Test-Time Adaptation
A machine learning paradigm in which model parameters or states are adjusted at the inference or deployment stage rather than the training stage, enabling the model to adapt to test environments that differ from the training distribution. It is commonly used to improve model generalization and robustness.
Timeline (last 90 days)
T2-Z belongs to the test-time adaptation (TTA) paradigm, avoiding full-model fine-tuning for each new environment and compressing adaptation capability into a low-dimensional latent variable.
Test-Time Adaptation requires the model to dynamically adjust its own parameters during the inference phase using unlabeled test data in order to handle distribution shift.
Test-Time Adaptation (TTA) allows a model to perform online updates of batch normalization layer parameters using a small amount of target domain data during inference, serving as a lightweight approach to addressing domain shift.
All Facts (3)
Test-Time Adaptation (TTA) allows a model to perform online updates of batch normalization layer parameters using a small amount of target domain data during inference, serving as a lightweight approach to addressing domain shift.
70%UnverifiedT2-Z belongs to the test-time adaptation (TTA) paradigm, avoiding full-model fine-tuning for each new environment and compressing adaptation capability into a low-dimensional latent variable.
50%UnverifiedTest-Time Adaptation requires the model to dynamically adjust its own parameters during the inference phase using unlabeled test data in order to handle distribution shift.
50%