Images Speak Louder Than Scores: Failure Mode Escape for Enhancing Generative Quality
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_ | 1866908487739506688 |
|---|---|
| author | Shao, Jie Zhu, Ke Fu, Minghao Wang, Guo-hua Wu, Jianxin |
| author_facet | Shao, Jie Zhu, Ke Fu, Minghao Wang, Guo-hua Wu, Jianxin |
| contents | Diffusion models have achieved remarkable progress in class-to-image generation. However, we observe that despite impressive FID scores, state-of-the-art models often generate distorted or low-quality images, especially in certain classes. This gap arises because FID evaluates global distribution alignment, while ignoring the perceptual quality of individual samples. We further examine the role of CFG, a common technique used to enhance generation quality. While effective in improving metrics and suppressing outliers, CFG can introduce distribution shift and visual artifacts due to its misalignment with both training objectives and user expectations. In this work, we propose FaME, a training-free and inference-efficient method for improving perceptual quality. FaME uses an image quality assessment model to identify low-quality generations and stores their sampling trajectories. These failure modes are then used as negative guidance to steer future sampling away from poor-quality regions. Experiments on ImageNet demonstrate that FaME brings consistent improvements in visual quality without compromising FID. FaME also shows the potential to be extended to improve text-to-image generation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_09598 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Images Speak Louder Than Scores: Failure Mode Escape for Enhancing Generative Quality Shao, Jie Zhu, Ke Fu, Minghao Wang, Guo-hua Wu, Jianxin Computer Vision and Pattern Recognition Diffusion models have achieved remarkable progress in class-to-image generation. However, we observe that despite impressive FID scores, state-of-the-art models often generate distorted or low-quality images, especially in certain classes. This gap arises because FID evaluates global distribution alignment, while ignoring the perceptual quality of individual samples. We further examine the role of CFG, a common technique used to enhance generation quality. While effective in improving metrics and suppressing outliers, CFG can introduce distribution shift and visual artifacts due to its misalignment with both training objectives and user expectations. In this work, we propose FaME, a training-free and inference-efficient method for improving perceptual quality. FaME uses an image quality assessment model to identify low-quality generations and stores their sampling trajectories. These failure modes are then used as negative guidance to steer future sampling away from poor-quality regions. Experiments on ImageNet demonstrate that FaME brings consistent improvements in visual quality without compromising FID. FaME also shows the potential to be extended to improve text-to-image generation. |
| title | Images Speak Louder Than Scores: Failure Mode Escape for Enhancing Generative Quality |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2508.09598 |