Marco Donato

Parameter-Efficient Dual-Loss Adaptation with Logit Divergence: A Unified Approach for Adversarial Example Detection and Robust Inference