Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2506.16712 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908414798462976 |
|---|---|
| author | Chen, Bin Gao, Xinzge Hu, Chuanrui Yu, Penghang Zhang, Hua Bao, Bing-Kun |
| author_facet | Chen, Bin Gao, Xinzge Hu, Chuanrui Yu, Penghang Zhang, Hua Bao, Bing-Kun |
| contents | Generative Reward Models (GRMs) provide greater flexibility than scalar reward models in capturing human preferences, but their effectiveness is limited by poor reasoning capabilities. This often results in incomplete or overly speculative reasoning paths, leading to hallucinations or missing key information in complex tasks. We address this challenge with ReasonGRM, a three-stage generative reward modeling framework. In the first stage, Zero-RL is used to generate concise, outcome-directed reasoning paths that reduce the likelihood of critical omissions. In the second stage, we introduce a novel evaluation metric, $R^\star$, which scores reasoning paths based on their generation likelihood. This favors paths that reach correct answers with minimal exploration, helping to reduce hallucination-prone data during training. In the final stage, the model is further refined through reinforcement learning on challenging examples to enhance its preference discrimination capabilities. Experiments on three public benchmarks show that ReasonGRM achieves competitive or state-of-the-art performance, outperforming previous best GRMs by 1.8\% on average and surpassing proprietary models such as GPT-4o by up to 5.6\%. These results demonstrate the effectiveness of reasoning-aware training and highlight the importance of high-quality rationale selection for reliable preference modeling. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_16712 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | ReasonGRM: Enhancing Generative Reward Models through Large Reasoning Models Chen, Bin Gao, Xinzge Hu, Chuanrui Yu, Penghang Zhang, Hua Bao, Bing-Kun Computation and Language Artificial Intelligence Generative Reward Models (GRMs) provide greater flexibility than scalar reward models in capturing human preferences, but their effectiveness is limited by poor reasoning capabilities. This often results in incomplete or overly speculative reasoning paths, leading to hallucinations or missing key information in complex tasks. We address this challenge with ReasonGRM, a three-stage generative reward modeling framework. In the first stage, Zero-RL is used to generate concise, outcome-directed reasoning paths that reduce the likelihood of critical omissions. In the second stage, we introduce a novel evaluation metric, $R^\star$, which scores reasoning paths based on their generation likelihood. This favors paths that reach correct answers with minimal exploration, helping to reduce hallucination-prone data during training. In the final stage, the model is further refined through reinforcement learning on challenging examples to enhance its preference discrimination capabilities. Experiments on three public benchmarks show that ReasonGRM achieves competitive or state-of-the-art performance, outperforming previous best GRMs by 1.8\% on average and surpassing proprietary models such as GPT-4o by up to 5.6\%. These results demonstrate the effectiveness of reasoning-aware training and highlight the importance of high-quality rationale selection for reliable preference modeling. |
| title | ReasonGRM: Enhancing Generative Reward Models through Large Reasoning Models |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2506.16712 |