Research map / Reinforcement learning / Visual and representation RL
Spectrum Random Masking for Generalization in Image-based Reinforcement Learning
Yangru Huang, Peixi Peng, Yifan Zhao, Guangyao Chen, Yonghong Tian
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add SRM to DrQ | DrQ · DMControl-GB color-hard average | episode return: 603 -> 833 | Improved |
| add SRM to DrQ | DrQ · DMControl-GB video-hard average | episode return: 139 -> 227 | Improved |
| add SRM to DrQ | DrQ · DMControl-GB video-easy average | episode return: 578 -> 754 | Improved |
| add SRM to SVEA-C | SVEA-C · DMControl-GB video-easy average | episode return: 688 -> 855 | Improved |
| add SRM to SVEA-O | SVEA-O · DMControl-GB video-hard average | episode return: 468 -> 581 | Improved |
| add SRM to SVEA-O | SVEA-O · DMControl-GB color-easy average | episode return: 929 -> 942 | Improved |
| add SRM to SVEA-C | SVEA-C · DMControl-GB color-easy average | episode return: 866 -> 946 | Improved |
| random convolution, rotation, flip, cutout or dropout instead of SRM | DrQ · DMControl-GB | episode return | Hurt |
| random convolution instead of SRM | DrQ · DMControl-GB color-hard | episode return | Mixed |
| use erasing, noise or swapping masking strategy (ablation) | SRM · DMControl-GB | episode return | Improved |
| mixture of erasing, noise and swapping instead of one strategy (ablation) | SRM · DMControl-GB | episode return | Hurt |
| random masking band instead of fixed band (r1 0.1-0.4) (ablation) | SRM · DMControl-GB | episode return | Improved |
Closest papers by meaning
- Zero Shot Generalization of Vision-Based RL Without Data Augmentation (ICML 2025 · similarity 0.90)
- Focus On What Matters: Separated Models For Visual-Based RL Generalization (NeurIPS 2024 · similarity 0.89)
- A Simple Framework for Generalization in Visual RL under Dynamic Scene Perturbations (NeurIPS 2024 · similarity 0.88)
- The Surprising Ineffectiveness of Pre-Trained Visual Representations for Model-Based Reinforcement Learning (NeurIPS 2024 · similarity 0.88)
- Revisiting Data Augmentation in Deep Reinforcement Learning (ICLR 2024 · similarity 0.87)
- VIGOR: Zero-Shot Visual Generalization via Latent-Space Consistency in Model-Based Reinforcement Learning (NeurIPS 2026 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
