Task-Relevant Representation Decoupling for Visual Reinforcement Learning Generalization

2026-07-01 · Jinwen Wang, Youfang Lin, Xiaobo Hu, Qian Xu, Shuo Wang, Zhuo Chen, Kai Lv arxiv

Visual Reinforcement Learning (VRL) has achieved considerable success in solving control tasks. However, generalizing learned policies to new environments remains a major challenge, as agents often overfit to task-irrelevant features in the training environment. To solve this problem, we introduce the concept of decoupling observations into task-relevant and task-irrelevant representations. Building on this idea, we propose a self-supervised Task-Relevant Representation Decoupling (T2RD) algorithm for VRL. This algorithm consists of three components: task-relevant representation consistency, cross-reconstruction, and cross-dynamic prediction. The first two components achieve the decoupling of content and style features, but the resulting content representations are not necessarily task-relevant. To further refine task-relevant features from content representations, we design the third component that introduces dynamic prediction. T2RD achieves State-Of-The-Art (SOTA) generalization performance and sample efficiency in the DeepMind Control Suite and Robotic Manipulation tasks.

📄 PDF Abstract BibTeX arXiv:2607.00796

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement Learning

Masked World Models for Visual Control

2022-06-28 · Younggyo Seo, Danijar Hafner, Hao liu, Fangchen Liu 외

Visual model-based reinforcement learning (RL) has the potential to enable sample-efficient robot learning from visual observations. Yet the current approaches typically train a single model end-to-end for learning both …

Model-based Reinforcement Learning Reinforcement Learning (RL)Representation Learning

SeMAIL: Eliminating Distractors in Visual Imitation via Separated Models

2023-06-19 · Shenghua Wan, Yucen Wang, Minghao Shao, Ruying Chen 외

Model-based imitation learning (MBIL) is a popular reinforcement learning method that improves sample efficiency on high-dimension input sources, such as images and videos. Following the convention of MBIL research, exis…

Imitation Learning

Robust Scene Transfer for PointGoal Navigation via Privileged Sensor Guided Contrastive Learning

2026-06-03 · Amirhossein Zhalehmehrabi, Tiziano Tezze, Alberto Castelini, Alessandro Farinelli arxiv

We propose a sensor-guided adaptive contrastive learning framework for visual representation learning in PointGoal navigation. During training, privileged LiDAR sensing guides the contrastive objective through a geometry…

Representation Learning Reinforcement Learning Contrastive Learning PointGoal Navigation

Decoupling feature extraction from policy learning: assessing benefits of state representation learning in goal based robotics

2019-01-24 · Antonin Raffin, Ashley Hill, René Traoré, Timothée Lesort 외

Scaling end-to-end reinforcement learning to control real robots from vision presents a series of challenges, in particular in terms of sample efficiency. Against end-to-end learning, state representation learning can he…

reinforcement-learning Reinforcement Learning Reinforcement Learning (RL)Representation Learning

Dual Feature Decoupling for Fine-Grained OOD Detection

2026-06-04 · Xiaokun Li, Yaping Huang, Qingji Guan arxiv

Out-of-distribution detection (OOD) is an indispensable technique when applying machine learning models to real-world scenarios. Most existing OOD detection methods have been developed under the idealized assumption of l…

Out-of-Distribution Detection Medical Image Classification