InterMoE: Individual-Specific 3D Human Interaction Generation via Dynamic Temporal-Selective MoE
Lipeng Wang, Hongxing Fan, Haohua Chen, Zehuan Huang, Lu Sheng · 2025
Read InterMoE: Individual-Specific 3D Human Interaction Generation via Dynamic Temporal-Selective MoE by Lipeng Wang, Hongxing Fan, Haohua Chen, Zehuan Huang, Lu Sheng (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-13488.txt.
Description
Lipeng Wang and 4 co-authors present InterMoE, a system that generates 3D interactions between two people while keeping each person's distinct style. Older generators often washed out individual traits or drifted from the text prompt that describes the action. Preserving identity and prompt fidelity at the same time is the core challenge.
InterMoE routes motion features through a mixture of experts chosen anew over time. Routing uses both high-level text meaning and low-level motion context, so experts can focus on the moments that matter most. Tests cut the standard realism error by nine percent on InterHuman and twenty-two percent on InterX, reaching top results for person-specific interaction generation.
Catalog record
| CREATOR | Lipeng Wang, Hongxing Fan, Haohua Chen, Zehuan Huang, Lu Sheng |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 44 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-13488