Research map / Reinforcement learning / Visual and representation RL
Diffusion Guided Adaptive Augmentation for Generalization in Visual Reinforcement Learning
Jeong Woon Lee, Hyoseok Hwang
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 DGA2 augmentation | SAC · DMControl-GB Video Hard | episode return: 136 -> 616 | Improved |
| add DGA2 augmentation | DrQ · DMControl-GB Video Hard | episode return: 135 -> 680 | Improved |
| add DGA2 augmentation | SVEA · DMControl-GB Video Hard | episode return: 157 -> 705 | Improved |
| add DGA2 to PPO | PPO · Procgen | average rank: 4.75 -> 3.50 | Improved |
| add DGA2 to DrAC | DrAC · Procgen | average rank: 3.63 -> 1.50 | Improved |
| add DGA2 vs best prior (PIE-G) | PIE-G · Adroit Door Video Hard | success rate (%): 41.3 -> 53.0 | Improved |
| add DGA2 vs PIE-G on Hammer Video Hard | PIE-G · Adroit Hammer Video Hard | success rate (%): 45.7 -> 22.0 | Hurt |
| replace flip augmentation with DGA2 | SAC · DMControl-GB avg | episode return: 480 -> 793 | Improved |
| use Rotation augmentation | SAC · DMControl-GB avg | episode return: 493 -> 339 | Hurt |
| use Random Conv instead of DGA2 (ablation) | SAC · DMControl-GB avg | episode return: 793 -> 651 | Hurt |
| use adaptive domain shift instead of identical domain (ablation) | SAC · Walker walk 300K | episode return: 487 -> 567 | Improved |
| use adaptive domain shift instead of fixed-step domain switching (ablation) | SAC · Walker walk 300K | episode return: 453 -> 567 | Improved |
Closest papers by meaning
- Fourier Guided Adaptive Adversarial Augmentation for Generalization in Visual Reinforcement Learning (AAAI 2025 · similarity 0.90)
- A Simple Framework for Generalization in Visual RL under Dynamic Scene Perturbations (NeurIPS 2024 · similarity 0.88)
- Efficient Reinforcement Learning Through Adaptively Pretrained Visual Encoder (AAAI 2025 · similarity 0.88)
- What Effects the Generalization in Visual Reinforcement Learning: Policy Consistency with Truncated Return Prediction (AAAI 2024 · similarity 0.88)
- Zero Shot Generalization of Vision-Based RL Without Data Augmentation (ICML 2025 · similarity 0.87)
- Focus On What Matters: Separated Models For Visual-Based RL Generalization (NeurIPS 2024 · 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.
