Research map / Reinforcement learning / Visual and representation RL
Adaptive Online Replanning with Diffusion Models
Siyuan Zhou, Yilun Du, Shun Zhang, Mengdi Xu, Yikang Shen, Wei Xiao, Dit-Yan Yeung, Chuang Gan
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 |
|---|---|---|---|
| replan with RDM instead of fixed-plan Diffuser | Diffuser · Maze2D Large | normalized return: 123.0 -> 185.3 | Improved |
| replan with RDM instead of Diffuser | Diffuser · Multi2D Large | normalized return: 132.1 -> 195.4 | Improved |
| replace fixed-interval replanning with likelihood-based RDM | Decision Diffuser · Stochastic Hopper Med-Expert | normalized return: 49.4 -> 59.4 | Improved |
| replace fixed-interval replanning with RDM | Decision Diffuser · Stochastic Walker2d Med-Expert | normalized return: 71.1 -> 92.5 | Improved |
| compare RDM with IQL on stochastic Walker2d Medium | IQL · Stochastic Walker2d Medium | normalized return: 71.8 -> 70.7 | Mixed |
| use RDM instead of Decision Diffuser on RLBench | Decision Diffuser · RLBench open box | success rate: 52.9 -> 72.0 | Improved |
| use RDM instead of Decision Diffuser on RLBench | Decision Diffuser · RLBench close box | success rate: 46.0 -> 61.3 | Improved |
| use RDM instead of Decision Diffuser on RLBench | Decision Diffuser · RLBench close fridge | success rate: 84.6 -> 98.6 | Improved |
| replan with only N=40 diffusion steps instead of Diffuser (ablation) | Diffuser · Multi2D Large | normalized return: 129.4 -> 160.9 | Improved |
| increase replanning diffusion steps from N=40 to N=80 (ablation) | RDM · Multi2D Large | normalized return: 160.9 -> 195.4 | Improved |
| increase replanning diffusion steps from N=80 to N=256 (replan from scratch) (ablation) | RDM · Multi2D Large | normalized return: 195.4 -> 197.7 | No clear effect |
| replan on past context (RPDM) or state distance (SDM) instead of fixed intervals (DD) (ablation) | Decision Diffuser · Hopper medium-expert | normalized return | Improved |
Closest papers by meaning
- ReinFlow: Fine-tuning Flow Matching Policy with Online Reinforcement Learning (NeurIPS 2025 · similarity 0.87)
- Multi-Robot Motion Planning with Diffusion Models (ICLR 2025 · similarity 0.87)
- Task-Agnostic Pre-training and Task-Guided Fine-tuning for Versatile Diffusion Planner (ICML 2025 · similarity 0.87)
- Simple Hierarchical Planning with Diffusion (ICLR 2024 · similarity 0.87)
- Generative Trajectory Stitching through Diffusion Composition (NeurIPS 2025 · similarity 0.87)
- One-Step Diffusion Policy: Fast Visuomotor Policies via Diffusion Distillation (ICML 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.
