Research map / Label-efficient and robust learning / Test-time adaptation
MixTTA: Low-Rank Cross-Channel Mixing for Reliable Test-Time Adaptation
Mansoo Jung, Youngwook Kim, Jungwoo Lee
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 MixTTA low-rank cross-channel mixing | Tent (ResNet-GN) · ImageNet-C sev5 mild | accuracy: 28.8 -> 32.4 | Improved |
| add MixTTA | EATA (ResNet-GN) · ImageNet-C sev5 mild | accuracy: 45.5 -> 46.2 | Improved |
| add MixTTA | SAR (ResNet-GN) · ImageNet-C sev5 mild | accuracy: 37.2 -> 37.5 | No clear effect |
| add MixTTA | DeYO (ResNet-GN) · ImageNet-C sev5 mild | accuracy: 45.3 -> 44.4 | Hurt |
| add MixTTA | ReCAP (ResNet-GN) · ImageNet-C sev5 mild | accuracy: 46.8 -> 47.9 | Improved |
| add MixTTA under online imbalanced label shifts | EATA (ResNet-GN) · ImageNet-C sev5 label shifts | accuracy: 33.1 -> 41.3 | Improved |
| add MixTTA under label shifts | DeYO (ResNet-GN) · ImageNet-C sev5 label shifts | accuracy: 42.3 -> 42.1 | No clear effect |
| add MixTTA at batch size 1 | Tent / EATA (ResNet-GN) · ImageNet-C sev5 batch size 1 | accuracy: Tent 23.9 -> 30.2; EATA 37.2 -> 41.5 | Improved |
| add MixTTA at batch size 1 | DeYO (ResNet-GN) · ImageNet-C sev5 batch size 1 | accuracy: 47.0 -> 45.4 | Hurt |
| add MixTTA under mixed corruption shifts | Tent (ResNet-GN) · ImageNet-C sev5 mixed shifts | accuracy: 14.1 -> 10.5 | Hurt |
| add MixTTA under mixed corruption shifts | DeYO (ResNet-GN) · ImageNet-C sev5 mixed shifts | accuracy: 37.2 -> 40.1 | Improved |
| add MixTTA on stylized domain gap at batch size 1 | Tent / EATA / SAR (ResNet-GN) · ImageNet-Sketch batch size 1 | accuracy: Tent 28.5 -> 32.0; EATA 29.5 -> 33.3; SAR 31.1 -> 33.8 | Improved |
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
- SNAP: Low-Latency Test-Time Adaptation with Sparse Updates (NeurIPS 2025 · similarity 0.86)
- A Layer Selection Approach to Test Time Adaptation (AAAI 2025 · similarity 0.86)
- Buffer layers for Test-Time Adaptation (NeurIPS 2025 · similarity 0.86)
- Everything to the Synthetic: Diffusion-driven Test-time Adaptation via Synthetic-Domain Alignment (CVPR 2025 · similarity 0.86)
- Un-Mixing Test-Time Normalization Statistics: Combatting Label Temporal Correlation (ICLR 2024 · similarity 0.86)
- Unraveling Batch Normalization for Realistic Test-Time Adaptation (AAAI 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.
