File size: 2,540 Bytes
428c3e5 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 | ---
pipeline_tag: image-feature-extraction
---
# Image Tokenizer Needs Post-Training
This repository contains the official implementation and checkpoints for the paper [Image Tokenizer Needs Post-Training](https://huggingface.co/papers/2509.12474).
Project page: https://qiuk2.github.io/works/RobusTok/index.html
Code: https://github.com/qiuk2/RobusTok
<div align="center">
<img src="https://github.com/qiuk2/RobusTok/raw/main/assets/teaser.png" alt="Teaser" width="95%">
</div>
---
## TL;DR
We present RobusTok, a new image tokenizer with a two-stage training scheme:
Main training → constructs a robust latent space.
Post-training → aligns the generator’s latent distribution with its image space.
## Key highlights of Post-Training
- 🚀 **Better generative quality**: gFID 1.60 → 1.36.
- 🔑 **Generalizability**: applicable to both autoregressive & diffusion models.
- ⚡ **Efficiency**: strong results with only ~400M generative models.
---
## Model Zoo
| Generator \ Tokenizer | RobusTok w/o. P.T([weights](https://huggingface.co/qiuk6/RobusTok/resolve/main/main-train.pt?download=true)) | RobusTok w/. P.T ([weights](https://huggingface.co/qiuk6/RobusTok/resolve/main/post-train.pt?download=true)) |
|---|---:|---:|
| Base ([weights](https://huggingface.co/qiuk6/RobusTok/resolve/main/rar_b.bin?download=true)) | gFID = 1.83 | gFID = 1.60 |
| Large ([weights](https://huggingface.co/qiuk6/RobusTok/resolve/main/rar_l.bin?download=true)) | gFID = 1.60 | gFID = 1.36 |
---
## Usage
For detailed installation, training, and inference instructions, please refer to the [GitHub repository](https://github.com/qiuk2/RobusTok).
---
## Visualization
<div align="center">
<img src="https://github.com/qiuk2/RobusTok/raw/main/assets/ft-diff.png" alt="vis" width="95%">
<p>
visualization of 256×256 image generation before (top) and after (bottom) post-training. Three improvements are observed: (a) OOD mitigation, (b) Color fidelity, (c) detail refinement.
</p>
</div>
---
## Citation
If our work assists your research, feel free to give us a star ⭐ or cite us using
```bibtex
@misc{qiu2025imagetokenizerneedsposttraining,
title={Image Tokenizer Needs Post-Training},
author={Kai Qiu and Xiang Li and Hao Chen and Jason Kuen and Xiaohao Xu and Jiuxiang Gu and Yinyi Luo and Bhiksha Raj and Zhe Lin and Marios Savvides},
year={2025},
eprint={2509.12474},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2509.12474},
}
``` |