Research map / Label-efficient and robust learning / Test-time adaptation
AcTTA: Rethinking Test-Time Adaptation via Dynamic Activation
Hyeongyu Kim, Geonhui Han, Dosik Hwang
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 learnable activation (AcTTA) to TENT | TENT · CIFAR10-C WRN-28 | error rate (%): 18.51 -> 17.03 | Improved |
| add AcTTA to SAR | SAR · ImageNet-C ViT-B/16 | error rate (%): 53.90 -> 54.17 | Hurt |
| add AcTTA to DeYO | DeYO · ImageNet-C ViT-B/16 | error rate (%): 47.73 -> 49.64 | Hurt |
| add AcTTA to ROID | ROID · ImageNet-C ResNet50 (GN) | error rate (%): 64.39 -> 55.87 | Improved |
| add AcTTA to TENT at batch size 4 | TENT · ImageNet-C ViT-B/16 (batch 4) | error rate (%): 89.92 -> 62.07 | Improved |
| freeze BN affine, adapt only activation parameters (ablation) | AcTTA · CIFAR10-C WRN-28 | error rate (%): 18.06 -> 17.03 | Improved |
| adapt LN affine jointly with activation parameters (ablation) | AcTTA · ImageNet-C ViT-B/16 | error rate (%): 55.30 -> 52.37 | Improved |
| replace activation with PReLU/PGELU (ablation) | Tent · CIFAR10-C WRN-28 | error rate (%): 18.51 -> 29.62 | Hurt |
| replace activation with ACON (ablation) | Tent · CIFAR10-C WRN-28 | error rate (%): 18.51 -> 18.46 | No clear effect |
| replace activation with PAU (ablation) | Tent · ImageNet-C ViT-B/16 | error rate (%): 53.85 -> 99.96 | Hurt |
| make all activations learnable instead of ~50% (ablation) | AcTTA · CIFAR100-C WRN-40 | error rate (%): 34.05 -> 44.97 | Hurt |
| pixel-wise instead of channel-wise activation parameters (ablation) | AcTTA_TENT · CIFAR10-C WRN-28 (batch 128) | error rate (%): 17.03 -> 58.80 | 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 4 of 4 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
- A Layer Selection Approach to Test Time Adaptation (AAAI 2025 · similarity 0.90)
- SURGEON: Memory-Adaptive Fully Test-Time Adaptation via Dynamic Activation Sparsity (CVPR 2025 · similarity 0.90)
- Backpropagation-Free Test-Time Adaptation via Probabilistic Gaussian Alignment (NeurIPS 2025 · similarity 0.88)
- DPCore: Dynamic Prompt Coreset for Continual Test-Time Adaptation (ICML 2025 · similarity 0.88)
- ReservoirTTA: Prolonged Test-time Adaptation for Evolving and Recurring Domains (NeurIPS 2025 · similarity 0.88)
- 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.
