Imbalance in Balance: Online Concept Balancing in Generation Models
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_ | 1866915611493269504 |
|---|---|
| author | Shi, Yukai Ou, Jiarong Chen, Rui Yang, Haotian Wang, Jiahao Tao, Xin Wan, Pengfei Zhang, Di Gai, Kun |
| author_facet | Shi, Yukai Ou, Jiarong Chen, Rui Yang, Haotian Wang, Jiahao Tao, Xin Wan, Pengfei Zhang, Di Gai, Kun |
| contents | In visual generation tasks, the responses and combinations of complex concepts often lack stability and are error-prone, which remains an under-explored area. In this paper, we attempt to explore the causal factors for poor concept responses through elaborately designed experiments. We also design a concept-wise equalization loss function (IMBA loss) to address this issue. Our proposed method is online, eliminating the need for offline dataset processing, and requires minimal code changes. In our newly proposed complex concept benchmark Inert-CompBench and two other public test sets, our method significantly enhances the concept response capability of baseline models and yields highly competitive results with only a few codes released at https://github.com/KwaiVGI/IMBA-Loss. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_13345 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Imbalance in Balance: Online Concept Balancing in Generation Models Shi, Yukai Ou, Jiarong Chen, Rui Yang, Haotian Wang, Jiahao Tao, Xin Wan, Pengfei Zhang, Di Gai, Kun Computer Vision and Pattern Recognition Artificial Intelligence In visual generation tasks, the responses and combinations of complex concepts often lack stability and are error-prone, which remains an under-explored area. In this paper, we attempt to explore the causal factors for poor concept responses through elaborately designed experiments. We also design a concept-wise equalization loss function (IMBA loss) to address this issue. Our proposed method is online, eliminating the need for offline dataset processing, and requires minimal code changes. In our newly proposed complex concept benchmark Inert-CompBench and two other public test sets, our method significantly enhances the concept response capability of baseline models and yields highly competitive results with only a few codes released at https://github.com/KwaiVGI/IMBA-Loss. |
| title | Imbalance in Balance: Online Concept Balancing in Generation Models |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2507.13345 |