Research map / Point clouds / Point cloud classification
Action-Geometry Prediction with 3D Geometric Prior for Bimanual Manipulation
Chongyang Xu, Haipeng Li, Shen Cheng, Haoqiang Fan, Ziliang Feng, Shuaicheng Liu
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 2D semantic module (DINOv3 features) (ablation) | GAP policy · RoboTwin 2.0 ablation (4 tasks avg) | avg success rate (%): 24.4 -> 25.1 | Improved |
| add future 3D pointmap prediction (geometric imagination) (ablation) | GAP policy · RoboTwin 2.0 ablation (4 tasks avg) | avg success rate (%): 23.6 -> 25.1 | Improved |
| add 3D geometric module (pi3 features) on top of imagination-free model (ablation) | GAP policy · RoboTwin 2.0 ablation (4 tasks avg) | avg success rate (%): 21.0 -> 23.6 | Improved |
| RGB-only policy with pi3 3D prior vs point-cloud DP3 | GAP policy · RoboTwin 2.0 Dominant-select (16 tasks) | avg success rate (%): 61.2 -> 63.2 | Improved |
| RGB-only policy vs Xu et al. (2D future-frame prediction) | GAP policy · RoboTwin 2.0 Sync-bimanual (8 tasks) | avg success rate (%): 47.6 -> 51.3 | Improved |
| RGB-only policy vs G3Flow | GAP policy · RoboTwin 2.0 Seq-coordinate (8 tasks) | avg success rate (%): 46.3 -> 50.4 | Improved |
| RGB-only policy vs DP3 on Place Dual Shoes | GAP policy · RoboTwin 2.0 Place Dual Shoes | success rate (%): 17.7 -> 43.3 | Improved |
| GAP vs DP3 on Beat Block Hammer | GAP policy · RoboTwin 2.0 Dominant-select (16 tasks) | success rate (%): 59.3 -> 56.3 | Hurt |
| GAP vs best baseline on real robot | GAP policy · Real-world (4 bimanual tasks) | avg success rate (%): 32.5 -> 40 | Improved |
Closest papers by meaning
- Diffusion-Based Imaginative Coordination for Bimanual Manipulation (ICCV 2025 · similarity 0.90)
- RDT-1B: a Diffusion Foundation Model for Bimanual Manipulation (ICLR 2025 · similarity 0.89)
- DynaRend: Learning 3D Dynamics via Masked Future Rendering for Robotic Manipulation (NeurIPS 2025 · similarity 0.88)
- Spatial-Temporal Graph Diffusion Policy with Kinematic Modeling for Bimanual Robotic Manipulation (CVPR 2025 · similarity 0.88)
- BimArt: A Unified Approach for the Synthesis of 3D Bimanual Interaction with Articulated Objects (CVPR 2025 · similarity 0.88)
- BiAssemble: Learning Collaborative Affordance for Bimanual Geometric Assembly (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.
