Research map / Reinforcement learning / Visual and representation RL
Learning Diffusion Policy from Primitive Skills for Robot Manipulation
Zhihao Gu, Ming Yang, Difan Zou, Dong Xu
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 cross-attention for vision-language injection (ablation) | DP baseline · LIBERO-Long | success rate (%): 50.5 -> 81.0 | Improved |
| add cross-attention for vision-language injection (ablation) | DP baseline · CALVIN ABC->D | average length: 1.13 -> 3.59 | Improved |
| add AdaLN prior injection of time and proprioception (ablation) | DP + cross-attention · LIBERO-Long | success rate (%): 81.0 -> 86.0 | Improved |
| add AdaLN prior injection of time and proprioception (ablation) | DP + cross-attention · CALVIN ABC->D | average length: 3.59 -> 4.01 | Improved |
| add primitive skill abstraction (skill-dependent FFN) (ablation) | DP + prior injection · LIBERO-Long | success rate (%): 86.0 -> 91.5 | Improved |
| add primitive skill abstraction (skill-dependent FFN) (ablation) | DP + prior injection · CALVIN ABC->D | average length: 4.01 -> 4.32 | Improved |
| add compositional prompt ensemble (ablation) | SDP without CPE · LIBERO-Long | success rate (%): 91.5 -> 93.8 | Improved |
| add compositional prompt ensemble (ablation) | SDP without CPE · CALVIN ABC->D | average length: 4.32 -> 4.49 | Improved |
| parameterize FFN by skill instead of FiLM conditioning (ablation) | SDP · LIBERO-Long | success rate (%): 92.5 -> 93.8 | Improved |
| parameterize FFN by skill instead of element-wise addition (ablation) | SDP · LIBERO-Long | success rate (%): 90.9 -> 93.8 | Improved |
| parameterize FFN by skill instead of concatenation (ablation) | SDP · CALVIN ABC->D | average length: 4.24 -> 4.49 | Improved |
| use SDP instead of MoDE (larger model) | diffusion policy · CALVIN ABC->D | average length: 3.92 -> 4.49 | Mixed |
Closest papers by meaning
- SkillDiffuser: Interpretable Hierarchical Planning via Skill Abstractions in Diffusion-Based Task Execution (CVPR 2024 · similarity 0.92)
- Think Small, Act Big: Primitive Prompt Learning for Lifelong Robot Manipulation (CVPR 2025 · similarity 0.90)
- Conditioning Matters: Training Diffusion Policies is Faster Than You Think (NeurIPS 2025 · similarity 0.89)
- Implicit Drifting Policy: One-Step Action Generation via Conditional Expert Geometry (NeurIPS 2026 · similarity 0.89)
- Zero-Shot Robotic Manipulation with Pre-Trained Image-Editing Diffusion Models (ICLR 2024 · similarity 0.89)
- In-Context Policy Adaptation via Cross-Domain Skill Diffusion (AAAI 2025 · 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.
