@misc{indiciae739597cdfaaf,
  title = {Rethinking Contrastive Loss in CLIP Post-training: A Complementary Framework with Frozen Text Encoder},
  author = {Zidan Wang and Yaqian Li and Xiaokai Zhang and Kaiwen Long and Kun He and Hanpeng Liu},
  year = {2026},
  url = {https://arxiv.org/abs/2610.11374},
  note = {Source identifier: 2610.11374}
}