@misc{indiciaebf7a0390f6e7,
  title = {Geometry-Aware CLIP Retrieval via Local Cross-Modal Alignment and Steering},
  author = {Nirmalendu Prakash and Narmeen Fatimah Oozeer and Xin Su and Phillip Howard and Shaan Shah and Zoe Wanying He and Shuang Wu and Shivam Raval and Roy Ka-Wei Lee and Meenakshi Khosla and Amir Abdullah},
  year = {2026},
  url = {https://arxiv.org/abs/2604.16487},
  note = {Source identifier: 2604.16487}
}