Weighted-Reward Preference Optimization for Implicit Model Fusion

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yang, Ziyi, Wan, Fanqi, Zhong, Longguang, Shi, Tianyuan, Quan, Xiaojun
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909511416020992
author Yang, Ziyi
Wan, Fanqi
Zhong, Longguang
Shi, Tianyuan
Quan, Xiaojun
author_facet Yang, Ziyi
Wan, Fanqi
Zhong, Longguang
Shi, Tianyuan
Quan, Xiaojun
contents While fusing heterogeneous open-source LLMs with varying architectures and sizes can potentially integrate the strengths of different models, existing fusion methods face significant challenges, such as vocabulary alignment and merging distribution matrices. These procedures are not only complex but also prone to introducing noise and errors. In this paper, we propose an implicit fusion method, Weighted-Reward Preference Optimization (WRPO), which leverages preference optimization between the source LLMs and the target LLM to transfer their capabilities effectively. WRPO eliminates the need for vocabulary alignment and matrix fusion and can be efficiently scaled to accommodate various LLMs. To address distributional deviations between the source and target LLMs, WRPO introduces a progressive adaptation strategy that gradually shifts reliance on preferred examples from the target LLM to the source LLMs. Extensive experiments on the MT-Bench, AlpacaEval-2, and Arena-Hard benchmarks demonstrate that WRPO consistently outperforms existing knowledge fusion methods and various fine-tuning baselines. When applied to LLaMA3-8B-Instruct as the target model, WRPO achieves a length-controlled win rate of 55.9% against GPT-4-Preview-1106 on AlpacaEval-2 and a win rate of 46.2% against GPT-4-0314 on Arena-Hard. Our code is available at https://github.com/SLIT-AI/WRPO.
format Preprint
id arxiv_https___arxiv_org_abs_2412_03187
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Weighted-Reward Preference Optimization for Implicit Model Fusion
Yang, Ziyi
Wan, Fanqi
Zhong, Longguang
Shi, Tianyuan
Quan, Xiaojun
Computation and Language
While fusing heterogeneous open-source LLMs with varying architectures and sizes can potentially integrate the strengths of different models, existing fusion methods face significant challenges, such as vocabulary alignment and merging distribution matrices. These procedures are not only complex but also prone to introducing noise and errors. In this paper, we propose an implicit fusion method, Weighted-Reward Preference Optimization (WRPO), which leverages preference optimization between the source LLMs and the target LLM to transfer their capabilities effectively. WRPO eliminates the need for vocabulary alignment and matrix fusion and can be efficiently scaled to accommodate various LLMs. To address distributional deviations between the source and target LLMs, WRPO introduces a progressive adaptation strategy that gradually shifts reliance on preferred examples from the target LLM to the source LLMs. Extensive experiments on the MT-Bench, AlpacaEval-2, and Arena-Hard benchmarks demonstrate that WRPO consistently outperforms existing knowledge fusion methods and various fine-tuning baselines. When applied to LLaMA3-8B-Instruct as the target model, WRPO achieves a length-controlled win rate of 55.9% against GPT-4-Preview-1106 on AlpacaEval-2 and a win rate of 46.2% against GPT-4-0314 on Arena-Hard. Our code is available at https://github.com/SLIT-AI/WRPO.
title Weighted-Reward Preference Optimization for Implicit Model Fusion
topic Computation and Language
url https://arxiv.org/abs/2412.03187