mlsanityData to 14 Oct 2026

Task-Relevant Representation Decoupling for Visual Reinforcement Learning Generalization

arXiv 2607.00796 ↗First posted 1 Jul 2026cs.LGcited by 0

On the map: Deep reinforcement learning methods → RL for robot locomotion and manipulation

Abstract (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.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in RL for robot locomotion and manipulation, one of 15 directions in Deep reinforcement learning methods. Builds on 39 of its 54 references that are in the map. First posted on arXiv 1 Jul 2026.

Builds on

39 of 54 references.

Closest ancestors · 1

Same direction, newest first

  1. 2024ViSaRL: Visual Reinforcement Learning Guided by Human Saliency16 Mar

Other references · 38

  1. 2018Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actorcited by 4,885
  2. 2020Bootstrap your own latent: A new approach to self-supervised Learningcited by 4,750
  3. 2018DeepMind Control Suitecited by 956
  4. 2015End-to-End Training of Deep Visuomotor Policiescited by 875
  5. 2020CURL: Contrastive Unsupervised Representations for Reinforcement Learningcited by 760
  6. 2023VoxPoser: Composable 3D Value Maps for Robotic Manipulation with Language Modelscited by 750
  7. 2020Image Augmentation Is All You Need: Regularizing Deep Reinforcement Learning from Pixelscited by 585
  8. 2020Reinforcement Learning with Augmented Datacited by 468
  9. 2020Learning Invariant Representations for Reinforcement Learning without Reconstructioncited by 436
  10. 2020Robust and Generalizable Visual Representation Learning via Random Convolutionscited by 142
  11. 2020Generalization in Reinforcement Learning by Soft Data Augmentationcited by 137
  12. 2023Q-Transformer: Scalable Offline Reinforcement Learning via Autoregressive Q-Functionscited by 131
26 more
  1. 2020Self-Supervised Policy Adaptation during Deploymentcited by 123
  2. 2021Stabilizing Deep Q-Learning with ConvNets and Vision Transformers under Data Augmentationcited by 105
  3. 2016Disentangling factors of variation in deep representations using adversarial trainingcited by 87
  4. 2017Unsupervised Person Re-identification: Clustering and Fine-tuningcited by 73
  5. 2019Unsupervised Part-Based Disentangling of Object Shape and Appearancecited by 67
  6. 2020Dreaming: Model-based Reinforcement Learning by Latent Imagination without Reconstructioncited by 66
  7. 2018Image-to-image translation for cross-domain disentanglementcited by 61
  8. 2022Pre-Trained Image Encoder for Generalizable Visual Reinforcement Learningcited by 60
  9. 2023TACO: Temporal Latent Action-Driven Contrastive Loss for Visual Reinforcement Learningcited by 57
  10. 2019Learning Disentangled Representations via Mutual Information Estimationcited by 52
  11. 2022Look Closer: Bridging Egocentric and Third-Person Views with Transformers for Robotic Manipulationcited by 49
  12. 2021SECANT: Self-Expert Cloning for Zero-Shot Generalization of Visual Policiescited by 43
  13. 2018End-to-end Active Object Tracking and Its Real-world Deployment via Reinforcement Learningcited by 35
  14. 2022Look where you look! Saliency-guided Q-networks for generalization in visual Reinforcement Learningcited by 34
  15. 2021Unsupervised Visual Attention and Invariance for Reinforcement Learningcited by 30
  16. 2021Towards Distraction-Robust Active Visual Trackingcited by 30
  17. 2024Empowering Embodied Visual Tracking with Visual Foundation Models and Offline RLcited by 23
  18. 2022Don't Touch What Matters: Task-Aware Lipschitz Data Augmentation for Visual Reinforcement Learningcited by 21
  19. 2018CRAVES: Controlling Robotic Arm with a Vision-based Economic Systemcited by 18
  20. 2023Learning from Visual Observation via Offline Pretrained State-to-Go Transformercited by 14
  21. 2023Normalization Enhances Generalization in Visual Reinforcement Learningcited by 13
  22. 2023Generalizable Visual Reinforcement Learning with Segment Anything Modelcited by 12
  23. 2019Disentangling Content and Style via Unsupervised Geometry Distillationcited by 9
  24. 2021Rethinking Content and Style: Exploring Bias for Unsupervised Disentanglementcited by 7
  25. 2024Certified Robust Accuracy of Neural Networks Are Bounded due to Bayes Errorscited by 7
  26. 2022Retriever: Learning Content-Style Representation as a Token-Level Bipartite Graphcited by 6

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2024Focus On What Matters: Separated Models For Visual-Based RL Generalization
  2. 2025Salience-Invariant Consistent Policy Learning for Generalization in Visual Reinforcement Learning
  3. 2022Task-Induced Representation Learning
  4. 2022Masked World Models for Visual Control
  5. 2022Generalization with Lossy Affordances: Leveraging Broad Offline Data for Learning Visuomotor Tasks
  6. 2023Generalizable Task Representation Learning for Offline Meta-Reinforcement Learning with Data Limitations
  7. 2024DMC-VB: A Benchmark for Representation Learning for Control with Visual Distractors
  8. 2025Self-Predictive Dynamics for Generalization of Vision-based Reinforcement Learning