PRM-as-a-Judge: A Dense Evaluation Paradigm for Fine-Grained Robotic Auditing

TL;DR

This paper introduces PRM-as-a-Judge, a dense evaluation framework using Process Reward Models to assess robotic policy execution from videos, capturing detailed progress and diagnosing subtle behavioral issues beyond success/failure metrics.

Contribution

It proposes a novel dense evaluation paradigm with the OPD metric system, formalizes properties for evaluation, and empirically demonstrates superior micro-resolution in robotic auditing.

Findings

PRM judges outperform existing methods in micro-scale progress discrimination

The OPD metric system effectively formalizes execution quality

Structured audits reveal behavioral signatures and failure modes

Abstract

Current robotic evaluation is still largely dominated by binary success rates, which collapse rich execution processes into a single outcome and obscure critical qualities such as progress, efficiency, and stability. To address this limitation, we propose PRM-as-a-Judge, a dense evaluation paradigm that leverages Process Reward Models (PRMs) to audit policy execution directly from trajectory videos by estimating task progress from observation sequences. Central to this paradigm is the OPD (Outcome-Process-Diagnosis) metric system, which explicitly formalizes execution quality via a task-aligned progress potential. We characterize dense robotic evaluation through two axiomatic properties: macro-consistency, which requires additive and path-consistent aggregation, and micro-resolution, which requires sensitivity to fine-grained physical evolution. Under this formulation, potential-based PRM judges provide a natural instantiation of dense evaluation, with macro-consistency following directly from the induced scalar potential. We empirically validate the micro-resolution property using RoboPulse, a diagnostic benchmark specifically designed for probing micro-scale progress discrimination, where several trajectory-trained PRM judges outperform discriminative similarity-based methods and general-purpose foundation-model judges. Finally, leveraging PRM-as-a-Judge and the OPD metric system, we conduct a structured audit of mainstream policy paradigms across long-horizon tasks, revealing behavioral signatures and failure modes that are invisible to outcome-only metrics.