PRM-as-a-Judge 1.5 Toolkit for Fine-Grained Robot Process Assessment
August 13, 2026
This toolkit converts robotic rollout videos into dense progress curves to provide metrics beyond binary success rates. It introduces three new metrics to characterize failure-side progress, post-drawdown recovery, and success-side execution quality for embodied models.
HOW THIS AFFECTS YOU
●
builderThis provides more granular telemetry for debugging robotic control policies.
●
researcherYou can move beyond binary success metrics to analyze specific failure modes in embodied AI.