| title | MixTok: Adaptive Visual Tokenization with Global 1D Tokens and Sparse 2D Refinement |
|---|---|
| authors | Chenge Li, Yefei He, Xinyu Liu, Feng Chen, Hong Zhou, Bohan Zhuang |
| date | 2026-10-09 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2610.13181 |
| affiliation | Zhejiang University |
| upvotes | 7 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | Efficient visual tokenization requires balancing representation compactness with the preservation of fine-grained image content. Conventional grid-based tokenizers allocate tokens uniformly across spatial locations despite non-uniform visual complexity, whereas compact 1D tokenizers aggregate global information but can struggle to preserve local details. We introduce MixTok, a hybrid visual tokenizer that combines a compact 1D global representation with sparse 2D tokens for local refinement. A content-adaptive router ranks candidate locations by refinement priority and determines the number of 2D tokens for each image under a target average token budget. At the highest-scoring locations, 2D token features replace the spatial features recovered from the 1D representation, and a shared decoder reconstructs the image from the resulting mixed features. On ImageNet 256 x 256, MixTok achieves highly competitive reconstruction and generation performance, with an rFID of 1.195 and a gFID of 1.95 at respective average budgets of approximately 128 and 131 content tokens per image. |
cite
@misc{li2026mixtok,
title={MixTok: Adaptive Visual Tokenization with Global 1D Tokens and Sparse 2D Refinement},
author={Chenge Li and Yefei He and Xinyu Liu and Feng Chen and Hong Zhou and Bohan Zhuang},
year={2026},
eprint={2610.13181},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2610.13181}
}