Modeling and Understanding Neural Weights of Humanoid Policies with Diffusion

原文：Modeling and Understanding Neural Weights of Humanoid Policies with Diffusion

作者：Huixuan Cheng, Bike Zhang, Shankar Sastry, Koushil Sreenath

来源：arXiv cs.AI（人工智能）

正文

Computer Science > Robotics

arXiv:2610.20606v1 (cs)

[Submitted on 14 Oct 2026]

Title:Modeling and Understanding Neural Weights of Humanoid Policies with Diffusion

Authors:Huixuan Cheng, Bike Zhang, Shankar Sastry, Koushil Sreenath

View a PDF of the paper titled Modeling and Understanding Neural Weights of Humanoid Policies with Diffusion, by Huixuan Cheng and 3 other authors

View PDF

HTML (experimental)

Abstract:Recent advances in robot learning have driven rapid progress in humanoid control, enabling increasingly agile behaviors in the real world. However, humanoid control networks are typically obtained through conventional training procedures, while comparatively little attention has been given to directly modeling the distribution of their learned weights and generating functional networks in weight space. In this work, we investigate whether high-performing humanoid control networks can be generatively modeled directly through their weights. We adopt a latent diffusion framework that couples a linear autoencoder with a Diffusion Transformer. The autoencoder compresses actor network parameters into a compact latent space, and the Diffusion Transformer models the distribution of these latents. Trained on networks collected from reinforcement learning for the Unitree G1 robot, our approach generates deployable networks that perform various motions in the real world, including walking, dancing, and spin kick. Beyond generation, we further study the learned weight space to better understand the structure captured by these humanoid control networks. We identify a structural principle in the first-layer weights that consistently reflects the robot’s kinematic prior. Notably, this kinematic prior emerges under both Reinforcement Learning (RL) and Imitation Learning (IL) training, suggesting that it is not specific to a particular optimization process but may instead reflect common knowledge associated with humanoid embodiment. Together, these results demonstrate the feasibility of generative modeling in humanoid network weight space, while providing initial insights into the information encoded by their learned weights. Additional results are available on our project website: this https URL

Comments:

Project page: this https URL

Subjects:

Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)

Cite as:

arXiv:2610.20606 [cs.RO]

(or

arXiv:2610.20606v1 [cs.RO] for this version)

https://doi.org/10.48550/arXiv.2610.20606

Focus to learn more

arXiv-issued DOI via DataCite (pending registration)

主题

机器人 · 人工智能 · 机器学习

由「前沿雷达」于 2026-10-15 采集。正文取自原文页面，已保留出处链接。标题与正文版权归原作者所有。