English
Related papers

Related papers: Failure Modes of Maximum Entropy RLHF

200 papers

Offline preference optimization methods, such as Direct Preference Optimization (DPO), offer significant advantages in aligning Large Language Models (LLMs) with human values. However, achieving optimal performance with these methods…

Machine Learning · Computer Science 2026-04-29 Peng Liao , Peijia Zheng , Lingbo Li , Shangsong Liang , Lin Chen

Direct Preference Optimization (DPO) is a widely adopted offline algorithm for preference-based reinforcement learning from human feedback (RLHF), designed to improve training simplicity and stability by redefining reward functions.…

Computation and Language · Computer Science 2025-05-30 Gengxu Li , Tingyu Xia , Yi Chang , Yuan Wu

The Maximum Entropy Reinforcement Learning (MaxEnt RL) framework is a leading approach for achieving efficient learning and robust performance across many RL tasks. However, MaxEnt methods have also been shown to struggle with…

Machine Learning · Computer Science 2025-06-13 Ruipeng Zhang , Ya-Chien Chang , Sicun Gao

Direct Preference Optimization (DPO) simplifies reinforcement learning from human feedback (RLHF) for large language models (LLMs) by directly optimizing human preferences without an explicit reward model. We find that during DPO training,…

Computation and Language · Computer Science 2026-01-01 Junshu Pan , Wei Shen , Shulin Huang , Qiji Zhou , Yue Zhang

Direct Preference Optimization (DPO) has been successfully used to align large language models (LLMs) according to human preferences, and more recently it has also been applied to improving the quality of text-to-image diffusion models.…

Computer Vision and Pattern Recognition · Computer Science 2025-10-07 Shivanshu Shekhar , Shreyas Singh , Tong Zhang

Aligning large language models (LLMs) with human values and intentions is crucial for their utility, honesty, and safety. Reinforcement learning from human feedback (RLHF) is a popular approach to achieve this alignment, but it faces…

Machine Learning · Computer Science 2025-07-22 Junkang Wu , Xue Wang , Zhengyi Yang , Jiancan Wu , Jinyang Gao , Bolin Ding , Xiang Wang , Xiangnan He

Entropy Regularisation is a widely adopted technique that enhances policy optimisation performance and stability. A notable form of entropy regularisation is augmenting the objective with an entropy term, thereby simultaneously optimising…

Machine Learning · Computer Science 2024-07-26 Jean Seong Bjorn Choe , Jong-Kook Kim

Reinforcement Learning (RL) from Human Preference-based feedback is a popular paradigm for fine-tuning generative models, which has produced impressive models such as GPT-4 and Claude3 Opus. This framework often consists of two steps:…

Machine Learning · Computer Science 2024-04-17 Jonathan D. Chang , Wenhao Zhan , Owen Oertell , Kianté Brantley , Dipendra Misra , Jason D. Lee , Wen Sun

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 this problem in a…

Machine Learning · Computer Science 2024-12-05 Zhihan Liu , Miao Lu , Shenao Zhang , Boyi Liu , Hongyi Guo , Yingxiang Yang , Jose Blanchet , Zhaoran Wang

This work studies the problem of large language model (LLM) unlearning, aiming to remove unwanted data influences (e.g., copyrighted or harmful content) while preserving model utility. Despite the increasing demand for unlearning, a…

Computation and Language · Computer Science 2025-10-21 Chongyu Fan , Jiancheng Liu , Licong Lin , Jinghan Jia , Ruiqi Zhang , Song Mei , Sijia Liu

We consider the problem of learning the optimal policy for Markov decision processes with safety constraints. We formulate the problem in a reach-avoid setup. Our goal is to design online reinforcement learning algorithms that ensure safety…

Machine Learning · Computer Science 2026-01-21 Abhijit Mazumdar , Rafal Wisniewski , Manuela L. Bujorianu

Improving the multi-step reasoning ability of large language models (LLMs) with offline reinforcement learning (RL) is essential for quickly adapting them to complex tasks. While Direct Preference Optimization (DPO) has shown promise in…

Machine Learning · Computer Science 2024-12-30 Huaijie Wang , Shibo Hao , Hanze Dong , Shenao Zhang , Yilin Bao , Ziran Yang , Yi Wu

Sample efficiency is critical for online Reinforcement Learning from Human Feedback (RLHF). While existing works investigate sample-efficient online exploration strategies, the potential of utilizing misspecified yet relevant reward models…

Machine Learning · Computer Science 2025-05-20 Jiawei Huang , Bingcong Li , Christoph Dann , Niao He

We study the problem of training and fine-tuning expressive policies with online reinforcement learning (RL) given an offline dataset. Training expressive policy classes with online RL present a unique challenge of stable value…

Machine Learning · Computer Science 2026-05-01 Perry Dong , Qiyang Li , Dorsa Sadigh , Chelsea Finn

Reinforcement learning with verifiable rewards (RLVR) has become an effective paradigm for improving the reasoning ability of large language models. However, widely used RLVR algorithms, such as GRPO, often suffer from entropy collapse,…

Machine Learning · Computer Science 2026-05-13 Huimin Xu , Shuai Zhao , Xiaobao Wu , Anh Tuan Luu

Recent advancements have established Reinforcement Learning (RL) as a pivotal paradigm for aligning generative models with human intent. However, group-based optimization frameworks (e.g., GRPO) face a critical limitation: the rapid decay…

Computer Vision and Pattern Recognition · Computer Science 2026-05-18 Sujie Hu , Chubin Chen , Jiashu Zhu , Jiahong Wu , Xiangxiang Chu , Xiu Li

The rapidly increasing capabilities of large language models (LLMs) raise an urgent need to align AI systems with diverse human preferences to simultaneously enhance their usefulness and safety, despite the often conflicting nature of these…

Machine Learning · Computer Science 2024-03-06 Zixuan Liu , Xiaolin Sun , Zizhan Zheng

This paper studies the alignment process of generative models with Reinforcement Learning from Human Feedback (RLHF). We first identify the primary challenges of existing popular methods like offline PPO and offline DPO as lacking in…

Machine Learning · Computer Science 2024-05-02 Wei Xiong , Hanze Dong , Chenlu Ye , Ziqi Wang , Han Zhong , Heng Ji , Nan Jiang , Tong Zhang

Adapting large language models (LLMs) for specific tasks usually involves fine-tuning through reinforcement learning with human feedback (RLHF) on preference data. While these data often come from diverse labelers' groups (e.g., different…

Computation and Language · Computer Science 2024-05-31 Shyam Sundhar Ramesh , Yifan Hu , Iason Chaimalas , Viraj Mehta , Pier Giuseppe Sessa , Haitham Bou Ammar , Ilija Bogunovic

Model-free reinforcement learning algorithms have seen remarkable progress, but key challenges remain. Trust Region Policy Optimization (TRPO) is known for ensuring monotonic policy improvement through conservative updates within a trust…

Machine Learning · Computer Science 2025-07-29 Zhengpeng Xie , Qiang Zhang , Fan Yang , Marco Hutter , Renjing Xu