@misc{indiciae51d1858fa477,
  title = {Not All Attention Heads Are What You Need: Refining CLIP's Image Representation with Attention Ablation},
  author = {Feng Lin and Marco Chen and Haokui Zhang and Xiaotian Yu and Guangming Lu and Rong Xiao},
  year = {2025},
  url = {https://arxiv.org/abs/2507.00537},
  note = {Source identifier: 2507.00537}
}