README.md

---
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&times;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}, 
}
```