CURE: Consistency-under-Unified Semantic Regularization for Generalized Category Discovery

Yuwei Bian, Shidong Wang, Haofeng Zhang
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:8182-8203, 2026.

Abstract

Generalized Category Discovery (GCD) aims to learn semantically structured representations for discovering novel categories in unlabeled data using supervision from known classes. Most existing methods rely on self-supervised contrastive learning (CL) with consistency and uniformity objectives. We identify an inherent optimization conflict between these objectives: while uniformity enforces global feature dispersion, it can hinder the formation of class-discriminative and semantically coherent structures. To address this issue, we propose a two-stage framework that decouples representation learning from self-contrastive regularization. The first stage learns category-anchored representations aligned with known class prototypes, while the second stage extends the representation space to novel categories via a consistency objective enhanced with unified semantic regularization. We further introduce a Semantic Exploration Energy mechanism to capture shared semantics across categories and mitigate information loss caused by prototype orthogonalization. The resulting framework, termed Consistency-under-Unified Semantic Regularization(CURE), achieves state-of-the-art performance on multiple benchmarks and substantially reduces the performance gap between known and novel categories.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-bian26d, title = {{CURE}: Consistency-under-Unified Semantic Regularization for Generalized Category Discovery}, author = {Bian, Yuwei and Wang, Shidong and Zhang, Haofeng}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {8182--8203}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/bian26d/bian26d.pdf}, url = {https://proceedings.mlr.press/v306/bian26d.html}, abstract = {Generalized Category Discovery (GCD) aims to learn semantically structured representations for discovering novel categories in unlabeled data using supervision from known classes. Most existing methods rely on self-supervised contrastive learning (CL) with consistency and uniformity objectives. We identify an inherent optimization conflict between these objectives: while uniformity enforces global feature dispersion, it can hinder the formation of class-discriminative and semantically coherent structures. To address this issue, we propose a two-stage framework that decouples representation learning from self-contrastive regularization. The first stage learns category-anchored representations aligned with known class prototypes, while the second stage extends the representation space to novel categories via a consistency objective enhanced with unified semantic regularization. We further introduce a Semantic Exploration Energy mechanism to capture shared semantics across categories and mitigate information loss caused by prototype orthogonalization. The resulting framework, termed Consistency-under-Unified Semantic Regularization(CURE), achieves state-of-the-art performance on multiple benchmarks and substantially reduces the performance gap between known and novel categories.} }
Endnote
%0 Conference Paper %T CURE: Consistency-under-Unified Semantic Regularization for Generalized Category Discovery %A Yuwei Bian %A Shidong Wang %A Haofeng Zhang %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-bian26d %I PMLR %P 8182--8203 %U https://proceedings.mlr.press/v306/bian26d.html %V 306 %X Generalized Category Discovery (GCD) aims to learn semantically structured representations for discovering novel categories in unlabeled data using supervision from known classes. Most existing methods rely on self-supervised contrastive learning (CL) with consistency and uniformity objectives. We identify an inherent optimization conflict between these objectives: while uniformity enforces global feature dispersion, it can hinder the formation of class-discriminative and semantically coherent structures. To address this issue, we propose a two-stage framework that decouples representation learning from self-contrastive regularization. The first stage learns category-anchored representations aligned with known class prototypes, while the second stage extends the representation space to novel categories via a consistency objective enhanced with unified semantic regularization. We further introduce a Semantic Exploration Energy mechanism to capture shared semantics across categories and mitigate information loss caused by prototype orthogonalization. The resulting framework, termed Consistency-under-Unified Semantic Regularization(CURE), achieves state-of-the-art performance on multiple benchmarks and substantially reduces the performance gap between known and novel categories.
APA
Bian, Y., Wang, S. & Zhang, H.. (2026). CURE: Consistency-under-Unified Semantic Regularization for Generalized Category Discovery. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:8182-8203 Available from https://proceedings.mlr.press/v306/bian26d.html.

Related Material