Research map / Video / Video generation
MoEG-HOI: Mixture of Expert Groups for One-Stage Hand-Object Interaction Motion Generation with Hand-Finger-Joint Semantic Guidance
Hang Xu, Yang Xiao, Changlong Jiang, Haohong Kuang, Kaidi Zhang, Min Du, Ran 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 LLM-refined hand/finger/joint text guidance (ablation) | MoEG-HOI · ARCTIC | FID: 0.867 -> 0.726 | Improved |
| add Hand/Finger/Joint expert group structure (ablation) | MoEG-HOI · ARCTIC | FID: 0.893 -> 0.867 | Improved |
| add hierarchically sized experts (ablation) | MoEG-HOI · ARCTIC | FID: 0.937 -> 0.893 | Improved |
| decoupled encoding of action and timestep in router instead of concatenation (ablation) | MoEG-HOI router · ARCTIC | Contact Ratio: 0.833 -> 0.920 | Improved |
| add action-aware routing (ablation) | MoEG-HOI router · ARCTIC | FID: 0.974 -> 0.726 | Improved |
| add noise-aware (timestep) routing (ablation) | MoEG-HOI router · ARCTIC | FID: 0.820 -> 0.726 | Improved |
| use task-specific MoEG-HOI instead of generic DiT-MoE (ablation) | diffusion MoE · ARCTIC | FID: 0.971 -> 0.726 | Improved |
| use 3 experts per group instead of 2 or 4 (ablation) | MoEG-HOI · ARCTIC | FID: 0.812 (M=2), 0.853 (M=4) -> 0.726 (M=3) | Improved |
| use max expert dimension 1024 instead of 512 or 2048 (ablation) | MoEG-HOI · ARCTIC | FID: 0.910 (512), 0.916 (2048) -> 0.726 (1024) | Improved |
| use MoEG-HOI instead of MF-MDM | HOI motion generation · ARCTIC | Penetration: 0.170 -> 0.253 | Hurt |
| use MoEG-HOI instead of MF-MDM | HOI motion generation · ARCTIC | PSKL-J (gt.p.): 0.3020 -> 0.3334 | Hurt |
| use one-stage MoEG-HOI instead of multi-stage MF-MDM | HOI motion generation · H2O | FID: 10.732 -> 1.648 | Improved |
Closest papers by meaning
- SViMo: Synchronized Diffusion for Video and Motion Generation in Hand-object Interaction Scenarios (NeurIPS 2025 · similarity 0.88)
- CG-HOI: Contact-Guided 3D Human-Object Interaction Generation (CVPR 2024 · similarity 0.88)
- HOIGPT: Learning Long-Sequence Hand-Object Interaction with Language Models (CVPR 2025 · similarity 0.88)
- Text2HOI: Text-guided 3D Motion Generation for Hand-Object Interaction (CVPR 2024 · similarity 0.88)
- InterAct: Advancing Large-Scale Versatile 3D Human-Object Interaction Generation (CVPR 2025 · similarity 0.88)
- LatentHOI: On the Generalizable Hand Object Motion Generation with Latent Hand Diffusion. (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.
