Research map / Reinforcement learning / Visual and representation RL
From Noise to Control: Parameterized Diffusion Policies
Renhao Zhang, Haotian Fu, Mingxi Jia, George Konidaris, Yilun Du, Bruno Castro da Silva
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 geometry alignment loss (soft-DTW) to latent (ablation) | PDP · CloseDrawer (Scene 3) | success rate (%): 11.7 -> 100.0 | Improved |
| add geometry alignment loss (soft-DTW) to latent (ablation) | PDP · PickUpCup (Scene 3) | success rate (%): 62.5 -> 82.5 | Improved |
| use continuous latent instead of discrete one-hot (ablation) | PDP · CloseDrawer (Scene 3) | success rate (%): 0.0 -> 100.0 | Improved |
| use continuous latent instead of discrete one-hot (ablation) | PDP · PickUpCup (Scene 3) | success rate (%): 0.0 -> 82.5 | Improved |
| use global modulation instead of concatenation (ablation) | PDP · CloseDrawer (Scene 3) | success rate (%): 82.5 -> 100.0 | Improved |
| use global modulation instead of concatenation (ablation) | PDP · PickUpCup (Scene 3) | success rate (%): 25.0 -> 82.5 | Improved |
| add behavior latent conditioning with global modulation to DP (ablation) | PDP · CloseDrawer (Scene 3) | success rate (%): 40.0 -> 100.0 | Improved |
| add behavior latent conditioning with global modulation to DP (ablation) | PDP · PickUpCup (Scene 3) | success rate (%): 15.8 -> 82.5 | Improved |
| use encoder warm-start instead of random init (100 steps) (ablation) | PDP · CloseDrawer (Scene 3) | success rate (%): 2.5 -> 100.0 | Improved |
| use encoder warm-start instead of random init (100 steps) (ablation) | PDP · PickUpCup (Scene 3) | success rate (%): 53.3 -> 82.5 | Improved |
| add 10 latent optimization steps on warm-start (ablation) | PDP · PickUpCup (Scene 3) | success rate (%): 61.7 -> 81.7 | Improved |
| add 100 instead of 10 latent optimization steps on warm-start (ablation) | PDP · PickUpCup (Scene 3) | success rate (%): 81.7 -> 82.5 | No clear effect |
Closest papers by meaning
- Diffusion Policy Policy Optimization (ICLR 2025 · similarity 0.91)
- One-Step Diffusion Policy: Fast Visuomotor Policies via Diffusion Distillation (ICML 2025 · similarity 0.91)
- Make-An-Agent: A Generalizable Policy Network Generator with Behavior-Prompted Diffusion (NeurIPS 2024 · similarity 0.89)
- Diffusion-ES: Gradient-free Planning with Diffusion for Autonomous and Instruction-guided Driving (CVPR 2024 · similarity 0.89)
- Conditioning Matters: Training Diffusion Policies is Faster Than You Think (NeurIPS 2025 · similarity 0.89)
- Score Regularized Policy Optimization through Diffusion Behavior (ICLR 2024 · similarity 0.89)
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.
