Active Real-World Factor-Based Evaluation for Generalist Robot Policies
For robotics researchers, this provides a sample-efficient method to rigorously evaluate generalist policies, addressing the intractable combinatorial space of task factors.
The paper proposes an active evaluation framework for generalist robot policies that uses probabilistic surrogate models to adaptively select evaluation configurations, achieving 20-40% savings in trials compared to random testing across 2331 real-world evaluations.
Generalist robot manipulation policies trained on large, diverse datasets have shown remarkable promise across a wide range of tasks. However, rigorously evaluating these policies remains a fundamental challenge. Real-world performance depends on a large combinatorial space of task factors including object poses and camera viewpoints, making full, exhaustive evaluation intractable. Additionally, real hardware evaluation is slow and resource-intensive, so current practice is to use narrow test suites that can miss critical failure modes and misrepresent true deployment readiness. We propose an active evaluation framework that addresses this challenge by treating policy evaluation as a sequential experimental design problem. Our approach fits a probabilistic surrogate model over a structured space of task factors and adaptively selects evaluation configurations to maximize information gain over the policy's performance distribution, allowing for sample-efficient characterization of policy behavior across unseen conditions and a systematic identification of failure-prone regions. We conduct 2331 real-world evaluations across 3 tasks with 3 factor variations and find that our approach typically saves the evaluator at least 20-40% of trials compared to typical random testing.