@misc{indiciaed8b365c50f2c,
  title = {Geometric Disentanglement of Text Embeddings for Subject-Consistent Text-to-Image Generation using A Single Prompt},
  author = {Shangxun Li and Youngjung Uh},
  year = {2025},
  url = {https://arxiv.org/abs/2512.16443},
  note = {Source identifier: 2512.16443}
}