README.md

metadata

license: other
license_name: research-and-educational-use-license
license_link: https://github.com/SNU-VGILab/improvedSelfDistillation/blob/main/LICENSE

improved Self-Distillation

Pretrained weights and evaluation assets for Stabilizing Consistency Training: A Flow Map Analysis and Self-Distillation.

Code and instructions are available in the GitHub repository.

Contents

• outputs/: pretrained checkpoints
• buffers/vaes/: VAE checkpoints and latent statistics
• buffers/refs/: reference files for FID evaluation

Checkpoint	Network	Steps	FID50K
2026.02.15KST14.22.08-base4	FlowMapTiT-B/4 (SD-VAE, TrigFlow)	400K	14.58
2026.01.18KST19.26.11-xlarge1	ADiT-XL/1 (VA-VAE, Linear)	600K	2.30

Usage

Place the downloaded outputs and buffers directories at the top level of the code repository, then run the provided training or evaluation scripts.

bash eval.sh

Citation

@misc{kim2026stabilizingconsistencytrainingflow,
      title={Stabilizing Consistency Training: A Flow Map Analysis and Self-Distillation},
      author={Youngjoong Kim and Duhoe Kim and Woosung Kim and Jaesik Park},
      year={2026},
      eprint={2601.22679},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2601.22679},
}