Research map / Vision-language and multimodal / Robot action and navigation
StructPolicy: Structure-Guided Imitation Learning Robust to Visual Domain Shifts
Zehao Du, Jiude Wei, Cewu Lu, Jianhua Sun
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 StructPolicy to IL backbones | DP / DP3 / Lift3D · MetaWorld (15 tasks) | success rate (%): 65.2 -> 77.5 (DP); 83.9 -> 90.5 (Lift3D) | Improved |
| add StructPolicy to IL backbones | MDT-V / FLOWER etc. · CALVIN D->D | avg. chain length: 4.35 -> 4.42 (FLOWER) | Improved |
| add StructPolicy | FLOWER · CALVIN ABCD->D | avg. chain length: 4.67 -> 4.70 | Improved |
| use Structure Map instead of mask/keypoint/pose/bbox cues (ablation) | DP · MetaWorld (15 tasks) | success rate (%): 71.1 -> 77.5 vs best alternative | Improved |
| use Structure Map instead of simpler object cues (ablation) | DP3 · MetaWorld (15 tasks) | success rate (%): 73.6 -> 79.2 vs best alternative | Improved |
| use structured instead of uniform point sampling (ablation) | StructEncoder · MetaWorld (15 tasks) | success rate (%): 80.5 -> 90.5 | Improved |
| use StructEncoder instead of PointTransformer (ablation) | map encoder · MetaWorld (15 tasks) | success rate (%): 80.8 -> 90.5 | Improved |
| add affordance stream to structure stream (ablation) | StructEncoder · MetaWorld (15 tasks) | success rate (%): 85.2 -> 90.5 | Improved |
| add structure stream to affordance stream (ablation) | StructEncoder · MetaWorld (15 tasks) | success rate (%): 74.6 -> 90.5 | Improved |
| use MLP fusion instead of attention fusion (ablation) | StructEncoder · MetaWorld (15 tasks) | success rate (%): 88.4 -> 90.5 | Improved |
| provide ground-truth structure template (ablation) | StructCon · MetaWorld (15 tasks) | success rate (%): 90.5 -> 91.0 | No clear effect |
| provide ground-truth parameters (ablation) | StructCon · MetaWorld (15 tasks) | success rate (%): 91.0 -> 98.4 | Improved |
Closest papers by meaning
- Robots Pre-train Robots: Manipulation-Centric Robotic Representation from Large-Scale Robot Datasets (ICLR 2025 · similarity 0.88)
- Rethinking Latent Redundancy in Behavior Cloning: An Information Bottleneck Approach for Robot Manipulation (ICML 2025 · similarity 0.87)
- What's the Move? Hybrid Imitation Learning via Salient Points (ICLR 2025 · similarity 0.87)
- SAM-E: Leveraging Visual Foundation Model with Sequence Imitation for Embodied Manipulation (ICML 2024 · similarity 0.87)
- GravMAD: Grounded Spatial Value Maps Guided Action Diffusion for Generalized 3D Manipulation (ICLR 2025 · similarity 0.87)
- Scaffolding Dexterous Manipulation with Vision-Language Models (NeurIPS 2025 · 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.
