daDPO: Distribution-Aware DPO for Distilling Conversational Abilities

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Zhengze, Wang, Shiqi, Shen, Yiqun, Guo, Simin, Lin, Dahua, Wang, Xiaoliang, Cam-Tu, Nguyen, Tan, Fei
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915350553034752
author Zhang, Zhengze
Wang, Shiqi
Shen, Yiqun
Guo, Simin
Lin, Dahua
Wang, Xiaoliang
Cam-Tu, Nguyen
Tan, Fei
author_facet Zhang, Zhengze
Wang, Shiqi
Shen, Yiqun
Guo, Simin
Lin, Dahua
Wang, Xiaoliang
Cam-Tu, Nguyen
Tan, Fei
contents Large language models (LLMs) have demonstrated exceptional performance across various applications, but their conversational abilities decline sharply as model size decreases, presenting a barrier to their deployment in resource-constrained environments. Knowledge distillation with Direct Preference Optimization (dDPO) has emerged as a promising approach to enhancing the conversational abilities of smaller models using a larger teacher model. However, current methods primarily focus on 'black-box' KD, which only uses the teacher's responses, overlooking the output distribution offered by the teacher. This paper addresses this gap by introducing daDPO (Distribution-Aware DPO), a unified method for preference optimization and distribution-based distillation. We provide rigorous theoretical analysis and empirical validation, showing that daDPO outperforms existing methods in restoring performance for pruned models and enhancing smaller LLM models. Notably, in in-domain evaluation, our method enables a 20% pruned Vicuna1.5-7B to achieve near-teacher performance (-7.3% preference rate compared to that of dDPO's -31%), and allows Qwen2.5-1.5B to occasionally outperform its 7B teacher model (14.0% win rate).
format Preprint
id arxiv_https___arxiv_org_abs_2506_15717
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle daDPO: Distribution-Aware DPO for Distilling Conversational Abilities
Zhang, Zhengze
Wang, Shiqi
Shen, Yiqun
Guo, Simin
Lin, Dahua
Wang, Xiaoliang
Cam-Tu, Nguyen
Tan, Fei
Machine Learning
Artificial Intelligence
Computation and Language
Large language models (LLMs) have demonstrated exceptional performance across various applications, but their conversational abilities decline sharply as model size decreases, presenting a barrier to their deployment in resource-constrained environments. Knowledge distillation with Direct Preference Optimization (dDPO) has emerged as a promising approach to enhancing the conversational abilities of smaller models using a larger teacher model. However, current methods primarily focus on 'black-box' KD, which only uses the teacher's responses, overlooking the output distribution offered by the teacher. This paper addresses this gap by introducing daDPO (Distribution-Aware DPO), a unified method for preference optimization and distribution-based distillation. We provide rigorous theoretical analysis and empirical validation, showing that daDPO outperforms existing methods in restoring performance for pruned models and enhancing smaller LLM models. Notably, in in-domain evaluation, our method enables a 20% pruned Vicuna1.5-7B to achieve near-teacher performance (-7.3% preference rate compared to that of dDPO's -31%), and allows Qwen2.5-1.5B to occasionally outperform its 7B teacher model (14.0% win rate).
title daDPO: Distribution-Aware DPO for Distilling Conversational Abilities
topic Machine Learning
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2506.15717