SAFE: Scalable Failure Estimation for Vision-Language-Action ModelsPublished in Preprint, 2025Share on Twitter Facebook LinkedIn Previous Next