Uncertainty-aware Reward Model: Teaching Reward Models to Know What is Unknown
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909488837033984 |
|---|---|
| author | Lou, Xingzhou Yan, Dong Shen, Wei Yan, Yuzi Xie, Jian Zhang, Junge |
| author_facet | Lou, Xingzhou Yan, Dong Shen, Wei Yan, Yuzi Xie, Jian Zhang, Junge |
| contents | Reward models (RMs) are essential for aligning large language models (LLM) with human expectations. However, existing RMs struggle to capture the stochastic and uncertain nature of human preferences and fail to assess the reliability of reward predictions. To address these challenges, we introduce the Uncertainty-aware Reward Model (URM) and its ensemble variant, URME. URM employs a probabilistic value head to capture aleatoric uncertainty by modeling the distribution of disentangled human preference attributes. URME further quantifies epistemic uncertainty by examining discrepancies among individual URMs within the ensemble, enabling identification of unreliable evaluations. Our empirical evaluations demonstrate that URM achieves strong performance on RewardBench, outperforming competitive large-scale models. Additionally, extensive experiments, including best-of-n sampling (BoN), iterative direct preference optimization (iterative DPO), and proximal policy optimization (PPO), demonstrate that URM and URME significantly enhance LLMs' generation quality. Notably, reward predictions with lower uncertainty are far more reliable, demonstrate significantly higher quality, and result in substantially improved alignment. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_00847 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Uncertainty-aware Reward Model: Teaching Reward Models to Know What is Unknown Lou, Xingzhou Yan, Dong Shen, Wei Yan, Yuzi Xie, Jian Zhang, Junge Machine Learning Reward models (RMs) are essential for aligning large language models (LLM) with human expectations. However, existing RMs struggle to capture the stochastic and uncertain nature of human preferences and fail to assess the reliability of reward predictions. To address these challenges, we introduce the Uncertainty-aware Reward Model (URM) and its ensemble variant, URME. URM employs a probabilistic value head to capture aleatoric uncertainty by modeling the distribution of disentangled human preference attributes. URME further quantifies epistemic uncertainty by examining discrepancies among individual URMs within the ensemble, enabling identification of unreliable evaluations. Our empirical evaluations demonstrate that URM achieves strong performance on RewardBench, outperforming competitive large-scale models. Additionally, extensive experiments, including best-of-n sampling (BoN), iterative direct preference optimization (iterative DPO), and proximal policy optimization (PPO), demonstrate that URM and URME significantly enhance LLMs' generation quality. Notably, reward predictions with lower uncertainty are far more reliable, demonstrate significantly higher quality, and result in substantially improved alignment. |
| title | Uncertainty-aware Reward Model: Teaching Reward Models to Know What is Unknown |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2410.00847 |