Research map / 3D vision / 3D generation
Human-Object Interaction via Automatically Designed VLM-Guided Motion Policy
Zekai Deng, Ye Shi, Kaiyang Ji, Lan Xu, Shaoli Huang, Jingya Wang
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 VLM-guided RMD multi-task formulation vs InterPhys | InterPhys · InterPlay Hybrid | Completion Rate (%): 27.5 -> 53.8 | Improved |
| add top-view image input for VLM planner instead of text-only LLM (ablation) | Ours · InterPlay Static | Completion Rate (%): 62.8 -> 75.1 | Improved |
| add top-view image input for VLM planner instead of text-only LLM (ablation) | Ours · InterPlay Hybrid | Completion Rate (%): 39.9 -> 53.8 | Improved |
| add part-level object decomposition (vs single entity) (ablation) | Ours · Single-task Carry | Completion Rate (%): 75.1 -> 88.3 | Improved |
| add multi-part human-object RMD (vs one human part, one object part) (ablation) | Ours · Single-task Carry | Completion Rate (%): 11.7 -> 88.3 | Improved |
| add multi-part human-object RMD (vs one-one) (ablation) | Ours · Single-task Lie | Completion Rate (%): 28.1 -> 62.0 | Improved |
| add relative position encoding (ablation) | Ours · Single-task Lie | Completion Rate (%): 36.8 -> 62.0 | Improved |
| add relative velocity encoding (ablation) | Ours · Single-task Sit | Completion Rate (%): 76.1 -> 92.6 | Improved |
| add dynamic encoding w'_ij (ablation) | Ours · Single-task Open | Completion Rate (%): 62.1 -> 91.2 | Improved |
| use GPT-4V instead of LLaVA-1.6 as planner (ablation) | Ours · InterPlay-Hybrid | Completion Rate (%): 43.3 -> 53.8 | Improved |
| use GPT-4V instead of Qwen-VL-Max as planner (ablation) | Ours · InterPlay-Hybrid | Completion Rate (%): 47.7 -> 53.8 | Improved |
| use full prompt instead of simplified prompt (ablation) | Ours · InterPlay-Hybrid | Completion Rate (%): 45.3 -> 53.8 | 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.
- TokenHSI — TokenHSI: Unified Synthesis of Physical Human-Scene Interactions through Task Tokenization (CVPR 2025) · better on 6 of 6 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
- InterDreamer: Zero-Shot Text to 3D Dynamic Human-Object Interaction (NeurIPS 2024 · similarity 0.89)
- CG-HOI: Contact-Guided 3D Human-Object Interaction Generation (CVPR 2024 · similarity 0.88)
- HOI-Dyn: Learning Interaction Dynamics for Human-Object Motion Diffusion (NeurIPS 2025 · similarity 0.88)
- Guiding Human-Object Interactions with Rich Geometry and Relations (CVPR 2025 · similarity 0.88)
- InterMimic: Towards Universal Whole-Body Control for Physics-Based Human-Object Interactions (CVPR 2025 · similarity 0.88)
- InterAct: Advancing Large-Scale Versatile 3D Human-Object Interaction Generation (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.
