Research map / Continual learning / Online and replay-based
iManip: Skill-Incremental Learning for Robotic Manipulation
Zexin Zheng, Jia-Feng Cai, Xiao-Ming Wu, Yi-Lin Wei, Yu-Ming Tang, Ancong Wu, Wei-Shi Zheng
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 temporal replay strategy (TRS) (ablation) | iManip · RLBench B5-1N1 | success rate (%): 20.7 -> 49.3 (+28.6%) | Improved |
| add temporal replay strategy (TRS) (ablation) | iManip · RLBench B5-5N1 | success rate (%): 5.2 -> 27.6 (+22.4%) | Improved |
| add extendable PerceiverIO (EPIO) (ablation) | iManip · RLBench B5-1N1 | success rate (%): 49.3 -> 54.0 (+4.7%) | Improved |
| add extendable PerceiverIO (EPIO) (ablation) | iManip · RLBench B5-5N1 | success rate (%): 27.6 -> 32.4 (+4.8%) | Improved |
| add knowledge distillation (DIS) (ablation) | iManip · RLBench B5-1N1 | success rate (%): 54.0 -> 56.7 (+2.7%) | Improved |
| add knowledge distillation (DIS) (ablation) | iManip · RLBench B5-5N1 | success rate (%): 32.4 -> 36.0 (+3.6%) | Improved |
| freeze encoder and extendable PerceiverIO, train only decoder with new prompts (ablation) | iManip · RLBench B5-1N1 slide block | old-skill success rate (%): 43.2 -> 57.6 | Improved |
| freeze the policy decoder (ablation) | iManip · RLBench B5-1N1 slide block | new-skill success rate (%): 60.0 -> 24.0 | Hurt |
| larger distillation weight lambda_dis (ablation) | iManip · RLBench | success rate (%) | Hurt |
| increase dimension d_new of appended PerceiverIO weights beyond 8 (ablation) | iManip · RLBench | success rate (%) | Hurt |
| increase action prompt length from 4 to 32 (ablation) | iManip · RLBench | success rate (%) | Improved |
| increase replays per keyframe from 1 to 5 (ablation) | iManip · RLBench | success rate (%) | Improved |
Closest papers by meaning
- Think Small, Act Big: Primitive Prompt Learning for Lifelong Robot Manipulation (CVPR 2025 · similarity 0.88)
- Multi-Stage Manipulation with Demonstration-Augmented Reward, Policy, and World Model Learning (ICML 2025 · similarity 0.87)
- SCaR: Refining Skill Chaining for Long-Horizon Robotic Manipulation via Dual Regularization (NeurIPS 2024 · similarity 0.87)
- Policy Compatible Skill Incremental Learning via Lazy Learning Interface (NeurIPS 2025 · similarity 0.86)
- RoboGen: Towards Unleashing Infinite Data for Automated Robot Learning via Generative Simulation (ICML 2024 · similarity 0.86)
- ManipTrans: Efficient Dexterous Bimanual Manipulation Transfer via Residual Learning (CVPR 2025 · similarity 0.86)
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.
