## 🏃‍➡️ Coming Soon Thank you for your interest in our project! We are currently preparing and testing the code and will be releasing it soon. ## What will be Included This repository contains the code implementation of the experiments presented in the paper [*Beyond Masked and Unmasked: Discrete Diffusion Models via Partial Masking*](https://arxiv.org/abs/2505.18495). - :whale: **Docker support** for enhanced reproducibility - :cd: **Pretrained weights** for inference and evaluation - :mag: **Sampling process visualizations** to understand model behavior - :microscope: **Code for all experiments** in our paper: - Toy experiments on synthetic data - Text generation on OpenWebText - Image generation on CIFAR-10 & ImageNet-32 ## Overview ### Toy Examples - **Dataset**: 2D Synthetic Dataset - **Folder**: [mdm-prime/toy](/toy) -

Demo (click me)

### Text Generation - **Dataset**: OpenWebText (OWT) - **Folder**: [mdm-prime/text](/text) -

Demo (click me)

### Image Generation - **Dataset**: CIFAR-10, ImageNet-32 - **Folder**: [mdm-prime/image](/image) -

Demo (click me)

## License This code implementation is developed based on the following repositories. - [kuleshov-group/mdlm](https://github.com/kuleshov-group/mdlm) (at commit `3ecb6dc`), licensed under the `Apache-2.0` license. - [facebookresearch/flow_matching](https://github.com/facebookresearch/flow_matching) (at commit `47c4396`), licensed under the `CC BY-NC 4.0` license. Further changes based on this repository are licensed under the `Apache-2.0` and `CC BY-NC 4.0` licenses. ## Citing MDM-Prime If you find our code useful, please consider citing our paper. ```bib @article{chao2025mdmprime, title={{Beyond Masked and Unmasked: Discrete Diffusion Models via Partial Masking}}, author={Chen-Hao Chao, Wei-Fang Sun, Hanwen Liang, Chun-Yi Lee, Rahul G. Krishnan}, journal={\tt arXiv:2505.18495 [cs.LG]}, year={2025}, } ```

Openwebtext

Topic Statistics

Related Topics