Research map / 3D vision / 3D generation
SpatialActor: Exploring Disentangled Spatial Representations for Robust Robotic Manipulation
Hao Shi, Bin Xie, Yingfei Liu, Yang Yue, Tiancai Wang, Haoqiang Fan, Xiangyu Zhang, Gao Huang
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 |
|---|---|---|---|
| decouple semantics and geometry (ablation) | RVT-2-style baseline · RLBench 18 tasks, heavy noise | average success rate (%): 57.0 -> 68.7 | Improved |
| decouple semantics and geometry (ablation) | RVT-2-style baseline · RLBench 18 tasks, no noise | average success rate (%): 81.4 -> 85.1 | Improved |
| add Semantic-guided Geometric Module (SGM) (ablation) | decoupled model · RLBench 18 tasks, heavy noise | average success rate (%): 68.7 -> 73.9 | Improved |
| add Semantic-guided Geometric Module (SGM) (ablation) | decoupled model · RLBench 18 tasks, no noise | average success rate (%): 85.1 -> 86.4 | Improved |
| add Spatial Transformer (SPT) (ablation) | decoupled model with SGM · RLBench 18 tasks, heavy noise | average success rate (%): 73.9 -> 76.4 | Improved |
| add Spatial Transformer (SPT) (ablation) | decoupled model with SGM · RLBench 18 tasks, no noise | average success rate (%): 86.4 -> 87.4 | Improved |
| use SpatialActor instead of RVT-2 under heavy depth noise | RVT-2 · RLBench heavy noise | average success rate (%): 57.0 -> 76.4 | Improved |
| use SpatialActor instead of RVT-2 for few-shot adaptation | RVT-2 · RLBench few-shot (19 novel tasks, 10 demos) | average success rate (%): 46.9 -> 79.2 | Improved |
| use SpatialActor instead of RVT-2 on high-precision tasks | RVT-2 · RLBench Insert Peg | success rate (%): 40.0 -> 93.3 | Improved |
| use SpatialActor instead of RVT-2 on Close Jar and Open Drawer | RVT-2 · RLBench (18 tasks, 249 variations) | success rate (%): Close Jar 100.0 -> 94.0; Open Drawer 74.0 -> 82.0 | Mixed |
| use SpatialActor instead of RVT-2 in real-world Stack Cup and Stack Block | RVT-2 · Real-world | success rate (%): Stack Cup 30% -> 30%; Stack Block 40% -> 35% | No clear effect |
Closest papers by meaning
- Localizing, Structuring, and Rendering: Bridging 3D and 2D Vision-Language-Action Models for Robotic Manipulation (CVPR 2026 · similarity 0.91)
- OmniManip: Towards General Robotic Manipulation via Object-Centric Interaction Primitives as Spatial Constraints (CVPR 2025 · similarity 0.88)
- Lift3D Policy: Lifting 2D Foundation Models for Robust 3D Robotic Manipulation (CVPR 2025 · similarity 0.88)
- RoboRefer: Towards Spatial Referring with Reasoning in Vision-Language Models for Robotics (NeurIPS 2025 · similarity 0.88)
- SoFar: Language-Grounded Orientation Bridges Spatial Reasoning and Object Manipulation (NeurIPS 2025 · similarity 0.88)
- DynaRend: Learning 3D Dynamics via Masked Future Rendering for Robotic Manipulation (NeurIPS 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.
