Q-Planning Enables Robot Self-Improvement, Boosting Success from 25% to 80%

Researchers from Georgia Tech and others proposed Q-Planning, which equips large vision-motor BC policies with a small off-policy learning harness, enabling self-improvement that raises robot task success rates from 25% to 80%.

2026-08-26 ~ 2026-08-27 · 2 related posts

• Q-Planning boosts robot self-improvement from 25% to 80% success — chris_j_paxton · 2026-08-26
• Q-Planning: robot policies self-improve from 25% to 80% success, no new demos — rbhar90 · 2026-08-27