VLAConf is a one-class discriminative method that estimates step-wise task-success confidence for VLA models via anomaly scoring on frozen representations plus step-conditioned modeling, shown to be more efficient than ensemble or probability baselines on LIBERO and real robots.
Richards, Yixiao Sun, Edward Schmerling, and Marco Pavone
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2026 2representative citing papers
OpenVLA layer-16 activations allow a logistic probe to predict failure within 15 steps under occlusion (AUROC 0.972) better than baselines, with some transfer to camera jitter.
citing papers explorer
-
VLAConf: Calibrated Task-Success Confidence for Vision-Language-Action Models
VLAConf is a one-class discriminative method that estimates step-wise task-success confidence for VLA models via anomaly scoring on frozen representations plus step-conditioned modeling, shown to be more efficient than ensemble or probability baselines on LIBERO and real robots.
-
Early Warning Signals for OpenVLA Failure under Visual Distribution Shift
OpenVLA layer-16 activations allow a logistic probe to predict failure within 15 steps under occlusion (AUROC 0.972) better than baselines, with some transfer to camera jitter.