Taming False Positives in Out-of-Distribution Detection with Human Feedback

1 code implementation • 25 Apr 2024 • Harit Vishwakarma, Heguang Lin, Ramya Korlakai Vinayak

Empirical evaluation of our system on synthetic and benchmark OOD datasets shows that our method can maintain FPR at most $5\%$ while maximizing TPR.

Medical Diagnosis Out-of-Distribution Detection +1

Promises and Pitfalls of Threshold-based Auto-labeling

2 code implementations • NeurIPS 2023 • Harit Vishwakarma, Heguang Lin, Frederic Sala, Ramya Korlakai Vinayak

Given the long shelf-life and diverse usage of the resulting datasets, understanding when the data obtained by such auto-labeling systems can be relied on is crucial.

Geometry of the Minimum Volume Confidence Sets

no code implementations • 16 Feb 2022 • Heguang Lin, Mengze Li, Daniel Pimentel-Alarcón, Matthew Malloy

Prior work showed the minimum-volume confidence sets are the level-sets of a discontinuous function defined by an exact p-value.