How Do Large Vision-Language Models See Text in Image? Unveiling the Distinctive Role of OCR Heads.

Ingeol Baek, Hwan Chang, Sunghyun Ryu, Hwanhee Lee

Browse the full EMNLP paper archive.