PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
February 15, 20240 citationsOpen Access

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

View Full Paper
SKSaeed KhakiJLJinjin LiLMLan Ma

Key Points

  • Improved alignment with user intent seen in LLMs through new methods, enhancing interaction reliability with AI.
  • Results show RS-DPO surpasses traditional models like proximal policy optimization and previous direct methods by notable gains.
  • Assessment involves developing supervised fine-tuned policy models with multiple responses generated per prompt for contrastive evaluation and training adjustments to meet user needs effectively in various settings. Enhancements validate the function of rejection sampling in support of preference alignment to user feedback.

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 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.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Khaki et al. (2024) studied this question.

synapsesocial.com/papers/68e7917cb6db643587702dachttps://doi.org/10.48550/arxiv.2402.10038
Ask AI
Helpful
Bookmark
Share
View Full Paper