Boosting the visual interpretability of CLIP via adversarial fine-tuning.

Shizhan Gong, Haoyu Lei, Qi Dou, Farzan Farnia

Browse the full ICLR paper archive.