Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

Scaling Behavior Foundation Model for Humanoid Robots

Source

arXiv

arxiv.org/abs/2607.15163v1

Read the source

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

Researchers propose a scaling recipe for Behavior Foundation Models (BFMs) to improve the capabilities of humanoid robots. The recipe includes a learning paradigm, strategic synergy between data collection methods, and an expressive model architecture called Humanoid Transformer. Experiments show significant improvements in control fidelity and task generalization.

What the Document Says

  • The text discusses a study on scaling Behavior Foundation Models (BFMs) for humanoid robots.
  • Humanoid control requires natural whole-body coordination, precise real-time responses, and robust generalization across diverse environments.
  • BFMs have shown promise in achieving superior expressiveness, versatility, and generalization by leveraging large-scale behavioral data.
  • However, the optimal coordination of key factors for effective scaling remains unclear.
  • The study presents a scaling recipe with three core components: motion tracking learning paradigm, strategic synergy between rollout quantity and reference motion diversity, and Humanoid Transformer model architecture.
  • Humanoid Transformer facilitates the natural emergence of structured behavioral representations.
  • Extensive experiments in simulation and real-world deployment demonstrate significant improvements in control fidelity and task generalization.
  • The results reduce Mean Per-Keypoint Position Error (MPKPE) on the test set by over 10% in local mode and 82% in global mode compared to existing humanoid controllers.
  • The study concludes that BFMs provide a principled and effective foundation for scalable and general-purpose humanoid control.

Key Concepts Mentioned

  • Humanoid robots
  • Behavior Foundation Models (BFMs)
  • Scaling recipe
  • Learning paradigm
  • Motion tracking
  • On-policy rollout quantity
  • Reference motion diversity
  • Humanoid Transformer
  • Control fidelity
  • Task generalization
  • Mean Per-Keypoint Position Error (MPKPE)

Claims Made by the Document

  • BFMs can improve the capabilities of humanoid robots
  • The presented scaling recipe can achieve substantial performance gains
  • The approach yields significant improvements in control fidelity and task generalization

Evidence Used

  • extensive experiments in simulation and real-world deployment
  • reduction of MPKPE by over 10% in local mode and 82% in global mode compared to existing humanoid controllers

Methods / Approach

  • motion tracking learning paradigm
  • strategic synergy between on-policy rollout quantity and reference motion diversity
  • Humanoid Transformer model architecture

Limitations Stated by the Authors

The text does not mention any limitations as stated by the authors.

Limitations Noted by the Agent

Agent interpretation — not from the source.

The data provided does not allow me to determine if there are any inherent limitations in the study's methodology or results.

Open Questions

  • What specific improvements were made to existing humanoid controllers?
  • What other factors could impact the scaling of BFMs beyond those addressed in this study?

Notes for Future Q&A

May include interpretation. Labeled where it does.

The text does not provide details on the specific improvements made to existing humanoid controllers or discuss potential factors that might affect the scaling of BFMs.

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.