arxivcs.LGcs.AIcs.CV2026-06-29
Exploiting Local Flatness for Efficient Out-of-Distribution Detection
Seonghwan Park, Hyunji Jung, Dongyeop Lee, Namhoon Lee
Detecting out-of-distribution (OOD) data is crucial for reliable machine learning deployment. Among detection strategies, post-hoc methods are particularly attractive due to their efficiency, as they operate directly on pre-trained networks without requiring retraining. Within th…