VPO: Leveraging the Number of Votes in Preference Optimization

TL;DR

This paper introduces VPO, a novel preference optimization framework that leverages vote counts in human preference data to improve language model training, outperforming existing methods.

Contribution

The paper proposes a new method, VPO, that uses vote counts with Bayesian MMSE estimation to better model preferences and enhance language model training.

Findings

VPO outperforms existing preference optimization algorithms.

Vote count information improves preference modeling accuracy.

Extensions of DPO and IPO with VPO yield better results.

Abstract

Direct Preference Optimization (DPO) trains a language model using human preference data, bypassing the explicit reward modeling phase of Reinforcement Learning from Human Feedback (RLHF). By iterating over sentence pairs in a preference dataset, DPO enhances generation quality by increasing the likelihood of producing preferred sentences over less favored ones. Preference datasets are typically created by selecting preferred sentences through a voting process involving multiple individuals, as opinions can vary due to the subjective nature of human preferences. While the number of votes offers insight into whether a sentence pair is clearly preferable or controversial, current methods do not fully leverage this information. In this paper, we introduce a technique that leverages user voting data to better align with diverse subjective preferences. We employ the Bayesian Minimum Mean Square Error (Bayesian MMSE) estimator to model the probability that one generation is preferable to another. Using this estimated probability as a target, we develop the Vote-based Preference Optimization (VPO) framework, which incorporates the number of votes on both sides to distinguish between controversial and obvious generation pairs. We show that previous algorithms, such as DPO and Identity Preference Optimization (IPO), can be extended using the proposed framework, termed VDPO and VIPO. Our experiments demonstrate that these proposed algorithms outperform various existing methods, including their base algorithms.