Cross-task Calibration for Asynchronous Federated Continual Learning

Yichen Li, Haozhao Wang, Hang Su, Yulong Li, Xiaoquan Yi, Yankai Jiang, Chuang Zhao, Imran Razzak, Ruixuan Li
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:70938-70953, 2026.

Abstract

Federated Continual Learning (FCL) aims to empower distributed devices to learn a sequence of tasks over time. However, existing FCL research largely relies on the impractical assumption of synchronous new task arrival. This overlooks the reality of asynchronous user behavior and system latencies, forcing more efficient clients to endure costly inactivity. The practical necessity of an asynchronous method gives rise to Asynchronous Federated Continual Learning (AFCL). The server constantly receives a mixture of updates from clients at different time steps, leading to a catastrophic task drift that corrupts the global model and prevents effective learning. In this paper, we introduce a novel Cross-task Calibration framework called C$^2$-AFCL that is the first to tackle task drift at a semantic level within an Asynchronous FCL setting. Its core is a two-stage orthogonal calibration mechanism. First, intra-client calibration uses task-aware caches to mitigate variance from local client drift. Second, and more critically, inter-task interference calibration dynamically estimates an interference subspace from historical task knowledge. New updates are orthogonally projected to isolate and remove components that conflict with this subspace, preserving previous knowledge while learning new tasks. Extensive experiments show that C$^2$-AFCL significantly outperforms existing methods, demonstrating robust and efficient learning in dynamic federated environments.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-li26gx, title = {Cross-task Calibration for Asynchronous Federated Continual Learning}, author = {Li, Yichen and Wang, Haozhao and Su, Hang and Li, Yulong and Yi, Xiaoquan and Jiang, Yankai and Zhao, Chuang and Razzak, Imran and Li, Ruixuan}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {70938--70953}, 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/li26gx/li26gx.pdf}, url = {https://proceedings.mlr.press/v306/li26gx.html}, abstract = {Federated Continual Learning (FCL) aims to empower distributed devices to learn a sequence of tasks over time. However, existing FCL research largely relies on the impractical assumption of synchronous new task arrival. This overlooks the reality of asynchronous user behavior and system latencies, forcing more efficient clients to endure costly inactivity. The practical necessity of an asynchronous method gives rise to Asynchronous Federated Continual Learning (AFCL). The server constantly receives a mixture of updates from clients at different time steps, leading to a catastrophic task drift that corrupts the global model and prevents effective learning. In this paper, we introduce a novel Cross-task Calibration framework called C$^2$-AFCL that is the first to tackle task drift at a semantic level within an Asynchronous FCL setting. Its core is a two-stage orthogonal calibration mechanism. First, intra-client calibration uses task-aware caches to mitigate variance from local client drift. Second, and more critically, inter-task interference calibration dynamically estimates an interference subspace from historical task knowledge. New updates are orthogonally projected to isolate and remove components that conflict with this subspace, preserving previous knowledge while learning new tasks. Extensive experiments show that C$^2$-AFCL significantly outperforms existing methods, demonstrating robust and efficient learning in dynamic federated environments.} }
Endnote
%0 Conference Paper %T Cross-task Calibration for Asynchronous Federated Continual Learning %A Yichen Li %A Haozhao Wang %A Hang Su %A Yulong Li %A Xiaoquan Yi %A Yankai Jiang %A Chuang Zhao %A Imran Razzak %A Ruixuan Li %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-li26gx %I PMLR %P 70938--70953 %U https://proceedings.mlr.press/v306/li26gx.html %V 306 %X Federated Continual Learning (FCL) aims to empower distributed devices to learn a sequence of tasks over time. However, existing FCL research largely relies on the impractical assumption of synchronous new task arrival. This overlooks the reality of asynchronous user behavior and system latencies, forcing more efficient clients to endure costly inactivity. The practical necessity of an asynchronous method gives rise to Asynchronous Federated Continual Learning (AFCL). The server constantly receives a mixture of updates from clients at different time steps, leading to a catastrophic task drift that corrupts the global model and prevents effective learning. In this paper, we introduce a novel Cross-task Calibration framework called C$^2$-AFCL that is the first to tackle task drift at a semantic level within an Asynchronous FCL setting. Its core is a two-stage orthogonal calibration mechanism. First, intra-client calibration uses task-aware caches to mitigate variance from local client drift. Second, and more critically, inter-task interference calibration dynamically estimates an interference subspace from historical task knowledge. New updates are orthogonally projected to isolate and remove components that conflict with this subspace, preserving previous knowledge while learning new tasks. Extensive experiments show that C$^2$-AFCL significantly outperforms existing methods, demonstrating robust and efficient learning in dynamic federated environments.
APA
Li, Y., Wang, H., Su, H., Li, Y., Yi, X., Jiang, Y., Zhao, C., Razzak, I. & Li, R.. (2026). Cross-task Calibration for Asynchronous Federated Continual Learning. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:70938-70953 Available from https://proceedings.mlr.press/v306/li26gx.html.

Related Material