HALO (ICML 2026 Spotlight): certified-stable RL keeps humanoids coordinated with unpredictable humans
Scobleizer · x · 2026-10-30
HALO, a human-robot collaboration paper from CMU and UT Arlington accepted as an ICML 2026 Spotlight/Oral, introduces Heterogeneous-Agent Lyapunov policy Optimization. It enforces a per-step Lyapunov decrease condition to stabilize decentralized policy learning against the rationality gap between robots and humans, with sim-to-real humanoid experiments showing better generalization. Scobleizer adds that robots will work alongside humans for decades rather than replace them.
More from Embodied
- Unitree Go2 and G1 robot dog firmware 0-day research slides published after coordinated disclosure — dyn___ · 2026-10-30
- Scoble: Robots repaving streets may be decades away, but construction data is gold — Scobleizer · 2026-10-30
- Robotics startup Machenta exits stealth with $40M seed, UK's largest-ever robotics seed round — Ed__Johns · 2026-10-30
- Stanford's I3L boosts whole-body robot deployment success by 68.3 points — jiajunwu_cs · 2026-10-30
- Standard Bots doc goes live: 7 years to sell its first 100 robots — Rewkang · 2026-10-30
- Neuralink ships 8 brain-interface features in 72 hours, lets ALS patient order coffee by thought — XFreeze · 2026-10-30