Research map / Image generation / Fast diffusion sampling
Contractive Diffusion Policies: Robust Action Diffusion via Contractive Score-Based Sampling with Differential Equations
Amin Abyaneh, Charlotte Morissette, Mohamad H. Danesh, Anas Houssaini, David Meger, Gregory Dudek, Hsiu-Chin Lin
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 contraction loss to DBC-DiT (CDP-DiT) | DBC-DiT · Robomimic average | success rate: 0.64 -> 0.78 | Improved |
| add contraction loss to EDP | EDP · D4RL average | normalized return: 61.2 -> 65.7 | Improved |
| add contraction loss to EDP | EDP · D4RL Hopper medium-replay | normalized return: 55.1 -> 63.5 | Improved |
| add contraction loss to EDP | EDP · D4RL Franka Kitchen complete | normalized return: 32.9 -> 51.0 | Improved |
| add contraction loss to EDP | EDP · D4RL Walker2D medium | normalized return: 81.7 -> 86.5 | Improved |
| add contraction loss to EDP | EDP · D4RL Antmaze medium diverse | normalized return: 25.6 -> 31.8 | Improved |
| add contraction loss to EDP | EDP · D4RL HalfCheetah medium | normalized return: 46.2 -> 46.0 | Hurt |
| add contraction loss to EDP | EDP · D4RL Walker2D medium-expert | normalized return: 109.8 -> 109.4 | No clear effect |
| use CDP-Unet instead of CDP-DiT backbone | CDP · Robomimic average | success rate: 0.78 -> 0.90 | Improved |
| add contraction loss to DP-Unet (CDP-Unet) | DP-Unet · Robomimic average | success rate: 0.88 -> 0.90 | Improved |
| add contraction loss to DBC-DiT | DBC-DiT · Robomimic Transport-L | success rate: 0.13 -> 0.48 | Improved |
| add contraction loss to EDP | EDP · D4RL training time (s per 100k steps) | time: 4594 -> 5236 | Hurt |
Closest papers by meaning
- Score Regularized Policy Optimization through Diffusion Behavior (ICLR 2024 · similarity 0.88)
- CCS: Controllable and Constrained Sampling with Diffusion Models via Initial Noise Perturbation (NeurIPS 2025 · similarity 0.88)
- DiRecT: Safe Diffusion-Based Planning via Receding-Horizon Denoising (NeurIPS 2026 · similarity 0.88)
- Contractive Dynamical Imitation Policies for Efficient Out-of-Sample Recovery (ICLR 2025 · similarity 0.87)
- Prior-Guided Diffusion Planning for Offline Reinforcement Learning (NeurIPS 2025 · similarity 0.87)
- Hierarchical Variational Policies for Reward-Guided Diffusion (NeurIPS 2026 · 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.
