@misc{indiciae0da1c1c976e4,
  title = {Attributes Should Come from Images, Not Class Names: Distribution-Conditioned Attribute Selection for Vision-Language Models},
  author = {Gautam Rajendrakumar Gare and Jia Shi and Zhiqiu Lin and Deepak Pathak and John Galeotti and Deva Ramanan},
  year = {2026},
  url = {https://arxiv.org/abs/2607.18695},
  note = {Source identifier: 2607.18695}
}