Research map / Label-efficient and robust learning / Test-time adaptation
Robust Test-Time Adaptation in Dynamic Scenarios
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 RoTTA instead of the strongest baseline, NOTE | CIFAR10-C severity 5 PTTA · CIFAR10-C severity 5 | classification error: 31.1 -> 25.2 | Helped |
| Use RoTTA instead of the strongest baseline, LAME | CIFAR100-C severity 5 PTTA · CIFAR100-C severity 5 | classification error: 40.5 -> 35.0 | Helped |
| Use RoTTA instead of the strongest baseline, LAME | DomainNet PTTA · DomainNet | average classification error: 71.4 -> 69.2 | Helped |
| Add robust batch normalization (RBN) instead of test-time BN (ablation) | RoTTA · CIFAR10-C | classification error: 75.4 -> 25.2 | Helped |
| Add robust batch normalization (RBN) instead of test-time BN (ablation) | RoTTA · CIFAR100-C | classification error: 51.3 -> 35.0 | Helped |
| Add CSTU category-balanced memory sampling instead of adapting directly on the stream (ablation) | RoTTA · CIFAR10-C | classification error: 47.1 -> 25.2 | Helped |
| Add CSTU category-balanced memory sampling instead of adapting directly on the stream (ablation) | RoTTA · CIFAR100-C | classification error: 46.3 -> 35.0 | Helped |
| Add robust teacher-student training with timeliness weighting instead of entropy minimization (ablation) | RoTTA · CIFAR10-C | classification error: 78.2 -> 25.2 | Helped |
| Add robust teacher-student training with timeliness weighting instead of entropy minimization (ablation) | RoTTA · CIFAR100-C | classification error: 95.0 -> 35.0 | Helped |
| Balance sample timeliness and uncertainty in memory selection rather than using either alone (ablation) | RoTTA · CIFAR100-C | classification error: 57.5 -> 35.0; 38.9 -> 35.0 | Helped |
| Set the RBN statistics update rate to an intermediate value rather than updating too aggressively or gently (ablation) | RoTTA · CIFAR100-C | classification error: 39.0; 36.0; 35.0; 36.0; 38.1; 41.5 | Mixed |
| Set the teacher EMA update rate to an intermediate value (ablation) | RoTTA · CIFAR100-C | classification error: 44.8; 39.1; 37.1; 35.0; 37.6; 43.6 | Mixed |
Closest papers by meaning
- Persistent Test-time Adaptation in Recurring Testing Scenarios (NeurIPS 2024 · similarity 0.88)
- Evaluation of Test-Time Adaptation Under Computational Time Constraints (ICML 2024 · similarity 0.87)
- Active Test-Time Adaptation: Theoretical Analyses and An Algorithm (ICLR 2024 · similarity 0.87)
- Bayesian Weight Enhancement with Steady-State Adaptation for Test-time Adaptation in Dynamic Environments (ICML 2025 · similarity 0.87)
- Buffer layers for Test-Time Adaptation (NeurIPS 2025 · similarity 0.87)
- AETTA: Label-Free Accuracy Estimation for Test-Time Adaptation (CVPR 2024 · 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.