aiqbench analysis
PartiGen advances humanoid motion generation by integrating anatomical structure into latent diffusion models, enabling more precise text-to-motion translation for robots like the Unitree G1. The framework demonstrates significant improvements in kinematic consistency and execution success rates through part-specific attention and autoregressive primitive generation. This work represents a meaningful step forward in aligning generative AI with the physical constraints of robotic hardware.
PartiGen: Part-Aware Representations for Language-Conditioned Humanoid Motion Generation
arXiv:2610.13373v1 Announce Type: new Abstract: Translating language into humanoid motion requires coordinating body parts, determining motion duration, and producing references suitable for robot execution. We introduce PartiGen, a part-aware latent diffusion framework that incorporates anatomical structure into a shared whole-body motion representation. Part-specific attention processes anatomically masked frame features, and its outputs are fused to combine local limb information with whole-body context. Conditioned on text and motion history, the model generates motion primitives autoregressively. Forward-kinematics supervision encourages kinematic consistency, while a learned end-of-sequence predictor selects a frame-level endpoint after generation. On HumanML3D retargeted to the Unitree G1, PartiGen achieves a segment-level FID of 3.165, a 16.8% reduction over TextOp, and improves R@1 by 1.4 percentage points. In the duration study with 16-frame primitives, endpoint prediction reduces mean absolute duration error from 12.471 to 3.080 frames compared with a fixed-length cutoff. We further evaluate robot execution using a pretrained, frozen GRIT tracker: 409 of 500 generated references (81.8%) pass the posture-based stability screen in MuJoCo. Following simulation screening and manual selection, 26 of 27 attempted motions are judged successful on the real G1.
Abstract:Translating language into humanoid motion requires coordinating body parts, determining motion duration, and producing references suitable for robot execution. We introduce PartiGen, a part-aware latent diffusion framework that incorporates anatomical structure into a shared whole-body motion representation. Part-specific attention processes anatomically masked frame features, and its outputs are fused to combine local limb information with whole-body context. Conditioned on text and motion history, the model generates motion primitives autoregressively. Forward-kinematics supervision encourages kinematic consistency, while a learned end-of-sequence predictor selects a frame-level endpoint after generation. On HumanML3D retargeted to the Unitree G1, PartiGen achieves a segment-level FID of 3.165, a 16.8% reduction over TextOp, and improves R@1 by 1.4 percentage points. In the duration study with 16-frame primitives, endpoint prediction reduces mean absolute duration error from 12.471 to 3.080 frames compared with a fixed-length cutoff. We further evaluate robot execution using a pretrained, frozen GRIT tracker: 409 of 500 generated references (81.8%) pass the posture-based stability screen in MuJoCo. Following simulation screening and manual selection, 26 of 27 attempted motions are judged successful on the real G1.
Robotics (cs.RO)
arXiv:2610.13373 [cs.RO]
(or arXiv:2610.13373v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2610.13373 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 270 words
Related coverage
Dyna-2: General Purpose Physical Agents with Hour-Long Autonomy — Jason Ma, Dyna Robotics
Latent Space · Oct 13, 2026
Streaming Multi-Track Timeline Control for 3D Human Motion Generation
arXiv · Oct 7, 2026
J++ Lens: Jacobian Filtering Enables More Faithful Workspace Lenses
LessWrong · Oct 8, 2026
LiveLaw Supreme Court Half-Yearly Digest: January - June 2026 [Part 2]
livelaw.in · Sep 13, 2026