[NeurIPS'25] Learnable Sampler Distillation for Discrete Diffusion Models

This repository officially houses the PyTorch implementation of the paper titled "Learnable Sampler Distillation for Discrete Diffusion Moldes", which is presented at NeurIPS 2025 as a spotlight.

Project Roadmap

To-Do

• Release training and evaluation code.
• Release code on more baselines.
• Remove redundant code.

Completed

• Release raw base code and core algorithm implementation.

Installation

Please refer to JYS for Installation.

References

This repository is heavily based on

• CTMC
• SEDD
• JYS