@misc{indiciaefda27ccd390a,
  title = {Composable Visual Tokenizers with Generator-Free Diagnostics of Learnability},
  author = {Bingchen Zhao and Qiushan Guo and Ye Wang and Yixuan Huang and Zhonghua Zhai and Yu Tian},
  year = {2026},
  url = {https://arxiv.org/abs/2602.03339},
  note = {Source identifier: 2602.03339}
}