RS-DPO: A Hybrid Rejection Sampling and Direct Preference Optimization Method for Alignment of Large Language Models

Saeed Khaki, JinJin Li, Lan Ma, Liu Yang, Prathap Ramachandra


Abstract
Reinforcement learning from human feedback (RLHF) has been extensively employed to align large language models with user intent. However, proximal policy optimization (PPO) based RLHF is occasionally unstable requiring significant hyperparameter finetuning, and computationally expensive to maximize the estimated reward during alignment. Recently, direct preference optimization (DPO) is proposed to address those challenges. However, DPO often relies on contrastive responses generated from human annotator and alternative LLM, instead of the policy model, limiting the effectiveness of the RLHF. In this paper, we addresses both challenges by systematically combining rejection sampling (RS) and DPO. Our proposed method, RS-DPO, initiates with the development of a supervised fine-tuned policy model (SFT). A varied set of k responses per prompt are sampled directly from the SFT model. RS-DPO identifies pairs of contrastive samples based on their reward distribution. Finally, we apply DPO with the contrastive samples to align the model to human preference. Our experiments indicate that our proposed method effectively fine-tunes LLMs with limited resource environments, leading to improved alignment with user intent. Furthermore, it outperforms existing methods, including RS, PPO, and DPO.
Anthology ID:
2024.findings-naacl.108
Volume:
Findings of the Association for Computational Linguistics: NAACL 2024
Month:
June
Year:
2024
Address:
Mexico City, Mexico
Editors:
Kevin Duh, Helena Gomez, Steven Bethard
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
1665–1680
Language:
URL:
https://aclanthology.org/2024.findings-naacl.108
DOI:
10.18653/v1/2024.findings-naacl.108
Bibkey:
Cite (ACL):
Saeed Khaki, JinJin Li, Lan Ma, Liu Yang, and Prathap Ramachandra. 2024. RS-DPO: A Hybrid Rejection Sampling and Direct Preference Optimization Method for Alignment of Large Language Models. In Findings of the Association for Computational Linguistics: NAACL 2024, pages 1665–1680, Mexico City, Mexico. Association for Computational Linguistics.
Cite (Informal):
RS-DPO: A Hybrid Rejection Sampling and Direct Preference Optimization Method for Alignment of Large Language Models (Khaki et al., Findings 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.findings-naacl.108.pdf