IT$^3$: Idempotent Test-Time Training

Nikita Durasov, Assaf Shocher, Doruk Oner, Gal Chechik, Alexei A Efros, Pascal Fua
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:14867-14883, 2025.

Abstract

Deep learning models often struggle when deployed in real-world settings due to distribution shifts between training and test data. While existing approaches like domain adaptation and test-time training (TTT) offer partial solutions, they typically require additional data or domain-specific auxiliary tasks. We present Idempotent Test-Time Training (IT3), a novel approach that enables on-the-fly adaptation to distribution shifts using only the current test instance, without any auxiliary task design. Our key insight is that enforcing idempotence—where repeated applications of a function yield the same result—can effectively replace domain-specific auxiliary tasks used in previous TTT methods. We theoretically connect idempotence to prediction confidence and demonstrate that minimizing the distance between successive applications of our model during inference leads to improved out-of-distribution performance. Extensive experiments across diverse domains (including image classification, aerodynamics prediction, and aerial segmentation) and architectures (MLPs, CNNs, GNNs) show that IT3 consistently outperforms existing approaches while being simpler and more widely applicable. Our results suggest that idempotence provides a universal principle for test-time adaptation that generalizes across domains and architectures.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-durasov25a, title = {{IT}$^3$: Idempotent Test-Time Training}, author = {Durasov, Nikita and Shocher, Assaf and Oner, Doruk and Chechik, Gal and Efros, Alexei A and Fua, Pascal}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {14867--14883}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/durasov25a/durasov25a.pdf}, url = {https://proceedings.mlr.press/v267/durasov25a.html}, abstract = {Deep learning models often struggle when deployed in real-world settings due to distribution shifts between training and test data. While existing approaches like domain adaptation and test-time training (TTT) offer partial solutions, they typically require additional data or domain-specific auxiliary tasks. We present Idempotent Test-Time Training (IT3), a novel approach that enables on-the-fly adaptation to distribution shifts using only the current test instance, without any auxiliary task design. Our key insight is that enforcing idempotence—where repeated applications of a function yield the same result—can effectively replace domain-specific auxiliary tasks used in previous TTT methods. We theoretically connect idempotence to prediction confidence and demonstrate that minimizing the distance between successive applications of our model during inference leads to improved out-of-distribution performance. Extensive experiments across diverse domains (including image classification, aerodynamics prediction, and aerial segmentation) and architectures (MLPs, CNNs, GNNs) show that IT3 consistently outperforms existing approaches while being simpler and more widely applicable. Our results suggest that idempotence provides a universal principle for test-time adaptation that generalizes across domains and architectures.} }
Endnote
%0 Conference Paper %T IT$^3$: Idempotent Test-Time Training %A Nikita Durasov %A Assaf Shocher %A Doruk Oner %A Gal Chechik %A Alexei A Efros %A Pascal Fua %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-durasov25a %I PMLR %P 14867--14883 %U https://proceedings.mlr.press/v267/durasov25a.html %V 267 %X Deep learning models often struggle when deployed in real-world settings due to distribution shifts between training and test data. While existing approaches like domain adaptation and test-time training (TTT) offer partial solutions, they typically require additional data or domain-specific auxiliary tasks. We present Idempotent Test-Time Training (IT3), a novel approach that enables on-the-fly adaptation to distribution shifts using only the current test instance, without any auxiliary task design. Our key insight is that enforcing idempotence—where repeated applications of a function yield the same result—can effectively replace domain-specific auxiliary tasks used in previous TTT methods. We theoretically connect idempotence to prediction confidence and demonstrate that minimizing the distance between successive applications of our model during inference leads to improved out-of-distribution performance. Extensive experiments across diverse domains (including image classification, aerodynamics prediction, and aerial segmentation) and architectures (MLPs, CNNs, GNNs) show that IT3 consistently outperforms existing approaches while being simpler and more widely applicable. Our results suggest that idempotence provides a universal principle for test-time adaptation that generalizes across domains and architectures.
APA
Durasov, N., Shocher, A., Oner, D., Chechik, G., Efros, A.A. & Fua, P.. (2025). IT$^3$: Idempotent Test-Time Training. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:14867-14883 Available from https://proceedings.mlr.press/v267/durasov25a.html.

Related Material