Research map / Reinforcement learning / Visual and representation RL
Beyond Mimicry: Learning Whole-Body Human-Humanoid Interaction from Human-Human Demonstrations
Wei-Jin Huang, Yue-Yi Zhang, Yi-Lin Wei, Zhi-Wei Xia, Juantao Tan, Yuan-Ming Li, Zhilin Zhao, Wei-Shi Zheng
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 contact preservation loss Lcon (ablation) | PAIR · contact@0.35m | contact F1: 0.821 -> 0.841 | Improved |
| add human partner adaptation (ablation) | PAIR · contact@0.35m | contact F1: 0.823 -> 0.841 | Improved |
| use two-stage coarse-to-fine optimization instead of single stage (ablation) | PAIR · contact@0.35m | contact F1: 0.788 -> 0.841 | Improved |
| add contact loss and two-stage on smoothness (ablation) | PAIR · retargeting | jerk mean: 0.0008 vs 0.0007-0.0008 without | No clear effect |
| PAIR vs ImitationNet | PAIR · contact@0.35m | contact F1: 0.502 -> 0.841 | Improved |
| add Phase Attention (when to act) (ablation) | D-STAR · HHoI sim, avg 6 tasks | success rate (%): 65.9 -> 75.4 | Improved |
| add Multi-Scale Spatial module (where to act) (ablation) | D-STAR · HHoI sim, avg 6 tasks | success rate (%): 64.3 -> 75.4 | Improved |
| remove both PA and MSS vs removing one (ablation) | D-STAR · HHoI sim, avg 6 tasks | success rate (%): w/o PA+MSS 68.3 vs w/o PA 65.9 and w/o MSS 64.3 | Mixed |
| D-STAR vs Transformer policy | D-STAR · HHoI sim, avg 6 tasks | success rate (%): 64.3 -> 75.4 | Improved |
| D-STAR vs Diffusion Policy on High-Five | D-STAR · HHoI sim, High-Five | success rate (%): 3.7 -> 40.7 | Improved |
| Naive mimicry (BC only) without decoupling | Naive Mimicry · HHoI sim | success rate (%): 0.0 average | Hurt |
Closest papers by meaning
- HumanoidArena: Benchmarking Egocentric Hierarchical Whole-body Learning (NeurIPS 2026 · similarity 0.89)
- InterMimic: Towards Universal Whole-Body Control for Physics-Based Human-Object Interactions (CVPR 2025 · similarity 0.88)
- SUGAR: A Scalable Human-Video-Driven Generalizable Humanoid Loco-Manipulation Learning Framework (NeurIPS 2026 · similarity 0.88)
- H-GAP: Humanoid Control with a Generalist Planner (ICLR 2024 · similarity 0.87)
- Reconstructing Close Human Interaction with Appearance and Proxemics Reasoning (CVPR 2025 · similarity 0.87)
- Humanoid Horizon: Extending Task Horizon in Whole-Body Loco-Manipulation via Parallel Training, Dynamic Starting, and Reward Gating (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.
