TULiP turns weight perturbations around a converged classifier into an OOD detection score, showing strong near-OOD performance on the OpenOOD benchmark.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
TULiP: Test-time Uncertainty Estimation via Linearization and Weight Perturbation
TULiP turns weight perturbations around a converged classifier into an OOD detection score, showing strong near-OOD performance on the OpenOOD benchmark.