Research map / Vision-language and multimodal / Robot action and navigation
Generalizable Coarse-to-Fine Robot Manipulation via Language-Aligned 3D Keypoints
Jianshu Hu, Lidi Wang, Shujia Li, Yunpeng Jiang, Xiao Li, Paul Weng, Yutong Ban
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 object position reasoning before step instruction and keypoint (ablation) | CLAP planner base · GemBench avg | average success rate (%): 48.0 -> 48.9 | Improved |
| add language plan data and plan reasoning (two-round inference) (ablation) | CLAP planner base · GemBench avg | average success rate (%): 48.0 -> 48.3 | No clear effect |
| include previous step instruction as short-term memory (ablation) | CLAP planner with plan and object reasoning · GemBench Level 4 | success rate (%): 2.0 -> 30.4 (with other components) | Improved |
| add object position dataset for auxiliary 3D object detection (ablation) | CLAP · GemBench avg | average success rate (%): 60.1 -> 62.0 | Improved |
| use pretrained SigLIP/DINOv2 encoders with 3D position embedding (ablation) | CLAP fine-grained action predictor · GemBench avg | average success rate (%): 57.8 -> 62.0 | Improved |
| CLAP on Level 1 placements compared with strongest baseline | CLAP · GemBench Level 1 | success rate (%): 94.3 -> 83.9 (vs 3D-LOTUS) | Hurt |
| train CLAP with 20 trajectories per variation instead of 100 | CLAP · GemBench avg | average success rate (%): 12% higher than prior SOTA | Improved |
| CLAP vs RVT2 in real world with 10 demos per task | CLAP · Real-world tasks | average success rate (%): 22.5% -> 77.3% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- BridgeVLA — BridgeVLA: Input-Output Alignment for Efficient 3D Manipulation Learning with Vision-Language Models (NeurIPS 2025) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- GravMAD: Grounded Spatial Value Maps Guided Action Diffusion for Generalized 3D Manipulation (ICLR 2025 · similarity 0.89)
- OmniManip: Towards General Robotic Manipulation via Object-Centric Interaction Primitives as Spatial Constraints (CVPR 2025 · similarity 0.89)
- PIVOT-R: Primitive-Driven Waypoint-Aware World Model for Robotic Manipulation (NeurIPS 2024 · similarity 0.89)
- Exploring the Limits of Vision-Language-Action Manipulation in Cross-task Generalization (NeurIPS 2025 · similarity 0.88)
- FineVLA: Fine-Grained Instruction Alignment for Steerable Vision-Language-Action Policies (NeurIPS 2026 · similarity 0.88)
- Phoenix: A Motion-based Self-Reflection Framework for Fine-grained Robotic Action Correction (CVPR 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.
