robotics
The Robot Policy Watches Its Own Progress
Linear probes trained on internal activations flag executions that depart from the progress pattern seen in successful demonstrations.
Summary
Linear probes trained on internal activations flag executions that depart from the progress pattern seen in successful demonstrations.
Runtime monitors for vision-language-action policies often add another foundation model to watch the first. Progress Probes instead trains lightweight linear probes on the policy's own activations from demonstrations. The monitor estimates how far a task should have advanced and flags trajectories that diverge from that expected progress. In simulated and real manipulation tests with the open-source pi-0.5 policy, the probes matched or exceeded vision-language-model monitors and approaches that require failure examples. The evidence covers the tested policy and tasks; internal progress signals do not guarantee detection of every failure mode.
Why it matters
Linear probes trained on internal activations flag executions that depart from the progress pattern seen in successful demonstrations.
Limits and context
- The evidence covers the tested policy and tasks; internal progress signals do not guarantee detection of every failure mode.
Key claims
Linear probes trained on internal activations flag executions that depart from the progress pattern seen in successful demonstrations.
Qualification: The evidence covers the tested policy and tasks; internal progress signals do not guarantee detection of every failure mode.
Evidence: source-2026-10-23-014
Sources
- arXiv preprint 2610.31613, original v1arXiv · primary research
Corrections
No corrections have been recorded for this story.