Focusing by Contrastive Attention: Enhancing VLMs' Visual Reasoning

Publication
arXiv preprint arXiv:2509.06461