ClearGCD: Mitigating Shortcut Learning For Robust Generalized Category Discovery
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866912735116132352 |
|---|---|
| author | Lyu, Kailin He, Jianwei Xiao, Long Zeng, Jianing Fan, Liang Shu, Lin Hao, Jie |
| author_facet | Lyu, Kailin He, Jianwei Xiao, Long Zeng, Jianing Fan, Liang Shu, Lin Hao, Jie |
| contents | In open-world scenarios, Generalized Category Discovery (GCD) requires identifying both known and novel categories within unlabeled data. However, existing methods often suffer from prototype confusion caused by shortcut learning, which undermines generalization and leads to forgetting of known classes. We propose ClearGCD, a framework designed to mitigate reliance on non-semantic cues through two complementary mechanisms. First, Semantic View Alignment (SVA) generates strong augmentations via cross-class patch replacement and enforces semantic consistency using weak augmentations. Second, Shortcut Suppression Regularization (SSR) maintains an adaptive prototype bank that aligns known classes while encouraging separation of potential novel ones. ClearGCD can be seamlessly integrated into parametric GCD approaches and consistently outperforms state-of-the-art methods across multiple benchmarks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_22892 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | ClearGCD: Mitigating Shortcut Learning For Robust Generalized Category Discovery Lyu, Kailin He, Jianwei Xiao, Long Zeng, Jianing Fan, Liang Shu, Lin Hao, Jie Computer Vision and Pattern Recognition Machine Learning In open-world scenarios, Generalized Category Discovery (GCD) requires identifying both known and novel categories within unlabeled data. However, existing methods often suffer from prototype confusion caused by shortcut learning, which undermines generalization and leads to forgetting of known classes. We propose ClearGCD, a framework designed to mitigate reliance on non-semantic cues through two complementary mechanisms. First, Semantic View Alignment (SVA) generates strong augmentations via cross-class patch replacement and enforces semantic consistency using weak augmentations. Second, Shortcut Suppression Regularization (SSR) maintains an adaptive prototype bank that aligns known classes while encouraging separation of potential novel ones. ClearGCD can be seamlessly integrated into parametric GCD approaches and consistently outperforms state-of-the-art methods across multiple benchmarks. |
| title | ClearGCD: Mitigating Shortcut Learning For Robust Generalized Category Discovery |
| topic | Computer Vision and Pattern Recognition Machine Learning |
| url | https://arxiv.org/abs/2511.22892 |