Salvato in:
| Autori principali: | , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2509.19300 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866917038430093312 |
|---|---|
| author | Chen, Chen Guo, Pengsheng Song, Liangchen Lu, Jiasen Qian, Rui Wang, Xinze Fu, Tsu-Jui Liu, Wei Yang, Yinfei Schwing, Alex |
| author_facet | Chen, Chen Guo, Pengsheng Song, Liangchen Lu, Jiasen Qian, Rui Wang, Xinze Fu, Tsu-Jui Liu, Wei Yang, Yinfei Schwing, Alex |
| contents | Conditional generative modeling aims to learn a conditional data distribution from samples containing data-condition pairs. For this, diffusion and flow-based methods have attained compelling results. These methods use a learned (flow) model to transport an initial standard Gaussian noise that ignores the condition to the conditional data distribution. The model is hence required to learn both mass transport and conditional injection. To ease the demand on the model, we propose Condition-Aware Reparameterization for Flow Matching (CAR-Flow) -- a lightweight, learned shift that conditions the source, the target, or both distributions. By relocating these distributions, CAR-Flow shortens the probability path the model must learn, leading to faster training in practice. On low-dimensional synthetic data, we visualize and quantify the effects of CAR-Flow. On higher-dimensional natural image data (ImageNet-256), equipping SiT-XL/2 with CAR-Flow reduces FID from 2.07 to 1.68, while introducing less than 0.6% additional parameters. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_19300 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | CAR-Flow: Condition-Aware Reparameterization Aligns Source and Target for Better Flow Matching Chen, Chen Guo, Pengsheng Song, Liangchen Lu, Jiasen Qian, Rui Wang, Xinze Fu, Tsu-Jui Liu, Wei Yang, Yinfei Schwing, Alex Computer Vision and Pattern Recognition Conditional generative modeling aims to learn a conditional data distribution from samples containing data-condition pairs. For this, diffusion and flow-based methods have attained compelling results. These methods use a learned (flow) model to transport an initial standard Gaussian noise that ignores the condition to the conditional data distribution. The model is hence required to learn both mass transport and conditional injection. To ease the demand on the model, we propose Condition-Aware Reparameterization for Flow Matching (CAR-Flow) -- a lightweight, learned shift that conditions the source, the target, or both distributions. By relocating these distributions, CAR-Flow shortens the probability path the model must learn, leading to faster training in practice. On low-dimensional synthetic data, we visualize and quantify the effects of CAR-Flow. On higher-dimensional natural image data (ImageNet-256), equipping SiT-XL/2 with CAR-Flow reduces FID from 2.07 to 1.68, while introducing less than 0.6% additional parameters. |
| title | CAR-Flow: Condition-Aware Reparameterization Aligns Source and Target for Better Flow Matching |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2509.19300 |