RUAC adds an uncertainty head to SAM, trained with collaborative style-deformation attacks and Uncertainty-Accuracy Alignment, yielding stronger uncertainty-accuracy correlation across 23 zero-shot domains.
Since calibration metrics (AUROC, PAvPU) de- pend only on therankingof uncertainty values, not their absolute magnitudes, the approximation preserves all task-relevant information
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Segment Anything with Robust Uncertainty-Accuracy Correlation
RUAC adds an uncertainty head to SAM, trained with collaborative style-deformation attacks and Uncertainty-Accuracy Alignment, yielding stronger uncertainty-accuracy correlation across 23 zero-shot domains.