EVA: Bridging Performance and Human Alignment in Hard-Attention Vision Models for Image Classification

EVA: Bridging Performance and Human Alignment in Hard-Attention Vision Models for Image Classification | ResearchPod