paper-with-me

Papers

Preference as Reward, Maximum Preference Optimization with Importance Sampling

2023-12-27 · Zaifan Jiang, Xing Huang, Chao Wei

Preference learning is a key technology for aligning language models with human values. Reinforcement Learning from Human Feedback (RLHF) is a model-based algorithm to optimize preference learning, which first fits a reward model for preference scores and then optimizes the generating policy with an on-policy PPO algorithm to maximize the reward. The processing of RLHF is complex, time-consuming, and unstable. The Direct Preference Optimization (DPO) algorithm uses an off-policy algorithm to directly optimize the generating policy and eliminates the need for a reward model. DPO is more data-efficient and stable. However, DPO has a drawback of overfitting to the preference data and ignoring the KL-regularization term when the preference is deterministic. Identity mapping Preference Optimization(IPO) uses a root-finding MSE loss to incorporate KL-regularization. However, both DPO and IPO fail to properly address the KL-regularization term because the support of the preference distribution is not equal to the reference distribution. In this paper, we propose a simple and intuitive off-policy preference optimization algorithm from an importance sampling view, which we call Maximum Preference Optimization (MPO). MPO incorporates the off-policy KL-regularization term, making regularization truly effective. MPO achieves the best of both worlds by combining the objectives of RLHF and IPO while being an off-policy algorithm. Furthermore, MPO eliminates the need for a reward model and reference policy, simplifying the learning process and reducing memory usage.

📄 PDF Abstract BibTeX arXiv:2312.16430

Code (0)

등록된 구현이 없습니다.

Methods 이 논문이 사용한 방법론

DPO 설명 없음
Entropy Regularization 설명 없음
PPO Proximal Policy Optimization, or PPO, is a policy gradient method for reinforcement learning. The motivation was to have an algorithm with the data efficiency and reliable…

Similar Papers 제목 키워드 기반

MaPPO: Maximum a Posteriori Preference Optimization with Prior Knowledge

2025-07-27 · Guangchen Lan, Sipeng Zhang, Tianle Wang, Yuwei Zhang 외 arxiv

As the era of large language models (LLMs) unfolds, Preference Optimization (PO) methods have become a central approach to aligning LLMs with human preferences and improving performance. We propose Maximum a Posteriori P…

Computational EfficiencyBinary Classification

Offline Preference Optimization via Maximum Marginal Likelihood Estimation

2025-10-27 · Saeed Najafi, Alona Fyshe arxiv

Aligning Large Language Models (LLMs) with human preferences is crucial, but standard methods like Reinforcement Learning from Human Feedback (RLHF) are often complex and unstable. In this work, we propose a new, simpler…

Reinforcement Learning

Robust Reinforcement Learning from Corrupted Human Feedback

2024-06-21 · Alexander Bukharin, Ilgee Hong, Haoming Jiang, Zichong Li 외

Reinforcement learning from human feedback (RLHF) provides a principled framework for aligning AI systems with human preference data. For various reasons, e.g., personal bias, context ambiguity, lack of training, etc, hu…

reinforcement-learningReinforcement LearningText Generation

Provably Mitigating Overoptimization in RLHF: Your SFT Loss is Implicitly an Adversarial Regularizer

2024-05-26 · Zhihan Liu, Miao Lu, Shenao Zhang, Boyi Liu 외

Aligning generative models with human preference via RLHF typically suffers from overoptimization, where an imperfectly learned reward model can misguide the generative model to output undesired responses. We investigate…

Failure Modes of Maximum Entropy RLHF

2025-09-24 · Ömer Veysel Çağatan, Barış Akgün arxiv

In this paper, we show that Simple Preference Optimization (SimPO) can be derived as Maximum Entropy Reinforcement Learning, providing a theoretical foundation for this reference-free method. Motivated by SimPO's strong …

Reinforcement Learning