English
Related papers

Related papers: The Perils of Optimizing Learned Reward Functions:…

200 papers

In most machine learning training paradigms a fixed, often handcrafted, loss function is assumed to be a good proxy for an underlying evaluation metric. In this work we assess this assumption by meta-learning an adaptive loss function to…

Machine Learning · Computer Science 2019-05-16 Chen Huang , Shuangfei Zhai , Walter Talbott , Miguel Angel Bautista , Shih-Yu Sun , Carlos Guestrin , Josh Susskind

This work proposes `PET', a novel pessimistic reward fine-tuning method, to learn a pessimistic reward model robust against reward hacking in offline reinforcement learning from human feedback (RLHF). Traditional reward modeling techniques…

Machine Learning · Computer Science 2025-05-28 Yinglun Xu , Hangoo Kang , Tarun Suresh , Yuxuan Wan , Gagandeep Singh

Reinforcement learning from human feedback (RLHF) has emerged as a powerful technique to make large language models (LLMs) easier to use and more effective. A core piece of the RLHF process is the training and utilization of a model of…

Computers and Society · Computer Science 2023-11-29 Nathan Lambert , Thomas Krendl Gilbert , Tom Zick

Reinforcement learning with human feedback (RLHF), which learns a reward model from human preference data and then optimizes a policy to favor preferred responses, has emerged as a central paradigm for aligning large language models (LLMs)…

Machine Learning · Statistics 2025-09-29 Gen Li , Yuling Yan

Deep reinforcement learning can learn effective policies for a wide range of tasks, but is notoriously difficult to use due to instability and sensitivity to hyperparameters. The reasons for this remain unclear. When using standard…

Machine Learning · Computer Science 2020-03-17 Aviral Kumar , Abhishek Gupta , Sergey Levine

Large language models are typically aligned with human preferences by optimizing $\textit{reward models}$ (RMs) fitted to human feedback. However, human preferences are multi-faceted, and it is increasingly common to derive reward from a…

Reinforcement Learning from Human Feedback significantly enhances Natural Language Processing by aligning language models with human expectations. A critical factor in this alignment is the strength of reward models used during training.…

Computation and Language · Computer Science 2024-10-17 Yanjun Chen , Dawei Zhu , Yirong Sun , Xinghao Chen , Wei Zhang , Xiaoyu Shen

There are many algorithms for regret minimisation in episodic reinforcement learning. This problem is well-understood from a theoretical perspective, providing that the sequences of states, actions and rewards associated with each episode…

Machine Learning · Computer Science 2023-04-07 Benjamin Howson , Ciara Pike-Burke , Sarah Filippi

Reinforcement learning (RL) commonly relies on scalar rewards with limited ability to express temporal, conditional, or safety-critical goals, and can lead to reward hacking. Temporal logic expressible via the more general class of…

Artificial Intelligence · Computer Science 2025-11-26 Dominik Wagner , Leon Witzman , Luke Ong

Reward hacking -- where RL agents exploit gaps in misspecified reward functions -- has been widely observed, but not yet systematically studied. To understand how reward hacking arises, we construct four RL environments with misspecified…

Machine Learning · Computer Science 2022-02-15 Alexander Pan , Kush Bhatia , Jacob Steinhardt

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

Reinforcement learning from human feedback (RLHF) has emerged as a key technique for aligning the output of large language models (LLMs) with human preferences. To learn the reward function, most existing RLHF algorithms use the…

Machine Learning · Statistics 2026-02-11 Kai Ye , Hongyi Zhou , Jin Zhu , Francesco Quinzan , Chengchun Shi

Reinforcement Learning from Human Feedback (RLHF) allows us to train models, such as language models (LMs), to follow complex human preferences. In RLHF for LMs, we first train an LM using supervised fine-tuning, sample pairs of responses,…

Machine Learning · Computer Science 2025-07-22 Johannes Ackermann , Takashi Ishida , Masashi Sugiyama

We study reinforcement learning (RL) problems in which agents observe the reward or transition realizations at their current state before deciding which action to take. Such observations are available in many applications, including…

Machine Learning · Computer Science 2024-10-22 Nadav Merlis

Because it is difficult to precisely specify complex objectives, reinforcement learning policies are often optimized using proxy reward functions that only approximate the true goal. However, optimizing proxy rewards frequently leads to…

Machine Learning · Computer Science 2025-03-14 Cassidy Laidlaw , Shivam Singhal , Anca Dragan

Diffusion models excel at modeling complex data distributions, including those of images, proteins, and small molecules. However, in many cases, our goal is to model parts of the distribution that maximize certain properties: for example,…

Model-based Reinforcement Learning (MBRL) aims to make agents more sample-efficient, adaptive, and explainable by learning an explicit model of the environment. While the capabilities of MBRL agents have significantly improved in recent…

Machine Learning · Computer Science 2024-04-09 Ran Wei , Nathan Lambert , Anthony McDonald , Alfredo Garcia , Roberto Calandra

The success of Reinforcement Learning from Human Feedback (RLHF) critically depends on the quality of the reward model. However, while this quality is primarily evaluated through accuracy, it remains unclear whether accuracy fully captures…

Machine Learning · Computer Science 2026-03-02 Noam Razin , Zixuan Wang , Hubert Strauss , Stanley Wei , Jason D. Lee , Sanjeev Arora

When applying reinforcement learning from human feedback (RLHF), the reward is learned from data and, therefore, always has some error. It is common to mitigate this by regularizing the policy with KL divergence from a base model, with the…

Machine Learning · Computer Science 2024-11-11 Thomas Kwa , Drake Thomas , Adrià Garriga-Alonso

A safe and trustworthy use of Large Language Models (LLMs) requires an accurate expression of confidence in their answers. We propose a novel Reinforcement Learning approach that allows to directly fine-tune LLMs to express calibrated…

Computation and Language · Computer Science 2026-03-03 David Bani-Harouni , Chantal Pellegrini , Paul Stangel , Ege Özsoy , Kamilia Zaripova , Nassir Navab , Matthias Keicher