Research map / Label-efficient and robust learning / Test-time adaptation
In Search of Lost Online Test-Time Adaptation: A Survey
Zixin Wang, Yadan Luo, Liang Zheng, Zhuoxiao Chen, Sen Wang, Zi Helen Huang
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 |
|---|---|---|---|
| update all parameters instead of only LayerNorm (ablation) | CoTTA · CIFAR-10-C severity 5 (batch size 16) | error rate (%): 24.71 -> 52.47 | Hurt |
| add parameter reset to full-network updating (ablation) | CoTTA-ALL · CIFAR-10-C severity 5 (batch size 16) | error rate (%): 77.08 -> 52.47 | Improved |
| add parameter reset to LayerNorm-only updating (ablation) | CoTTA-LN · CIFAR-10-C severity 5 (batch size 16) | error rate (%): 24.68 -> 24.71 | No clear effect |
| add parameter reset to full-network updating (ablation) | CoTTA-ALL · ImageNet-C severity 5 (batch size 16) | error rate (%): 90.54 -> 85.16 | Improved |
| use LayerNorm-only updating instead of full update (ablation) | CoTTA · ImageNet-C severity 5 (batch size 16) | error rate (%): 85.16 -> 72.37 | Improved |
| apply Tent entropy minimization to ViT | Tent · CIFAR-10-C severity 5 (batch size 16) | error rate (%): 23.89 -> 24.90 | Hurt |
| apply Tent entropy minimization to ViT | Tent · ImageNet-C severity 5 (batch size 16) | error rate (%): 63.24 -> 77.66 | Hurt |
| apply ROID to ViT | ROID · CIFAR-10-C severity 5 (batch size 16) | error rate (%): 23.89 -> 13.36 | Improved |
| apply ROID to ViT | ROID · ImageNet-C severity 5 (batch size 16) | error rate (%): 63.24 -> 46.70 | Improved |
| use batch size 1 instead of 16 for ROID (ablation) | ROID · CIFAR-10-C severity 5 | error rate (%): 13.36 -> 89.99 | Hurt |
| use batch size 1 instead of 16 for Tent (ablation) | Tent · CIFAR-10-C severity 5 | error rate (%): 24.90 -> 52.25 | Hurt |
| apply SAR (sharpness-aware) on top of Tent-style update | SAR · ImageNet-C severity 5 (batch size 16) | error rate (%): 77.66 -> 52.70 | Improved |
Closest papers by meaning
- Stationary Latent Weight Inference for Unreliable Observations from Online Test-Time Adaptation (ICML 2024 · similarity 0.90)
- ReservoirTTA: Prolonged Test-time Adaptation for Evolving and Recurring Domains (NeurIPS 2025 · similarity 0.88)
- Unraveling Batch Normalization for Realistic Test-Time Adaptation (AAAI 2024 · similarity 0.87)
- Evaluation of Test-Time Adaptation Under Computational Time Constraints (ICML 2024 · similarity 0.87)
- A Layer Selection Approach to Test Time Adaptation (AAAI 2025 · similarity 0.87)
- Free on the Fly: Enhancing Flexibility in Test-Time Adaptation with Online EM (CVPR 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.
