HoloMotion-1 Technical Report

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Chen, Maiyue, Wang, Kaihui, Zhang, Bo, Ma, Xihan, Yang, Zhiyuan, Ren, Yi, Huang, Qijun, Zhu, Zihao, Wang, Yucheng, Su, Zhizhong
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866917509158928384
author Chen, Maiyue
Wang, Kaihui
Zhang, Bo
Ma, Xihan
Yang, Zhiyuan
Ren, Yi
Huang, Qijun
Zhu, Zihao
Wang, Yucheng
Su, Zhizhong
author_facet Chen, Maiyue
Wang, Kaihui
Zhang, Bo
Ma, Xihan
Yang, Zhiyuan
Ren, Yi
Huang, Qijun
Zhu, Zihao
Wang, Yucheng
Su, Zhizhong
contents In this report, we present HoloMotion-1, a humanoid motion foundation model for zero-shot whole-body motion tracking. A key innovation of HoloMotion-1 is to scale control-policy training with a large-scale hybrid motion corpus, where video-reconstructed motions from in-the-wild videos provide the dominant source of motion diversity, while curated motion-capture and in-house motion data provide higher-fidelity supervision and deployment-oriented coverage. This data regime enables HoloMotion-1 to move beyond conventional MoCap-only training and exposes the policy to substantially broader behaviors, capture conditions, and motion styles. Learning from such heterogeneous data introduces new challenges, including reconstruction noise, source-domain mismatch, uneven motion quality, and the need for temporal modeling under large behavioral variation. To address these challenges, HoloMotion-1 integrates large-capacity temporal modeling, a sparsely activated Mixture-of-Experts Transformer with KV-cache inference for real-time control, and a sequence-level training strategy that improves learning efficiency on extended motion sequences. Extensive experiments on multiple unseen motion benchmarks show that HoloMotion-1 generalizes robustly across diverse motion types and capture conditions, significantly improves tracking accuracy over prior methods, and transfers directly to a real humanoid robot without task-specific fine-tuning.
format Preprint
id arxiv_https___arxiv_org_abs_2605_15336
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle HoloMotion-1 Technical Report
Chen, Maiyue
Wang, Kaihui
Zhang, Bo
Ma, Xihan
Yang, Zhiyuan
Ren, Yi
Huang, Qijun
Zhu, Zihao
Wang, Yucheng
Su, Zhizhong
Robotics
Artificial Intelligence
In this report, we present HoloMotion-1, a humanoid motion foundation model for zero-shot whole-body motion tracking. A key innovation of HoloMotion-1 is to scale control-policy training with a large-scale hybrid motion corpus, where video-reconstructed motions from in-the-wild videos provide the dominant source of motion diversity, while curated motion-capture and in-house motion data provide higher-fidelity supervision and deployment-oriented coverage. This data regime enables HoloMotion-1 to move beyond conventional MoCap-only training and exposes the policy to substantially broader behaviors, capture conditions, and motion styles. Learning from such heterogeneous data introduces new challenges, including reconstruction noise, source-domain mismatch, uneven motion quality, and the need for temporal modeling under large behavioral variation. To address these challenges, HoloMotion-1 integrates large-capacity temporal modeling, a sparsely activated Mixture-of-Experts Transformer with KV-cache inference for real-time control, and a sequence-level training strategy that improves learning efficiency on extended motion sequences. Extensive experiments on multiple unseen motion benchmarks show that HoloMotion-1 generalizes robustly across diverse motion types and capture conditions, significantly improves tracking accuracy over prior methods, and transfers directly to a real humanoid robot without task-specific fine-tuning.
title HoloMotion-1 Technical Report
topic Robotics
Artificial Intelligence
url https://arxiv.org/abs/2605.15336