, we introduce the concepts of Error Propagation and Error Cessation, enhancing PRMs' ability to identify both effective self-correction behaviors and reasoning based on erroneous steps. Leveraging an LLM-based judger for annotation, we collect 1.7 million data samples to train a 7B PRM and evaluate it at both solution and step levels. Experimental results demonstrate that compared to existing open-source PRMs and PRMs trained on open-source datasets, our PRM achieves superior performance across various metrics, including search guidance, BoN, and F1 scores. Compared to widely used MC-based annotation methods, our annotation approach not only achieves higher data efficiency but also delivers superior performance. Detailed analysis is also conducted to demonstrate the stability and generalizability of our method.