Saved in:
Bibliographic Details
Main Authors: Zhao, Yiyang, Bai, Huiyu, Zhao, Xuejiao
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.08965
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910998967877632
author Zhao, Yiyang
Bai, Huiyu
Zhao, Xuejiao
author_facet Zhao, Yiyang
Bai, Huiyu
Zhao, Xuejiao
contents The ability to train high-performing reward models with few-shot data is critical for enhancing the efficiency and scalability of Reinforcement Learning from Human Feedback (RLHF). We propose a data augmentation and expansion framework that enables generative reward models trained on small datasets to achieve comparable performance to those trained on large-scale datasets. Traditional methods to train a generative reward model, such as Direct Preference Optimization (DPO), are constrained by inefficiencies in sample pairing and limited data diversity. This work introduces preference refinement, which employs Chain-of-Thought (CoT) sampling to uncover diverse and high-quality preference relationships. It also incorporates a perplexity-based scoring mechanism to assign nuanced preference levels and utilizes Multi-level Direct Preference Optimization (M-DPO) to enable the model to capture finer-grained preference differences between samples. Experimental results demonstrate that the proposed method significantly enhances data efficiency and model performance, enabling reward models trained in a few-shot setting to achieve results on par with those trained on large-scale datasets. This study underscores the potential of data-efficient strategies in advancing reward model optimization, offering a robust solution for low-resource RLHF applications.
format Preprint
id arxiv_https___arxiv_org_abs_2506_08965
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle GFRIEND: Generative Few-shot Reward Inference through EfficieNt DPO
Zhao, Yiyang
Bai, Huiyu
Zhao, Xuejiao
Machine Learning
Artificial Intelligence
The ability to train high-performing reward models with few-shot data is critical for enhancing the efficiency and scalability of Reinforcement Learning from Human Feedback (RLHF). We propose a data augmentation and expansion framework that enables generative reward models trained on small datasets to achieve comparable performance to those trained on large-scale datasets. Traditional methods to train a generative reward model, such as Direct Preference Optimization (DPO), are constrained by inefficiencies in sample pairing and limited data diversity. This work introduces preference refinement, which employs Chain-of-Thought (CoT) sampling to uncover diverse and high-quality preference relationships. It also incorporates a perplexity-based scoring mechanism to assign nuanced preference levels and utilizes Multi-level Direct Preference Optimization (M-DPO) to enable the model to capture finer-grained preference differences between samples. Experimental results demonstrate that the proposed method significantly enhances data efficiency and model performance, enabling reward models trained in a few-shot setting to achieve results on par with those trained on large-scale datasets. This study underscores the potential of data-efficient strategies in advancing reward model optimization, offering a robust solution for low-resource RLHF applications.
title GFRIEND: Generative Few-shot Reward Inference through EfficieNt DPO
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2506.08965