Research map / Label-efficient and robust learning / Test-time adaptation
Neural Collapse in Test-Time Adaptation
Xiao Chen, Zhongjing Du, Jiazhen Huang, Xu Jiang, Li Lu, Jingyan Jiang, Zhi Wang
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 L_NC feature-classifier alignment loss to entropy minimization (ablation) | Tent/entropy baseline · ImageNet-C (ViT-B/16) | avg accuracy: 51.13 -> 53.17 | Improved |
| add L_NC feature-classifier alignment loss to entropy minimization (ablation) | Tent/entropy baseline · CIFAR-10-C (ResNet50) | avg accuracy: 75.19 -> 75.31 | Improved |
| add L_NC to entropy loss with filtering and dynamic weight (ablation) | NCTTA · ImageNet-C (ViT-B/16) | avg accuracy: 53.97 -> 66.50 | Improved |
| add L_NC to entropy loss with filtering and dynamic weight (ablation) | NCTTA · CIFAR-10-C (ResNet50) | avg accuracy: 76.87 -> 78.10 | Improved |
| use hybrid targets (tuned alpha, k) instead of pseudo-label only (ablation) | NCTTA · ImageNet-C Contrast severity 5 | accuracy | Improved |
| choose InfoNCE-style vs L2 or Triplet L_NC (ablation) | NCTTA · CIFAR-10-C / ImageNet-C | avg accuracy | Mixed |
| use NCTTA instead of Tent | NCTTA · ImageNet-C CTTA | avg accuracy: 66.27 -> 71.27 | Improved |
| use NCTTA instead of Tent | NCTTA · ImageNet-C batch size 1 | avg accuracy: 49.19 -> 59.69 | Improved |
| use NCTTA instead of DeYO | NCTTA · Waterbirds | worst-group accuracy: 75.65 -> 76.56 | No clear effect |
| use NCTTA instead of DeYO | NCTTA · PACS | avg accuracy: 76.67 -> 77.36 | Improved |
| apply Tent adaptation to snow and frost corruptions | Tent · ImageNet-C (ViT-B/16) | accuracy: 22.15 -> 17.33 (Snow) | Hurt |
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 5 of 5 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
- Entropy is not Enough for Test-Time Adaptation: From the Perspective of Disentangled Factors (ICLR 2024 · similarity 0.87)
- ReservoirTTA: Prolonged Test-time Adaptation for Evolving and Recurring Domains (NeurIPS 2025 · similarity 0.87)
- Test-time Correlation Alignment (ICML 2025 · similarity 0.87)
- Neural Collapse Beyond the Unconstrained Features Model: Landscape, Dynamics, and Generalization in the Mean-Field Regime (ICML 2025 · similarity 0.86)
- A Layer Selection Approach to Test Time Adaptation (AAAI 2025 · similarity 0.86)
- Backpropagation-Free Test-Time Adaptation via Probabilistic Gaussian Alignment (NeurIPS 2025 · similarity 0.86)
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.
