Research map / Vision-language and multimodal / Robot action and navigation
Physically Ground Commonsense Knowledge for Articulated Object Manipulation with Analytic Concepts
Jiude Wei, Yuxuan Li, 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 |
|---|---|---|---|
| Ground MLLM knowledge on analytic concepts instead of semantic-level outputs (vs A3VLM) | MLLM-driven articulated object manipulation · PartNet-Mobility, novel test categories | average success rate (%): 32.1 -> 40.8 | Improved |
| Ground MLLM knowledge on analytic concepts (vs A3VLM) | MLLM-driven articulated object manipulation · PartNet-Mobility, training categories | average success rate (%): 37.4 -> 42.5 | Improved |
| Use analytic concepts with suction end-effector (vs A3VLM) | MLLM-driven manipulation · PartNet-Mobility, suction, test categories | average success rate (%): 66.4 -> 73.8 | Improved |
| Switch end-effector from suction to parallel gripper | A3VLM and ManipLLM · PartNet-Mobility, training categories | average success rate (%): A3VLM 72.4 -> 37.4; ManipLLM 58.0 -> 32.0; Ours 75.5 -> 42.5 | Hurt |
| Estimate grasp pose parameters with cGAN generator-discriminator instead of random sampling (ablation) | Ours · PartNet-Mobility, training categories | average success rate (%): 40.2 -> 42.5 | Improved |
| Estimate grasp pose parameters with cGAN generator-discriminator instead of random sampling (ablation) | Ours · PartNet-Mobility, testing categories | average success rate (%): 38.6 -> 40.8 | Improved |
| Replace structural parameter estimation with ground truth (ablation) | Ours · PartNet-Mobility, training categories | average success rate (%): 51.2 -> 72.0 | Improved |
| Replace 6-DoF pose estimation with ground truth (ablation) | Ours · PartNet-Mobility, training categories | average success rate (%): 72.0 -> 86.3 | Improved |
| Replace actionable part segmentation with ground truth (ablation) | Ours · PartNet-Mobility, training categories | average success rate (%): 42.5 -> 49.8 | Improved |
| Use analytic concepts on complex multi-joint category Table (vs A3VLM) | MLLM-driven manipulation · PartNet-Mobility, Table (test) | success rate (%): 40.0 -> 50.6 | Improved |
Closest papers by meaning
- PAC Bench: Do Foundation Models Understand Prerequisites for Executing Manipulation Policies? (NeurIPS 2025 · similarity 0.90)
- ManipLLM: Embodied Multimodal Large Language Model for Object-Centric Robotic Manipulation (CVPR 2024 · similarity 0.89)
- Discovering Conceptual Knowledge with Analytic Ontology Templates for Articulated Objects (AAAI 2025 · similarity 0.88)
- OmniManip: Towards General Robotic Manipulation via Object-Centric Interaction Primitives as Spatial Constraints (CVPR 2025 · similarity 0.88)
- ClevrSkills: Compositional Language And Visual Reasoning in Robotics (NeurIPS 2024 · similarity 0.87)
- Embodied-R1.5: Evolving Physical Intelligence via Embodied Foundation Models (NeurIPS 2026 · 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.
