Recode Hive
Test-Time Adaptation vs. Continual Test-Time Adaptation #1031
|
What's the difference between Test-Time Adaptation and Continual Test-Time Adaptation? |
Answered by
seth-woo
Mar 16, 2026
Replies: 1 comment
|
Well in both Test-Time Adaptation and Continual Test-Time Adaptation, model adaptation occurs during online deployment as opposed to other methods like Domain Generalization and Domain Adaptation where it occurs during offline training. Continual Test-Time Adaptation is simply Test-Time Adaptation that is robust to distributional shifts in target distributions and operates under settings where test data is unlabelled, where there are no explicit task boundaries and where there is a "strictly online" constraint (with usually a single gradient update). Hope that helps ! |
0 replies
Answer selected by
memetic-instigator
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Well in both Test-Time Adaptation and Continual Test-Time Adaptation, model adaptation occurs during online deployment as opposed to other methods like Domain Generalization and Domain Adaptation where it occurs during offline training. Continual Test-Time Adaptation is simply Test-Time Adaptation that is robust to distributional shifts in target distributions and operates under settings where test data is unlabelled, where there are no explicit task boundaries and where there is a "strictly online" constraint (with usually a single gradient update). Hope that helps !