Scaling Behavior Foundation Model for Humanoid Robots

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Weishuai Zeng, Kangning Yin, Xiaojie Niu, Shunlin Lu, Weixiang Zhong, Jiahe Chen, Feiyu Jia, Xiao Chen, Zirui Wang, Furui Xu, Ming Zhou, Kailin Li, Weinan Zhang, He Wang, Li Yi, Dahua Lin, Jiangmiao Pang, Jingbo Wang
• Venue: arXiv: cs.RO
• Published: 2026-07-16
• Collected: 2026-07-17
• URL: http://arxiv.org/abs/2607.15163v1
• PDF: https://arxiv.org/pdf/2607.15163v1
• DOI: UNKNOWN

Plain-Language Brief

The research focuses on improving the performance of Behavior Foundation Models (BFMs) used for controlling humanoid robots. The authors propose a scaling recipe that involves coordinating three core components: 1) learning paradigm of motion tracking, 2) strategic synergy between on-policy rollout quantity and reference motion diversity, and 3) expressive and scalable model architecture called Humanoid Transformer. Extensive experiments show significant improvements in control fidelity and task generalization for both simulation and real-world deployment.

What the Document Says

• The research discusses the scaling of Behavior Foundation Models (BFMs) for humanoid robot control.
• Key factors for effective scaling include learning paradigm, behavioral data, and model architecture coordination.
• The authors propose a scaling recipe involving motion tracking learning paradigm, strategic synergy between on-policy rollout quantity and reference motion diversity, and Humanoid Transformer model architecture.

Key Concepts Mentioned

• Behavior Foundation Models (BFMs)
• Humanoid robot control
• Scaling
• Learning paradigm
• Motion tracking
• On-policy rollout quantity
• Reference motion diversity
• Humanoid Transformer model architecture

Claims Made by the Document

• Substantial performance gains can be achieved through the coordination of key components for BFM scaling.

Evidence Used

• Extensive experiments in both simulation and real-world deployment demonstrate significant improvements in control fidelity and task generalization.

Methods / Approach

• Extensive experimentation in simulation and real-world environments

Limitations Stated by the Authors

The paper does not discuss limitations of their proposed approach explicitly.

Limitations Noted by the Agent

Agent interpretation — not from the source.

• The experiments only demonstrate the results on specific humanoid robots and scenarios. Generalizability to other robots or contexts is unknown.

Open Questions

• How can the proposed method be extended to different types of robots?
• What are the potential limitations of the Humanoid Transformer model architecture?

Notes for Future Q&A

May include interpretation. Labeled where it does.

• It is important to consider that the authors only discuss scaling BFMs for humanoid robots. Other applications or types of robots may require different approaches.

Raw Abstract / Excerpt

Humanoid control requires natural whole-body coordination, precise real-time responses to control signals, and robust generalization across diverse environmental contexts, making it a cornerstone for generalist embodied agents. Behavior Foundation Models (BFMs) have recently emerged as a promising solution to address these challenges by leveraging large-scale behavioral data to achieve superior expressiveness, versatility and generalization. However, despite growing interest in scaling BFMs to further improve their capabilities, it remains unclear how key factors, including the learning paradigm, behavioral data and model architecture should be coordinated to enable effective scaling. In this work, we revisit the scaling recipe for BFMs and demonstrate that substantial performance gains can be achieved through the coordination of three core components: 1) the learning paradigm of motion tracking that reformulates diverse humanoid control problems as the reproduction of integrated whole-body behaviors in the global frame; 2) the strategic synergy between on-policy rollout quantity and reference motion diversity; and 3) the expressive and scalable model architecture termed Humanoid Transformer that facilitates the natural emergence of structured behavioral representations. Through extensive experiments in both simulation and real-world deployment, we demonstrate that our approach yields significant improvements in control fidelity and task generalization, reducing Mean Per-Keypoint Position Error (MPKPE) on the test set by over 10% in local mode and 82% in global mode compared with existing humanoid controllers. These results establish BFM as a principled and effective foundation for scalable and general-purpose humanoid control.