Test-Time Compute for Humanoid Behavior Foundation Models

arXiv:2610.14896 · cs.RO · Submitted 2026-10-10 · Read on arXiv

cs.RO

Submitted: 2026-10-10

Updated: 2026-10-14

Code: https://github.com/google-deepmind/mujoco_warp

Project page: https://bfmttc.github.io

Terminology

• Foundation Model — 1× in this paper · explained in 3 episodes
• Foundation Models — 1× in this paper · explained in 7 episodes

Sources

• GR00T N1: An Open Foundation Model for Generalist Humanoid Robots
• Large Language Monkeys: Scaling Inference Compute with Repeated Sampling
• TEXEDO : Test Time Scaling for Controller-aware Language-conditioned Humanoid Motion Generation
• HoloMotion-1 Technical Report
• KungfuBot2: Learning Versatile Motion Skills for Humanoid Whole-Body Control
• Predictive Sampling: Real-time Behaviour Synthesis with MuJoCo
• OpenHLM: An Empirical Recipe for Whole-Body Humanoid Loco-Manipulation
• An Introduction to Zero-Order Optimization Techniques for Robotics
• ELASTIC: Efficiently Learning to Adaptively Scale Test-Time Compute for Generative Control Policies
• OmniClone: Engineering a Robust, All-Rounder Whole-Body Humanoid Teleoperation System
• HEFT: Heavy-Payload Full-size Humanoid Teleoperation with Privileged Motion Guidance and Windowed Payload Curriculum
• Enabling Robust Cloth Manipulation via Inference-Time Simulator-in-the-Loop Refinement
• Robust and Generalized Humanoid Motion Tracking
• SPIDER: Scalable Physics-Informed Dexterous Retargeting
• Consensus-based optimization (CBO): Towards Global Optimality in Robotics
• HumanoidArena: Benchmarking Egocentric Hierarchical Whole-body Learning
• VLK: Learning Humanoid Loco-Manipulation from Synthetic Interactions in Reconstructed Scenes
• OmniXtreme: Breaking the Generality Barrier in High-Dynamic Humanoid Control
• Teleopit: A Full-Embodiment Humanoid Teleoperation System
• GRAIL: Generating Humanoid Loco-Manipulation from 3D Assets and Video Priors

Related papers

• FMT x: An Efficient and Asymptotically Optimal Extension of the Fast Marching Tree for Dynamic Replanning
• MPCFormer: A physics-informed data-driven approach for explainable socially-aware autonomous driving
• RoboLab: A High-Fidelity Simulation Benchmark for Analysis of Task Generalist Policies
• ARC: A Reasoning Recipe for Robot Foundation Models
• HRDexDB: A 4D Dexterous Grasping Dataset Across Human and Multiple Robot Embodiments
• APT: Action Expert Pretraining Improves Instruction Generalization of Vision-Language-Action Policies