Holistic Utility Preference Learning for Listwise Alignment

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Zhou, Jiacong, Wang, Xianyun, Zhang, Min, Yu, Jun
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866911320724471808
author Zhou, Jiacong
Wang, Xianyun
Zhang, Min
Yu, Jun
author_facet Zhou, Jiacong
Wang, Xianyun
Zhang, Min
Yu, Jun
contents Aligning large language models with human preferences is essential for improving interaction quality and safety by ensuring outputs better reflect human values. A promising strategy involves Reinforcement Learning from Human Feedback (RLHF), starting with collecting and ranking responses generated by a supervised fine-tuning model to refine alignment. Existing methods such as Direct Preference Optimization (DPO) focus on pairwise comparisons, categorizing responses into preferred and less preferred pairs and optimizing pairwise margins. However, this pairwise approach cannot capture the holistic ranking relationships among multiple responses or effectively leverage the rich preference information available in list-wise comparisons. To address this challenge, this paper introduces \underline{D}irect \underline{R}anking \underline{P}reference \underline{O}ptimization (DRPO), a novel method that views human preference alignment as a Learning-to-Rank (LTR) task. Unlike pairwise methods, DRPO optimizes the preference ranking of entire response lists by computing holistic utility scores through NDCG, a standard LTR metric. To enable end-to-end optimization with the non-differentiable NDCG, we propose diffNDCG loss, a differentiable approximation facilitated by a sorting network. Furthermore, we introduce a novel margin-based Adaptive Rank Policy Score to enhance the discriminative quality of generated responses. Extensive experiments have shown that DRPO outperforms existing methods, enhancing the quality of the generated responses.
format Preprint
id arxiv_https___arxiv_org_abs_2410_18127
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Holistic Utility Preference Learning for Listwise Alignment
Zhou, Jiacong
Wang, Xianyun
Zhang, Min
Yu, Jun
Information Retrieval
Artificial Intelligence
Computation and Language
Machine Learning
Aligning large language models with human preferences is essential for improving interaction quality and safety by ensuring outputs better reflect human values. A promising strategy involves Reinforcement Learning from Human Feedback (RLHF), starting with collecting and ranking responses generated by a supervised fine-tuning model to refine alignment. Existing methods such as Direct Preference Optimization (DPO) focus on pairwise comparisons, categorizing responses into preferred and less preferred pairs and optimizing pairwise margins. However, this pairwise approach cannot capture the holistic ranking relationships among multiple responses or effectively leverage the rich preference information available in list-wise comparisons. To address this challenge, this paper introduces \underline{D}irect \underline{R}anking \underline{P}reference \underline{O}ptimization (DRPO), a novel method that views human preference alignment as a Learning-to-Rank (LTR) task. Unlike pairwise methods, DRPO optimizes the preference ranking of entire response lists by computing holistic utility scores through NDCG, a standard LTR metric. To enable end-to-end optimization with the non-differentiable NDCG, we propose diffNDCG loss, a differentiable approximation facilitated by a sorting network. Furthermore, we introduce a novel margin-based Adaptive Rank Policy Score to enhance the discriminative quality of generated responses. Extensive experiments have shown that DRPO outperforms existing methods, enhancing the quality of the generated responses.
title Holistic Utility Preference Learning for Listwise Alignment
topic Information Retrieval
Artificial Intelligence
Computation and Language
Machine Learning
url https://arxiv.org/abs/2410.18127