Research map / Label-efficient and robust learning / Test-time adaptation
Test-Time Classifier Adjustment Module for Model-Agnostic Domain Generalization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add T3A test-time classifier adjustment to ERM | ERM ResNet50 · Average of 4 DG datasets | accuracy: 68.6 -> 70.1 | Helped |
| add T3A to CORAL | CORAL ResNet50 · Average of 4 DG datasets | accuracy: 69.8 -> 70.4 | Helped |
| add T3A across 10 different backbones | ERM · Average of 4 DG datasets | accuracy: 75.1 -> 76.1 (HViT) | Helped |
| add T3A on TerraIncognita for some backbones (BiT-M-R50x3) | ERM BiT-M-R50x3 · TerraIncognita | accuracy: 52.5 -> 52.2 | No clear effect |
| replace T3A with Tent-C (entropy minimization of classifier via SGD) | ERM ResNet50 · Average of 4 DG datasets | accuracy: 68.6 -> 67.6 | Hurt |
| replace T3A with Tent-BN | ERM ResNet50 · Average of 4 DG datasets | accuracy: 68.6 -> 66.2 | Hurt |
| add T3A vs no adaptation on BN ResNet50 | ResNet50 w/ BN · Average of 4 DG datasets | accuracy: 67.8 -> 68.5 | Helped |
| SHOT/PL updating the feature extractor at test time | ResNet50 w/ BN · Average of 4 DG datasets | accuracy: 67.8 -> 62.3 (SHOT) | Hurt |
| fine-tune only the linear classifier versus none (ablation) | ResNet50 ERM · VLCS | accuracy: 72.5 -> 82.9 | Helped |
Closest papers by meaning
- Test-time Correlation Alignment (ICML 2025 · similarity 0.85)
- Lost Domain Generalization Is a Natural Consequence of Lack of Training Domains (AAAI 2024 · similarity 0.85)
- Gradient-Guided Annealing for Domain Generalization (CVPR 2025 · similarity 0.84)
- TTA-FedDG: Leveraging Test-Time Adaptation to Address Federated Domain Generalization (AAAI 2025 · similarity 0.84)
- Set Valued Predictions For Robust Domain Generalization (ICML 2025 · similarity 0.84)
- Everything to the Synthetic: Diffusion-driven Test-time Adaptation via Synthetic-Domain Alignment (CVPR 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.