Research map / Reinforcement learning / Imitation learning
Geometry-aware Policy Imitation
Yiming Li, Nael Darwiche, Amirreza Razmjoo, Sichao Liu, Yilun Du, Auke Ijspeert, Sylvain Calinon
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 |
|---|---|---|---|
| replace diffusion policy with GPI (state-based) | Push-T · Push-T state-based | avg score: 82.3 -> 85.8 (vs DDPM) | Improved |
| replace diffusion policy with GPI (vision-based) | Push-T · Push-T vision-based | avg score: 80.9 -> 83.3 (vs DDPM) | Improved |
| replace DDIM diffusion policy with GPI | Push-T state-based · inference time | ms per step: 65 ms -> 0.6 ms | Improved |
| replace diffusion policy with GPI | Push-T state-based · memory | MB: 252 MB -> 0.7 MB | Improved |
| GPI vs diffusion policy on Robomimic/Adroit | Robomimic and Adroit · Robomimic Square | success rate: 0.87 -> 0.82 | Mixed |
| GPI vs diffusion policy on Adroit Pen | Adroit · Adroit Pen | success rate: 0.89 -> 0.95 | Improved |
| vary number of neighbors K = 1,3,5,10 (ablation) | GPI · Push-T | score | No clear effect |
| vary action horizon up to 16 (ablation) | GPI · Push-T | score | No clear effect |
| use pretrained SAM + pose module as embedding instead of VAE (ablation) | GPI · Push-T vision-based | avg score: 88 -> 41 | Hurt |
| use VAE encoder instead of ResNet+PCA (ablation) | GPI · Push-T vision-based | avg score: 84 -> 88 | Improved |
| increase Gaussian noise on query state (ablation) | GPI · Push-T | score | Mixed |
| increase number of demonstrations from 1K to 20K (ablation) | GPI state-based · Push-T | success rate | Improved |
Closest papers by meaning
- Implicit Drifting Policy: One-Step Action Generation via Conditional Expert Geometry (NeurIPS 2026 · similarity 0.91)
- GO-DICE: Goal-Conditioned Option-Aware Offline Imitation Learning via Stationary Distribution Correction Estimation (AAAI 2024 · similarity 0.89)
- OSVI-WM: One-Shot Visual Imitation for Unseen Tasks using World-Model-Guided Trajectory Generation (NeurIPS 2025 · similarity 0.89)
- Policy Decorator: Model-Agnostic Online Refinement for Large Policy Model (ICLR 2025 · similarity 0.88)
- FlowPolicy: Enabling Fast and Robust 3D Flow-Based Policy via Consistency Flow Matching for Robot Manipulation (AAAI 2025 · similarity 0.88)
- Latent Diffusion Planning for Imitation Learning (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.
