GRAM: A Generative Foundation Reward Model for Reward Generalization
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866915751488651264 |
|---|---|
| author | Wang, Chenglong Gan, Yang Huo, Yifu Mu, Yongyu He, Qiaozhi Yang, Murun Li, Bei Xiao, Tong Zhang, Chunliang Liu, Tongran Zhu, Jingbo |
| author_facet | Wang, Chenglong Gan, Yang Huo, Yifu Mu, Yongyu He, Qiaozhi Yang, Murun Li, Bei Xiao, Tong Zhang, Chunliang Liu, Tongran Zhu, Jingbo |
| contents | In aligning large language models (LLMs), reward models have played an important role, but are standardly trained as discriminative models and rely only on labeled human preference data. In this paper, we explore methods that train reward models using both unlabeled and labeled data. Building on the generative models in LLMs, we develop a generative reward model that is first trained via large-scale unsupervised learning and then fine-tuned via supervised learning. We also show that by using label smoothing, we are in fact optimizing a regularized pairwise ranking loss. This result, in turn, provides a new view of training reward models, which links generative models and discriminative models under the same class of training objectives. The outcome of these techniques is a foundation reward model, which can be applied to a wide range of tasks with little or no further fine-tuning effort. Extensive experiments show that this model generalizes well across several tasks, including response ranking, reinforcement learning from human feedback, and task adaptation with fine-tuning, achieving significant performance improvements over several strong baseline models. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_14175 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | GRAM: A Generative Foundation Reward Model for Reward Generalization Wang, Chenglong Gan, Yang Huo, Yifu Mu, Yongyu He, Qiaozhi Yang, Murun Li, Bei Xiao, Tong Zhang, Chunliang Liu, Tongran Zhu, Jingbo Computation and Language Artificial Intelligence In aligning large language models (LLMs), reward models have played an important role, but are standardly trained as discriminative models and rely only on labeled human preference data. In this paper, we explore methods that train reward models using both unlabeled and labeled data. Building on the generative models in LLMs, we develop a generative reward model that is first trained via large-scale unsupervised learning and then fine-tuned via supervised learning. We also show that by using label smoothing, we are in fact optimizing a regularized pairwise ranking loss. This result, in turn, provides a new view of training reward models, which links generative models and discriminative models under the same class of training objectives. The outcome of these techniques is a foundation reward model, which can be applied to a wide range of tasks with little or no further fine-tuning effort. Extensive experiments show that this model generalizes well across several tasks, including response ranking, reinforcement learning from human feedback, and task adaptation with fine-tuning, achieving significant performance improvements over several strong baseline models. |
| title | GRAM: A Generative Foundation Reward Model for Reward Generalization |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2506.14175 |