VCT: Training Consistency Models with Variational Noise Coupling
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913873970331648 |
|---|---|
| author | Silvestri, Gianluigi Ambrogioni, Luca Lai, Chieh-Hsin Takida, Yuhta Mitsufuji, Yuki |
| author_facet | Silvestri, Gianluigi Ambrogioni, Luca Lai, Chieh-Hsin Takida, Yuhta Mitsufuji, Yuki |
| contents | Consistency Training (CT) has recently emerged as a strong alternative to diffusion models for image generation. However, non-distillation CT often suffers from high variance and instability, motivating ongoing research into its training dynamics. We propose Variational Consistency Training (VCT), a flexible and effective framework compatible with various forward kernels, including those in flow matching. Its key innovation is a learned noise-data coupling scheme inspired by Variational Autoencoders, where a data-dependent encoder models noise emission. This enables VCT to adaptively learn noise-todata pairings, reducing training variance relative to the fixed, unsorted pairings in classical CT. Experiments on multiple image datasets demonstrate significant improvements: our method surpasses baselines, achieves state-of-the-art FID among non-distillation CT approaches on CIFAR-10, and matches SoTA performance on ImageNet 64 x 64 with only two sampling steps. Code is available at https://github.com/sony/vct. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_18197 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | VCT: Training Consistency Models with Variational Noise Coupling Silvestri, Gianluigi Ambrogioni, Luca Lai, Chieh-Hsin Takida, Yuhta Mitsufuji, Yuki Machine Learning Computer Vision and Pattern Recognition Consistency Training (CT) has recently emerged as a strong alternative to diffusion models for image generation. However, non-distillation CT often suffers from high variance and instability, motivating ongoing research into its training dynamics. We propose Variational Consistency Training (VCT), a flexible and effective framework compatible with various forward kernels, including those in flow matching. Its key innovation is a learned noise-data coupling scheme inspired by Variational Autoencoders, where a data-dependent encoder models noise emission. This enables VCT to adaptively learn noise-todata pairings, reducing training variance relative to the fixed, unsorted pairings in classical CT. Experiments on multiple image datasets demonstrate significant improvements: our method surpasses baselines, achieves state-of-the-art FID among non-distillation CT approaches on CIFAR-10, and matches SoTA performance on ImageNet 64 x 64 with only two sampling steps. Code is available at https://github.com/sony/vct. |
| title | VCT: Training Consistency Models with Variational Noise Coupling |
| topic | Machine Learning Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2502.18197 |