Research map / Label-efficient and robust learning / Test-time adaptation
CAFA: Class-Aware Feature Alignment for Test-Time Adaptation
Sanghun Jung, Jungsoo Lee, Nanhee Kim, Amirreza Shaban, Byron Boots, Jaegul Choo
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 |
|---|---|---|---|
| replace class-agnostic global feature alignment with class-aware (CAFA) alignment (ablation) | TTA feature alignment · CIFAR10-C | classification error (%): 19.12 -> 12.13 | Improved |
| add inter-class distance enlargement to intra-class alignment (ablation) | CAFA · CIFAR10-C | classification error (%): 13.02 -> 12.13 | Improved |
| add inter-class distance enlargement to intra-class alignment (ablation) | CAFA · ImageNet-C | classification error (%): 59.12 -> 55.24 | Improved |
| update only BN affine parameters instead of all feature extractor parameters (ablation) | CAFA · CIFAR10-C | classification error (%): 12.66 -> 12.13 | Improved |
| update all parameters vs BN only on ImageNet-C (ablation) | CAFA · ImageNet-C | classification error (%) | No clear effect |
| use class-wise covariance instead of tied covariance (ablation) | CAFA · CIFAR10-C | classification error (%): 12.47 -> 12.13 | Improved |
| use Mahalanobis distance instead of Euclidean distance (ablation) | CAFA · ImageNet-C | classification error (%): 56.30 -> 55.24 | Improved |
| use Mahalanobis distance instead of class-aware KL-divergence (ablation) | CAFA · CIFAR10-C | classification error (%): 12.53 -> 12.13 | Improved |
| use CAFA instead of TENT under label shift | test-time adaptation · CIFAR10-C, label shift s=1.0 | classification error (%): 26.01 -> 17.61 | Improved |
| use CAFA instead of EATA on ImageNet-C | test-time adaptation · ImageNet-C | classification error (%): 52.00 -> 55.24 | Hurt |
| estimate source statistics from about 5% of training samples (ablation) | CAFA · CIFAR10-C and ImageNet-C | classification error (%) | No clear effect |
| use CAFA instead of TENT at batch size 1 | test-time adaptation · CIFAR10-C, batch size 1 | classification error (%): 17.66 -> 15.00 | Improved |
Later papers that compared with this method
2 later accepted papers report CAFA as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Alleviate and Mining: Rethinking Unsupervised Domain Adaptation for Mitochondria Segmentation from Pseudo-Label Perspective (AAAI 2025 · better on 6 of 6 shared results)
- MaskTwins: Dual-form Complementary Masking for Domain-Adaptive Image Segmentation (ICML 2025 · better on 4 of 4 shared results)
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 7 of 7 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
- Test-time Correlation Alignment (ICML 2025 · similarity 0.90)
- Test-time Adaptation for Regression by Subspace Alignment (ICLR 2025 · similarity 0.88)
- L-TTA: Lightweight Test-Time Adaptation Using a Versatile Stem Layer (NeurIPS 2024 · similarity 0.88)
- Backpropagation-Free Test-Time Adaptation via Probabilistic Gaussian Alignment (NeurIPS 2025 · similarity 0.88)
- A Layer Selection Approach to Test Time Adaptation (AAAI 2025 · similarity 0.87)
- Efficient Test-time Adaptation through Candidate Verification and Divergence Shifts (NeurIPS 2026 · 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.
