Benchmark RadarAI BENCHMARK PROFILE
PRM-as-a-Judge 1.5
A toolkit that converts robotic rollout videos into progress curves and computes metrics for failure progress, recovery, and execution quality, with a benchmark and evaluation suite for process reward models.
- Released
- 2026-08-14
- Readiness
- Paper only
- Primary field
- Robotics & Autonomous Systems
Why it matters
Supplies fine-grained process assessment beyond binary success, enabling more transparent and procedural evaluation of embodied models.
Motivation
Fine-grained robotic evaluation matters for understanding embodied models, going beyond binary success rates and rule-based process scores.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.