How VLAs (Really) Work In Open-World Environments

TL;DR

This paper critically analyzes vision-language-action models in robotics, highlighting limitations of current evaluation metrics and proposing new protocols to better assess safety and robustness in real-world scenarios.

Contribution

It provides a thorough evaluation of state-of-the-art VLAs on B1K, identifies safety and robustness issues, and introduces improved evaluation protocols for complex tasks.

Findings

Current metrics may overestimate performance by ignoring safety.

Models show inconsistent performance across different scenarios.

Proposed protocols better capture safety violations in complex environments.

Abstract

Vision-language-action models (VLAs) have been extensively used in robotics applications, achieving great success in various manipulation problems. More recently, VLAs have been used in long-horizon tasks and evaluated on benchmarks, such as BEHAVIOR1K (B1K), for solving complex household chores. The common metric for measuring progress in such benchmarks is success rate or partial score based on satisfaction of progress-agnostic criteria, meaning only the final states of the objects are considered, regardless of the events that lead to such states. In this paper, we argue that using such evaluation protocols say little about safety aspects of operation and can potentially exaggerate reported performance, undermining core challenges for future real-world deployment. To this end, we conduct a thorough analysis of state-of-the-art models on the B1K Challenge and evaluate policies in terms of robustness via reproducibility and consistency of performance, safety aspects of policies operations, task awareness, and key elements leading to the incompletion of tasks. We then propose evaluation protocols to capture safety violations to better measure the true performance of the policies in more complex and interactive scenarios. At the end, we discuss the limitations of the existing VLAs and motivate future research.