Research map / Label-efficient and robust learning / Test-time adaptation
Improved Test-Time Adaptation for Domain Generalization
Liang Chen, Yong Zhang, Yibing Song, Ying Shan, Lingqiao Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 a learnable weight subnetwork to the consistency loss (ablation) | ITTA · PACS | accuracy: 82.5±0.1 -> 83.8±0.3 | Improved |
| replace learnable consistency with entropy minimization (ablation) | ITTA · PACS | accuracy: 83.8±0.3 -> 82.4±0.8 | Hurt |
| replace learnable consistency with rotation estimation (ablation) | ITTA · PACS | accuracy: 83.8±0.3 -> 82.1±0.3 | Hurt |
| add test-time updating of the new adaptive parameters (ablation) | ITTA · PACS | accuracy: 82.8±0.3 -> 83.8±0.3 | Improved |
| update all original feature-extractor parameters instead of the added adaptive parameters (ablation) | ITTA · PACS | accuracy: 83.8±0.3 -> 83.0±0.2 | Hurt |
| update only batch-normalization parameters instead of the added adaptive parameters (ablation) | ITTA · PACS | accuracy: 83.8±0.3 -> 82.4±0.5 | Hurt |
| increase the number of TTT updates from one to three (ablation) | ITTA · PACS | accuracy: 83.8 ± 0.3 -> 83.9 ± 0.3 | No clear effect |
| switch from online to episodic updating (ablation) | ITTA · PACS | accuracy: 83.8 ± 0.3 -> 83.3 ± 0.4 | Hurt |
| vary the depth of the weight subnetwork or adaptive blocks (ablation) | ITTA · PACS | accuracy: 83.1 ± 0.6 to 83.8 ± 0.3 | No clear effect |
| use a learnable consistency loss rather than naive consistency with the alternative augmentation (ablation) | ITTA · PACS | accuracy: 79.7 ± 0.2 -> 80.8 ± 0.4 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- TENT — Tent: Fully Test-Time Adaptation by Entropy Minimization (ICLR 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- TTA-FedDG: Leveraging Test-Time Adaptation to Address Federated Domain Generalization (AAAI 2025 · similarity 0.89)
- A Versatile Framework for Continual Test-Time Domain Adaptation: Balancing Discriminability and Generalizability (CVPR 2024 · similarity 0.88)
- DPCore: Dynamic Prompt Coreset for Continual Test-Time Adaptation (ICML 2025 · similarity 0.88)
- Everything to the Synthetic: Diffusion-driven Test-time Adaptation via Synthetic-Domain Alignment (CVPR 2025 · similarity 0.87)
- Test-Time Domain Adaptation by Learning Domain-Aware Batch Normalization (AAAI 2024 · similarity 0.87)
- Test-time Correlation Alignment (ICML 2025 · similarity 0.87)
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.
