Research map / Reinforcement learning / Visual and representation RL
Ada-Diffuser: Latent-Aware Adaptive Diffusion for Decision-Making
Fan Feng, Selena Ge, Minghao Fu, Zijian Li, Yujia Zheng, Zeyu Tang, Yingyao Hu, Biwei Huang, Kun Zhang
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 latent identification module (ablation) | Ada-Diffuser planner · Cheetah-Wind-S | normalized return: -103.5 -> -73.5 | Improved |
| keep training the latent module (not frozen) (ablation) | Ada-Diffuser planner · Cheetah-Wind-S | normalized return: -110.4 -> -73.5 | Improved |
| add backward refinement step (ablation) | Ada-Diffuser planner · Cheetah-Wind-S | normalized return: -82.0 -> -73.5 | Improved |
| add zig-zag sampling (ablation) | Ada-Diffuser planner · Cheetah-Wind-S | normalized return: -91.6 -> -73.5 | Improved |
| use causal (increasing) noise schedule instead of same noise level (ablation) | Ada-Diffuser planner · Cheetah-Wind-S | normalized return: -89.7 -> -73.5 | Improved |
| add backward refinement step (ablation) | Ada-Diffuser policy · LIBERO | success rate: 83.9 -> 93.4 | Improved |
| add zig-zag sampling (ablation) | Ada-Diffuser policy · LIBERO | success rate: 91.4 -> 93.4 | Improved |
| add latent module (ablation) | Ada-Diffuser policy · LIBERO | success rate: 89.3 -> 93.4 | Improved |
| increase latent dimension to 4x-6x (ablation) | Ada-Diffuser planner · Cheetah-Wind-S | normalized return: -73.5 -> -102.4 (6x) | Hurt |
| vary latent dimension within 0.5x-2x (ablation) | Ada-Diffuser planner · Cheetah-Wind-S | normalized return: -85.2 (0.5x), -77.6 (2x) vs -73.5 | No clear effect |
| add latent factor modeling to DF planner | DF · Cheetah-Wind-E | normalized return: -105.8 -> -68.9 | Improved |
| Ada-Diffuser vs plug-in DynaMITE context module | DF · Ant-Dir-E | normalized return: 266.7 -> 285.3 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MetaDiffuser — MetaDiffuser: Diffusion Model as Conditional Planner for Offline Meta-RL (ICML 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Act to See, See to Act: Diffusion-Driven Perception-Action Interplay for Adaptive Policies (NeurIPS 2025 · similarity 0.90)
- Efficient Planning with Latent Diffusion (ICLR 2024 · similarity 0.89)
- Score Regularized Policy Optimization through Diffusion Behavior (ICLR 2024 · similarity 0.89)
- Prior-Guided Diffusion Planning for Offline Reinforcement Learning (NeurIPS 2025 · similarity 0.89)
- Latent Plan Transformer for Trajectory Abstraction: Planning as Latent Space Inference (NeurIPS 2024 · similarity 0.89)
- One-Step Diffusion Policy: Fast Visuomotor Policies via Diffusion Distillation (ICML 2025 · similarity 0.88)
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.
