中文
相关论文

相关论文: Off-Policy Self-Critical Training for Transformer …

200 篇论文

Offline goal-conditioned reinforcement learning (GCRL) aims at solving goal-reaching tasks with sparse rewards from an offline dataset. While prior work has demonstrated various approaches for agents to learn near-optimal policies, these…

机器人学 · 计算机科学 2024-03-05 Chenyang Cao , Zichen Yan , Renhao Lu , Junbo Tan , Xueqian Wang

Adversarial Imitation Learning (AIL) is a class of algorithms in Reinforcement learning (RL), which tries to imitate an expert without taking any reward from the environment and does not provide expert behavior directly to the policy…

机器学习 · 计算机科学 2020-05-05 Samin Yeasar Arnob

Recent advancements in off-policy Reinforcement Learning (RL) have significantly improved sample efficiency, primarily due to the incorporation of various forms of regularization that enable more gradient update steps than traditional…

The recent success of supervised learning methods on ever larger offline datasets has spurred interest in the reinforcement learning (RL) field to investigate whether the same paradigms can be translated to RL algorithms. This research…

机器学习 · 计算机科学 2021-02-12 Mengjiao Yang , Ofir Nachum

We investigate the ability of transformers to perform in-context reinforcement learning (ICRL), where a model must infer and execute learning algorithms from trajectory data without parameter updates. We show that a linear self-attention…

机器学习 · 统计学 2026-05-08 Haodong Liang , Lifeng Lai

Reinforcement learning (RL) has demonstrated significant potential for post-training language models and autoregressive visual generative models, but adapting RL to masked generative models remains challenging. The core factor is that…

计算机视觉与模式识别 · 计算机科学 2025-12-23 Guohui Zhang , Hu Yu , Xiaoxiao Ma , Yaning Pan , Hang Xu , Feng Zhao

To meet the requirements of real-world applications, it is essential to control generations of large language models (LLMs). Prior research has tried to introduce reinforcement learning (RL) into controllable text generation while most…

计算与语言 · 计算机科学 2024-03-19 Wendi Li , Wei Wei , Kaihe Xu , Wenfeng Xie , Dangyang Chen , Yu Cheng

Effectively leveraging large, previously collected datasets in reinforcement learning (RL) is a key challenge for large-scale real-world applications. Offline RL algorithms promise to learn effective policies from previously-collected,…

机器学习 · 计算机科学 2020-08-20 Aviral Kumar , Aurick Zhou , George Tucker , Sergey Levine

Reinforcement learning (RL) post-training has increasingly demonstrated strong ability to elicit reasoning behaviors in large language models (LLMs). For training efficiency, rollouts are typically generated in an off-policy manner using an…

人工智能 · 计算机科学 2026-02-02 Shiye Lei , Zhihao Cheng , Dacheng Tao

Off-policy updates are inevitable in reinforcement learning (RL) for large language models (LLMs) due to rollout staleness from asynchronous training and mismatches between training and inference engines. Naive importance sampling gives an…

机器学习 · 计算机科学 2026-05-11 Guobin Shen , Chenxiao Zhao , Xiang Cheng , Lei Huang , Xing Yu

Recent advancements in deep reinforcement learning (RL) have demonstrated notable progress in sample efficiency, spanning both model-based and model-free paradigms. Despite the identification and mitigation of specific bottlenecks in prior…

机器学习 · 计算机科学 2024-04-02 Yibo Wang , Jiang Zhao

Reinforcement Learning (RL) algorithms are known to scale poorly to environments with many available actions, requiring numerous samples to learn an optimal policy. The traditional approach of considering the same fixed action space in…

机器学习 · 计算机科学 2023-05-15 Leo Ardon , Alberto Pozanco , Daniel Borrajo , Sumitra Ganesh

Offline reinforcement learning (RL) is a challenging task, whose objective is to learn policies from static trajectory data without interacting with the environment. Recently, offline RL has been viewed as a sequence modeling problem, where…

机器学习 · 计算机科学 2023-03-08 Shengchao Hu , Li Shen , Ya Zhang , Dacheng Tao

Online interactions with the environment to collect data samples for training a Reinforcement Learning (RL) agent is not always feasible due to economic and safety concerns. The goal of Offline Reinforcement Learning is to address this…

机器学习 · 计算机科学 2021-10-05 Chi Zhang , Sanmukh Rao Kuppannagari , Viktor K Prasanna

Reinforcement learning is a promising framework for solving control problems, but its use in practical situations is hampered by the fact that reward functions are often difficult to engineer. Specifying goals and tasks for autonomous…

机器学习 · 计算机科学 2019-02-22 Justin Fu , Anoop Korattikara , Sergey Levine , Sergio Guadarrama

Off-policy learning methods seek to derive an optimal policy directly from a fixed dataset of prior interactions. This objective presents significant challenges, primarily due to the inherent distributional shift and value function…

机器学习 · 计算机科学 2026-02-03 Arip Asadulaev , Maksim Bobrin , Salem Lahlou , Dmitry Dylov , Fakhri Karray , Martin Takac

Reinforcement learning (RL) has achieved remarkable success in fields like robotics and autonomous driving, but adversarial attacks designed to mislead RL systems remain challenging. Existing approaches often rely on modifying the…

机器学习 · 计算机科学 2025-07-25 Junyong Jiang , Buwei Tian , Chenxing Xu , Songze Li , Lu Dong

A highly desirable property of a reinforcement learning (RL) agent -- and a major difficulty for deep RL approaches -- is the ability to generalize policies learned on a few tasks over a high-dimensional observation space to similar tasks…

机器学习 · 计算机科学 2022-03-17 Bogdan Mazoure , Ahmed M. Ahmed , Patrick MacAlpine , R Devon Hjelm , Andrey Kolobov

Offline reinforcement learning (RL) methods harness previous experiences to derive an optimal policy, forming the foundation for pre-trained large-scale models (PLMs). When encountering tasks not seen before, PLMs often utilize several…

机器学习 · 计算机科学 2024-11-05 Shengchao Hu , Wanru Zhao , Weixiong Lin , Li Shen , Ya Zhang , Dacheng Tao

Supervised fine-tuning (SFT) of large language models can be viewed as an off-policy learning problem, where expert demonstrations come from a fixed behavior policy while training aims to optimize a target policy. Importance sampling is the…

机器学习 · 计算机科学 2025-09-22 Shiwan Zhao , Xuyang Zhao , Jiaming Zhou , Aobo Kong , Qicheng Li , Yong Qin