HypCL: Adapting CLIP in Hyperbolic Space for Continual Learning

Quan Cheng, Hao Yu, Da-Wei Zhou, Lijun Zhang
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:19099-19111, 2026.

Abstract

Recently, vision-language models (e.g., CLIP) have been increasingly adopted for continual learning to mitigate catastrophic forgetting. However, existing CLIP-based methods typically freeze the backbone to preserve pre-trained knowledge, which limits the model’s ability to learn discriminative features for downstream tasks. In this paper, we introduce HypCL, a parameter-efficient framework that continually adapts CLIP in hyperbolic space for continual learning. Our key insight is that the exponentially expanding capacity of hyperbolic geometry naturally accommodates the growing class space and promotes stronger inter-class separation. Specifically, HypCL attaches task-specific adapters and composes their updates sequentially in the Poincaré ball. To exploit the enhanced feature separability of hyperbolic geometry, HypCL maintains visual prototypes computed from the adapted features, which serve as stable anchors for calibrating predictions at inference. Extensive experiments on standard class-incremental benchmarks demonstrate that HypCL consistently outperforms existing CLIP-based continual learning methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-cheng26p, title = {{H}yp{CL}: Adapting {CLIP} in Hyperbolic Space for Continual Learning}, author = {Cheng, Quan and Yu, Hao and Zhou, Da-Wei and Zhang, Lijun}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {19099--19111}, 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/cheng26p/cheng26p.pdf}, url = {https://proceedings.mlr.press/v306/cheng26p.html}, abstract = {Recently, vision-language models (e.g., CLIP) have been increasingly adopted for continual learning to mitigate catastrophic forgetting. However, existing CLIP-based methods typically freeze the backbone to preserve pre-trained knowledge, which limits the model’s ability to learn discriminative features for downstream tasks. In this paper, we introduce HypCL, a parameter-efficient framework that continually adapts CLIP in hyperbolic space for continual learning. Our key insight is that the exponentially expanding capacity of hyperbolic geometry naturally accommodates the growing class space and promotes stronger inter-class separation. Specifically, HypCL attaches task-specific adapters and composes their updates sequentially in the Poincaré ball. To exploit the enhanced feature separability of hyperbolic geometry, HypCL maintains visual prototypes computed from the adapted features, which serve as stable anchors for calibrating predictions at inference. Extensive experiments on standard class-incremental benchmarks demonstrate that HypCL consistently outperforms existing CLIP-based continual learning methods.} }
Endnote
%0 Conference Paper %T HypCL: Adapting CLIP in Hyperbolic Space for Continual Learning %A Quan Cheng %A Hao Yu %A Da-Wei Zhou %A Lijun 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-cheng26p %I PMLR %P 19099--19111 %U https://proceedings.mlr.press/v306/cheng26p.html %V 306 %X Recently, vision-language models (e.g., CLIP) have been increasingly adopted for continual learning to mitigate catastrophic forgetting. However, existing CLIP-based methods typically freeze the backbone to preserve pre-trained knowledge, which limits the model’s ability to learn discriminative features for downstream tasks. In this paper, we introduce HypCL, a parameter-efficient framework that continually adapts CLIP in hyperbolic space for continual learning. Our key insight is that the exponentially expanding capacity of hyperbolic geometry naturally accommodates the growing class space and promotes stronger inter-class separation. Specifically, HypCL attaches task-specific adapters and composes their updates sequentially in the Poincaré ball. To exploit the enhanced feature separability of hyperbolic geometry, HypCL maintains visual prototypes computed from the adapted features, which serve as stable anchors for calibrating predictions at inference. Extensive experiments on standard class-incremental benchmarks demonstrate that HypCL consistently outperforms existing CLIP-based continual learning methods.
APA
Cheng, Q., Yu, H., Zhou, D. & Zhang, L.. (2026). HypCL: Adapting CLIP in Hyperbolic Space for Continual Learning. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:19099-19111 Available from https://proceedings.mlr.press/v306/cheng26p.html.

Related Material