Research map / Reinforcement learning / Visual and representation RL
BFM-Zero: A Promptable Behavioral Foundation Model for Humanoid Control Using Unsupervised Reinforcement Learning
Yitang Li, Zhengyi Luo, Tonghe Zhang, Cunxi Dai, Anssi Kanervisto, Andrea Tirinzoni, Haoyang Weng, Kris Kitani, Mateusz Guzek, Ahmed Touati, Alessandro Lazaric, Matteo Pirotta and 1 more
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| train with domain randomization and history-based actor instead of privileged no-DR model (ablation) | BFM-Zero · LAFAN1 tracking, reward, pose reaching (sim) | performance vs priv: 2.47%, 25.86%, 10.65% worse | Hurt |
| add CEM latent-prompt optimization for single pose with 4 kg payload | BFM-Zero single-leg stand · Real G1 with payload | balance duration: within 5 s -> over 15 s | Improved |
| add dual-annealing trajectory optimization over latent prompts | BFM-Zero tracking of leaping motion · Altered ground friction | tracking error: reducing error by ~29.1% | Improved |
| scale up residual network capacity (ablation) | BFM-Zero tracking · AMASS-trained datasets, test mocap | tracking error | Improved |
| use residual architectures instead of plain MLP (ablation) | BFM-Zero · tracking ablation | tracking error | Improved |
| include LAFAN1 motions in training data (ablation) | BFM-Zero reward inference · reward tasks | return | Improved |
| infer rewards from the replay buffer instead of the LAFAN1 motion dataset (ablation) | BFM-Zero reward inference · reward tasks | average return | Hurt |
| add auxiliary regularization rewards (ablation) | BFM-Zero · simulation metrics vs real robot | sim score / real stability | Mixed |
Closest papers by meaning
- Zero-Shot Whole-Body Humanoid Control via Behavioral Foundation Models (ICLR 2025 · similarity 0.90)
- Hierarchical World Models as Visual Whole-Body Humanoid Controllers (ICLR 2025 · similarity 0.88)
- HumanoidArena: Benchmarking Egocentric Hierarchical Whole-body Learning (NeurIPS 2026 · similarity 0.88)
- Humanoid Horizon: Extending Task Horizon in Whole-Body Loco-Manipulation via Parallel Training, Dynamic Starting, and Reward Gating (NeurIPS 2026 · similarity 0.87)
- Universal Humanoid Motion Representations for Physics-Based Control (ICLR 2024 · similarity 0.87)
- SUGAR: A Scalable Human-Video-Driven Generalizable Humanoid Loco-Manipulation Learning Framework (NeurIPS 2026 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
