How VLAs (Really) Work In Open-World Environments
For robotics researchers, this work highlights critical flaws in evaluation metrics that undermine real-world deployment safety, though it is an incremental critique rather than a novel solution.
The paper argues that current evaluation protocols for VLAs in robotics, which focus only on final success rates, ignore safety and exaggerate performance. The authors analyze SOTA models on BEHAVIOR1K, propose new safety-aware evaluation protocols, and discuss limitations of existing VLAs.
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.