This repository contains the weights for Longcat-Image-Turbo, a few-step distilled version of Longcat-Image using the Continuous-Time Distribution Matching (CDM) method presented in Continuous-Time Distribution Matching for Few-Step Diffusion Distillation.
CDM migrates the Distribution Matching Distillation (DMD) framework from discrete anchoring to continuous optimization, allowing for high-quality image generation with very few steps (e.g., 4 NFE).
Algorithm Overview
Overview of Continuous-Time Distribution Matching (CDM). Top: Our approach employs a dynamic continuous time schedule during backward simulation, sampling intermediate anchors uniformly from (0, 1]. Bottom Left: CFG augmentation (CA) and distribution matching (DM) operate on this dynamic schedule to align text-image conditions and data distributions at on-trajectory anchors. Bottom Right: To address inter-anchor inconsistency, the proposed CDM objective explicitly extrapolates off-trajectory latents using the predicted velocity.
4-NFE Generation Results
SD3-Medium
LongCat
Inference
To use this model, please refer to the GitHub repository.
# Clone this repository
git clone https://github.com/byliutao/cdm.git
cd cdm
# Create and activate the inference environment
conda create -n cdm_infer python=3.10
conda activate cdm_infer
pip install -r config/requirements_infer.txt
# Run inference
python scripts/infer/longcat.py # LongCat
Training
# Create and activate the training environment
conda create -n cdm_train python=3.10
conda activate cdm_train
pip install -r config/requirements_train.txt
# Launch training with FSDP2
accelerate launch --config_file config/accelerate_fsdp2.yaml \
--num_processes 8 -m scripts.train \
--config config/config.py:longcat # LongCat
License
This project is licensed under the MIT License — see the LICENSE file for details.
Citation
If our work assists your research, please consider giving us a star ⭐ or citing us:
@misc{liu2026continuoustimedistributionmatchingfewstep,
title={Continuous-Time Distribution Matching for Few-Step Diffusion Distillation},
author={Tao Liu and Hao Yan and Mengting Chen and Taihang Hu and Zhengrong Yue and Zihao Pan and Jinsong Lan and Xiaoyong Zhu and Ming-Ming Cheng and Bo Zheng and Yaxing Wang},
year={2026},
eprint={2605.06376},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2605.06376},
}