English
Related papers

Related papers: Provably Mitigating Corruption, Overoptimization, …

200 papers

In this paper, we explore how directly pretraining a value model simplifies and stabilizes reinforcement learning from human feedback (RLHF). In reinforcement learning, value estimation is the key to policy optimization, distinct from…

Machine Learning · Computer Science 2026-01-27 Chenghua Huang , Lu Wang , Fangkai Yang , Pu Zhao , Zhixu Li , Qingwei Lin , Dongmei Zhang , Saravan Rajmohan , Qi Zhang

Direct Preference Optimization (DPO) has emerged as a stable, scalable, and efficient solution for language model alignment. Despite its empirical success, the optimization properties, particularly the impact of samplers on its convergence…

Machine Learning · Computer Science 2025-02-04 Ruizhe Shi , Runlong Zhou , Simon S. Du

Despite significant advances in long-context reasoning by large language models (LLMs), primarily through Online Reinforcement Learning (RL) methods, these approaches incur substantial computational costs and complexity. In contrast,…

Computation and Language · Computer Science 2025-05-06 Xiaoyu Tian , Sitong Zhao , Haotian Wang , Shuaiting Chen , Yiping Peng , Yunjie Ji , Han Zhao , Xiangang Li

Direct Preference Optimization (DPO) has been proposed as an effective and efficient alternative to reinforcement learning from human feedback (RLHF). However, neither RLHF nor DPO take into account the fact that learning certain…

Computer Vision and Pattern Recognition · Computer Science 2026-02-16 Florinel-Alin Croitoru , Vlad Hondru , Radu Tudor Ionescu , Nicu Sebe , Mubarak Shah

Post-alignment of large language models (LLMs) is critical in improving their utility, safety, and alignment with human intentions. Direct preference optimisation (DPO) has become one of the most widely used algorithms for achieving this…

Machine Learning · Computer Science 2025-01-06 Rasul Tutnov , Antoine Grosnit , Haitham Bou-Ammar

Reinforcement learning from human feedback (RLHF) is the canonical framework for large language model alignment. However, rising popularity in offline alignment algorithms challenge the need for on-policy sampling in RLHF. Within the…

Paraphrasing re-expresses meaning to enhance applications like text simplification, machine translation, and question-answering. Specific paraphrase types facilitate accurate semantic analysis and robust language models. However, existing…

Computation and Language · Computer Science 2025-06-04 Christopher Lee Lübbers

Reinforcement Learning from Human Feedback (RLHF) is commonly utilized to improve the alignment of Large Language Models (LLMs) with human preferences. Given the evolving nature of human preferences, continual alignment becomes more crucial…

Machine Learning · Computer Science 2024-12-24 Han Zhang , Lin Gui , Yu Lei , Yuanzhao Zhai , Yehong Zhang , Yulan He , Hui Wang , Yue Yu , Kam-Fai Wong , Bin Liang , Ruifeng Xu

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…

Machine Learning · Computer Science 2024-10-31 Jae Hyeon Cho , Minkyung Park , Byung-Jun Lee

Reinforcement learning from human feedback (RLHF) has become the leading approach for fine-tuning large language models (LLM). However, RLHF has limitations in multi-task learning (MTL) due to challenges of reward hacking and extreme…

Reinforcement learning from human feedback (RLHF) is a standard approach for fine-tuning large language models to follow instructions. As part of this process, learned reward models are used to approximately model human preferences.…

Machine Learning · Computer Science 2024-03-12 Thomas Coste , Usman Anwar , Robert Kirk , David Krueger

Large language models (LLMs) trained on webscale data can produce toxic outputs, raising concerns for safe deployment. Prior defenses, based on applications of DPO, NPO, and similar algorithms, reduce the likelihood of harmful…

Language model (LM) post-training (or alignment) involves maximizing a reward function that is derived from preference annotations. Direct Preference Optimization (DPO) is a popular offline alignment method that trains a policy directly on…

Machine Learning · Computer Science 2025-03-04 Adam Fisch , Jacob Eisenstein , Vicky Zayats , Alekh Agarwal , Ahmad Beirami , Chirag Nagpal , Pete Shaw , Jonathan Berant

Preference learning algorithms (e.g., RLHF and DPO) are frequently used to steer LLMs to produce generations that are more preferred by humans, but our understanding of their inner workings is still limited. In this work, we study the…

Machine Learning · Computer Science 2024-11-01 Angelica Chen , Sadhika Malladi , Lily H. Zhang , Xinyi Chen , Qiuyi Zhang , Rajesh Ranganath , Kyunghyun Cho

Direct Alignment Algorithms (DAAs), such as Direct Preference Optimisation (DPO) and Identity Preference Optimisation (IPO), have emerged as alternatives to online Reinforcement Learning from Human Feedback (RLHF) algorithms such as…

Computation and Language · Computer Science 2024-10-21 Zhengyan Shi , Sander Land , Acyr Locatelli , Matthieu Geist , Max Bartolo

The diversity of contexts in which large language models (LLMs) are deployed requires the ability to modify or customize default model behaviors to incorporate nuanced requirements and preferences. A convenient interface to specify such…

Machine Learning · Computer Science 2024-02-19 Moritz Stephan , Alexander Khazatsky , Eric Mitchell , Annie S Chen , Sheryl Hsu , Archit Sharma , Chelsea Finn

The ability to train high-performing reward models with few-shot data is critical for enhancing the efficiency and scalability of Reinforcement Learning from Human Feedback (RLHF). We propose a data augmentation and expansion framework that…

Machine Learning · Computer Science 2025-06-11 Yiyang Zhao , Huiyu Bai , Xuejiao Zhao

Reinforcement Learning from Human Feedback (RLHF) has achieved considerable success in aligning large language models (LLMs) by modeling human preferences with a learnable reward model and employing a reinforcement learning algorithm to…

Machine Learning · Computer Science 2025-05-20 Jianfeng Cai , Jinhua Zhu , Ruopei Sun , Yue Wang , Li Li , Wengang Zhou , Houqiang Li

Direct Preference Optimization (DPO) is an effective framework for aligning large language models with human preferences, but it struggles with complex reasoning tasks. DPO optimizes for the likelihood of generating preferred over…

Artificial Intelligence · Computer Science 2026-04-23 Darsh Kachroo , Adriana Caraeni , Arjun Prasaath Anbazhagan , Brennan Lagasse , Kevin Zhu

Aligning Large Language Models (LLMs) traditionally relies on costly training and human preference annotations. Self-alignment seeks to reduce these expenses by enabling models to align themselves. To further lower costs and achieve…

Computation and Language · Computer Science 2024-11-15 Somanshu Singla , Zhen Wang , Tianyang Liu , Abdullah Ashfaq , Zhiting Hu , Eric P. Xing