Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2504.14290 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917082331873280 |
|---|---|
| author | Ruan, Shouwei Wu, Zhenyu Huang, Yao Zhang, Ruochen Sun, Yitong Kang, Caixin Zhao, Shiji Wei, Xingxing |
| author_facet | Ruan, Shouwei Wu, Zhenyu Huang, Yao Zhang, Ruochen Sun, Yitong Kang, Caixin Zhao, Shiji Wei, Xingxing |
| contents | Content safety is a fundamental challenge for text-to-image (T2I) models, yet prevailing methods enforce a debilitating trade-off between safety and generation quality. We argue that mitigating this trade-off hinges on addressing systemic challenges in current T2I safety alignment across data, methods, and evaluation protocols. To this end, we introduce a unified framework for synergistic safety alignment. First, to overcome the flawed data paradigm that provides biased optimization signals, we develop LibraAlign-100K, the first large-scale dataset with dual annotations for safety and quality. Second, to address the myopic optimization of existing methods focus solely on safety reward, we propose Synergistic Preference Optimization (T2I-SPO), a novel alignment algorithm that extends the DPO paradigm with a composite reward function that integrates generation safety and quality to holistically model user preferences. Finally, to overcome the limitations of quality-agnostic and binary evaluation in current protocols, we introduce the Unified Alignment Score, a holistic, fine-grained metric that fairly quantifies the balance between safety and generative capability. Extensive experiments demonstrate that T2I-SPO achieves state-of-the-art safety alignment against a wide range of NSFW concepts, while better maintaining the model's generation quality and general capability |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_14290 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | The Path to Reconciling Quality and Safety in Text-to-Image Generation: Dataset, Method, and Evaluation Ruan, Shouwei Wu, Zhenyu Huang, Yao Zhang, Ruochen Sun, Yitong Kang, Caixin Zhao, Shiji Wei, Xingxing Computer Vision and Pattern Recognition Content safety is a fundamental challenge for text-to-image (T2I) models, yet prevailing methods enforce a debilitating trade-off between safety and generation quality. We argue that mitigating this trade-off hinges on addressing systemic challenges in current T2I safety alignment across data, methods, and evaluation protocols. To this end, we introduce a unified framework for synergistic safety alignment. First, to overcome the flawed data paradigm that provides biased optimization signals, we develop LibraAlign-100K, the first large-scale dataset with dual annotations for safety and quality. Second, to address the myopic optimization of existing methods focus solely on safety reward, we propose Synergistic Preference Optimization (T2I-SPO), a novel alignment algorithm that extends the DPO paradigm with a composite reward function that integrates generation safety and quality to holistically model user preferences. Finally, to overcome the limitations of quality-agnostic and binary evaluation in current protocols, we introduce the Unified Alignment Score, a holistic, fine-grained metric that fairly quantifies the balance between safety and generative capability. Extensive experiments demonstrate that T2I-SPO achieves state-of-the-art safety alignment against a wide range of NSFW concepts, while better maintaining the model's generation quality and general capability |
| title | The Path to Reconciling Quality and Safety in Text-to-Image Generation: Dataset, Method, and Evaluation |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2504.14290 |