pith. sign in

To minimize the expected NLL term Ew[NLL]≈NLL(µ) +1 2Tr(H(Σ)) , the optimizer is forced to reduce Σ significantly

Standard Training (Sharp Minima):Standard ERM often converges to sharp minima where the Hessian Hw has large eigenvalues (high curvature)

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it
browse 1 citing papers

fields

cs.CV 1

years

2026 1

verdicts

UNVERDICTED 1

representative citing papers

Segment Anything with Robust Uncertainty-Accuracy Correlation

cs.CV · 2026-05-11 · unverdicted · novelty 5.0 · 3 refs

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.

citing papers explorer

Showing 1 of 1 citing paper.

  • Segment Anything with Robust Uncertainty-Accuracy Correlation cs.CV · 2026-05-11 · unverdicted · none · ref 14 · 3 links

    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.