Research map / Label-efficient and robust learning / Test-time adaptation
Bilateral Information-aware Test-time Adaptation for Vision-Language Models
Jingwei Sun, Jianing Zhu, Jiangchao Yao, Gang Niu, Masashi Sugiyama, Bo Han
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 BITTA high-entropy unlearning to BAT | BAT, CLIP ViT-B-16 · CIFAR-10-C | mean accuracy: 73.34 -> 74.78 | Improved |
| add BITTA to BAT | BAT, CLIP ViT-B-32 · CIFAR-100-C | mean accuracy: 36.29 -> 38.23 | Improved |
| add BITTA to BAT | BAT, CLIP ViT-B-16 · ImageNet-C | mean accuracy: 31.06 -> 31.88 | Improved |
| add BITTA to TPT | TPT, ViT-B-16 · CIFAR-10-C | mean accuracy: 63.84 -> 65.37 | Improved |
| add BITTA to TPS | TPS, ViT-B-16 · CIFAR-10-C | mean accuracy: 64.28 -> 65.24 | Improved |
| add BITTA to CTPT | CTPT, ViT-B-16 · CIFAR-10-C | mean accuracy: 61.84 -> 62.61 | Improved |
| add BITTA to SAR | SAR, ViT-Base · ImageNet-C Gaussian | accuracy: 52.83 -> 56.27 | Improved |
| add BITTA to BAT | BAT, ViT-B-16 · CIFAR-10-C | ECE: 15.25 -> 12.93 | Improved |
| unlearning high-entropy samples instead of weight decay regularization (ablation) | BAT · CIFAR-10-C | mean accuracy: 73.79 -> 74.78 | Improved |
| unlearning instead of distribution dissolve regularization (ablation) | BAT · CIFAR-10-C | mean accuracy: 74.37 -> 74.78 | Improved |
| BITTA effect on individual corruption types | BAT, ViT-B-16 · ImageNet-C Frost/Brightness/Elastic | accuracy: -1.02, -0.64, -1.22 | Mixed |
| use a different prompt template (ablation) | BITTA · CIFAR-10-C ViT-B-16 | accuracy: 74.78 -> 75.55 | No clear effect |
Closest papers by meaning
- The Illusion of Progress? A Critical Look at Test-Time Adaptation for Vision-Language Models (NeurIPS 2025 · similarity 0.93)
- Efficient Test-Time Adaptation of Vision-Language Models (CVPR 2024 · similarity 0.92)
- Frustratingly Easy Test-Time Adaptation of Vision-Language Models (NeurIPS 2024 · similarity 0.92)
- Efficient Test-time Adaptation through Candidate Verification and Divergence Shifts (NeurIPS 2026 · similarity 0.92)
- DOTA: Distributional Test-time Adaptation of Vision-Language Models (NeurIPS 2025 · similarity 0.92)
- BoostAdapter: Improving Vision-Language Test-Time Adaptation via Regional Bootstrapping (NeurIPS 2024 · similarity 0.91)
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.
