Research map / Reinforcement learning / Offline RL
S2P: State-conditioned Image Synthesis for Data Augmentation in Offline Reinforcement Learning
Daesol Cho, Dongseok Shim, H. Jin Kim
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 S2P augmented images (50k extra) | IQL · DMControl cheetah run (mixed) | normalized score: 41.68 -> 88.53 | Improved |
| add S2P augmented images | SLAC-off · DMControl walker walk (mixed) | normalized score: 29.02 -> 92.60 | Improved |
| add S2P augmented images | CQL · DMControl reacher easy (expert) | normalized score: 57.68 -> 32.54 | Hurt |
| add S2P augmented images | IQL · DMControl finger spin (expert) | normalized score: 98.42 -> 94.42 | Hurt |
| rollout with offline-RL policy instead of random policy (ablation) | CQL+S2P · cheetah run (random) | normalized score: -0.69 -> 11.77 | Improved |
| rollout with random policy instead of offline-RL policy (ablation) | IQL+S2P · cheetah run (expert) | normalized score: 79.20 -> 87.18 | Improved |
| replace Max Var with Ens Var uncertainty (ablation) | CQL+S2P · cheetah run (random) | normalized score: 11.77 -> 19.83 | Improved |
| replace Max Var with Ens Var uncertainty (ablation) | IQL+S2P · cheetah run (expert) | normalized score: 87.18 -> 62.41 | Hurt |
| increase rollout horizon from 1 to 5 steps (ablation) | IQL+S2P · cheetah run (mixed) | normalized score: 88.53 -> 66.38 | Hurt |
| increase rollout horizon from 1 to 5 steps (ablation) | SLAC-off+S2P · cheetah run (expert) | normalized score: 14.41 -> 17.17 | Improved |
| add S2P data to policy-constraint methods | BC · cheetah mixed | normalized score: 51.81 -> 0.06 | Hurt |
| replace S2P images with Dreamer images (ablation) | IQL · cheetah run (mixed) | normalized score: 88.53 -> 2.09 | Hurt |
Closest papers by meaning
- Temporal Distance-aware Transition Augmentation for Offline Model-based Reinforcement Learning (ICML 2025 · similarity 0.89)
- Video-Enhanced Offline Reinforcement Learning: A Model-Based Approach (ICML 2025 · similarity 0.88)
- Planning, Fast and Slow: Online Reinforcement Learning with Action-Free Offline Data via Multiscale Planners (ICML 2024 · similarity 0.88)
- Making Offline RL Online: Collaborative World Models for Offline Visual Reinforcement Learning (NeurIPS 2024 · similarity 0.88)
- Scaling Offline Model-Based RL via Jointly-Optimized World-Action Model Pretraining (ICLR 2025 · similarity 0.88)
- Scaling Offline RL via Efficient and Expressive Shortcut Models (NeurIPS 2025 · 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.
