P-BFM

Model family: P-BFM is a perceptive behavioral foundation model for humanoid whole-body control on varied terrain. It separates terrain-independent intent from execution conditioned on geometry and explicit root instructions. The public implementation provides training and evaluation workflows.

Papers & reports (1)

P-BFM: Perceptive Behavioral Foundation Models for Humanoid Control via Unsupervised Reinforcement Learning

Paper available

arXiv: 2610.15833
Authors & affiliations (10)

Author order: Original order

  1. Xue Wang

    Not provided

  2. Yitang Li

    Not provided

  3. Yiming Zhang

    Not provided

  4. Siqin Wang

    Not provided

  5. Junhan Sun

    Not provided

  6. Zhihao Liu

    Not provided

  7. Yaojie Shen

    Not provided

  8. Jingbo Yuan

    Not provided

  9. Hao Zhao

    Not provided

First published
2026-10-11
Latest paper revision
Not provided
Venue
arXiv

Covered model versions

Paper ↗PDF ↗

How to access and use

Public code resources
Public access

Training / fine-tuning code

Access
Public access

Verified on · 2026-11-21

Code license
CC BY-NC 4.0
Release information

SourceVerified on · 2026-11-21

Checks before deployment

Available weights or code do not establish plug-and-play robot compatibility. Check the required configuration, data and interfaces in the source documentation.

Deployment and usage resources

Control and inputs / outputs

Model type
Motion & whole-body control
Input
Depth

Scope and training

Output

Actions

Robot form

Humanoid

Training

Reinforcement learning

Submit a source or correction