Robot policies are being rebuilt around control-loop evidence
Robot learning is being judged inside the control loop. The strongest papers add future-change priors, drift monitors, critics, world-model rollouts, and cheaper motion data to make vision-language-action (VLA) policies…