Research map / Label-efficient and robust learning / Test-time adaptation
Towards Stable Test-time Adaptation in Dynamic Wild World
Shuaicheng Niu, Jiaxiang Wu, Yifan Zhang, Zhiquan Wen, Yaofo Chen, Peilin Zhao, Mingkui Tan
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 |
|---|---|---|---|
| use SAR instead of EATA | ResNet50-GN · ImageNet-C, severity 5, online imbalanced label shifts | accuracy (%): 31.6 -> 37.2 | Improved |
| use SAR instead of EATA | VitBase-LN · ImageNet-C, severity 5, online imbalanced label shifts | accuracy (%): 49.9% vs. 58.0% | Improved |
| use SAR instead of EATA | ResNet50-GN · ImageNet-C, mixed corruption types, severity 5 | accuracy (%): 38.1 -> 38.3 | Improved |
| add batch-agnostic group or layer normalization instead of batch normalization | online test-time adaptation · ImageNet-C, mixed shifts, small batches, and imbalanced label shifts | stability | Improved |
| add reliable entropy filtering (ablation) | pure entropy minimization on ResNet50-GN · ImageNet-C, severity 5, online imbalanced label shifts | average accuracy (%): 24.6%->31.7% | Improved |
| add reliable entropy filtering (ablation) | pure entropy minimization on VitBase-LN · ImageNet-C, severity 5, online imbalanced label shifts | average accuracy (%): 47.6%\rightarrow 53.1% | Improved |
| add sharpness-aware minimization to reliable entropy (ablation) | ResNet50-GN · ImageNet-C, severity 5, online imbalanced label shifts | average accuracy (%): 31.7%->37.0% | Improved |
| add sharpness-aware minimization to reliable entropy (ablation) | VitBase-LN · ImageNet-C, shot noise, severity 5 | accuracy (%): 35.7%\rightarrow 47.6% | Improved |
| add model recovery to reliable entropy plus sharpness-aware minimization (ablation) | VitBase-LN · ImageNet-C, severity 5, online imbalanced label shifts | average accuracy (%): 54.5%\rightarrow 58.2% | Improved |
| add model recovery to reliable entropy plus sharpness-aware minimization (ablation) | ResNet50-GN · ImageNet-C, severity 5, online imbalanced label shifts | average accuracy (%): 37.0 -> 37.0 | No clear effect |
| use large-threshold gradient clipping instead of SAR (ablation) | online test-time adaptation · ImageNet-C, shot noise, severity 5 | accuracy | Hurt |
| use small-threshold gradient clipping instead of SAR (ablation) | online test-time adaptation · ImageNet-C, shot noise, severity 5 | accuracy | No clear effect |
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 2 of 3 shared results
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
- Towards Real-World Test-Time Adaptation: Tri-net Self-Training with Balanced Normalization (AAAI 2024 · similarity 0.92)
- A Layer Selection Approach to Test Time Adaptation (AAAI 2025 · similarity 0.92)
- Evaluation of Test-Time Adaptation Under Computational Time Constraints (ICML 2024 · similarity 0.91)
- Bayesian Weight Enhancement with Steady-State Adaptation for Test-time Adaptation in Dynamic Environments (ICML 2025 · similarity 0.91)
- Buffer layers for Test-Time Adaptation (NeurIPS 2025 · similarity 0.90)
- Unified Entropy Optimization for Open-Set Test-Time Adaptation (CVPR 2024 · similarity 0.90)
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.
