WAM-Policy Brings Language-Conditioned Loco-Manipulation to Arm-Free Quadruped Robots
Maincode · hf · 2026-11-20
Researchers from USYD, UNSW, Maincode and La Trobe present WAM-Policy, the first language-conditioned policy enabling arm-free quadrupeds to perform loco-manipulation—walking on three legs while manipulating with a front leg.
- Two stages: Cross-Morphology Policy Adaptation and Acceleration adapts an arm-centric world action model (WAM) to quadrupeds and distills it into a student (one video + two action steps) via on-policy self-distillation (OPSD); then Task-Specific RL plus multi-teacher on-policy distillation (M-OPD) merges slow specialists into one few-step student.
- On QLM-Bench (10 simulated tasks, 1,000 demonstrations synthesized without teleoperation): transferred policy 77.0%, OPSD student keeps 76.7% while sampling 14x faster, RL+M-OPD lifts it to 85.0%.
- Deployed on a real Unitree Go2 Air at a fraction of the inference cost.
More from Embodied
- AI writes all code to make a real robotic arm play chess, open-sourced — FinanceYF5 · 2026-11-20
- KAIST's VSpark Speeds Up Autoregressive VLA Robot Models 2.1x with Speculative Decoding — kaist-ai · 2026-11-20
- RoboColosseum Aims to Benchmark Robot Policies in the Real World at Unprecedented Scale — DJiafei · 2026-11-20
- Dev Tours Seeed Studio's Shenzhen Factory: PCBA Lines and Robot Arm Assembly Up Close — haydendevs · 2026-11-20
- Musk says Optimus robotic hands could be given to people in need — CyberRobooo · 2026-11-20
- Zeromatter opens its pre-name pitch deck 3.5 years after pitching simulation infra for physical intelligence — soleio · 2026-11-20