Research map / Label-efficient and robust learning / Test-time adaptation
To Adapt or Not to Adapt? Selective Adaptation for Vision-Language Models
Yuwei Liang, Yuwei Liang, Jian Liang, Ran He, Tieniu Tan
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 |
|---|---|---|---|
| select adaptation with CAS instead of MCM | TPT · ImageNet and variants (avg) | AUC: 60.88 -> 89.42 | Improved |
| select adaptation with CAS instead of MCM | ZERO · ImageNet and variants (avg) | AUC: 61.06 -> 92.62 | Improved |
| select adaptation with CAS instead of Energy | TPT · Fine-grained datasets (avg) | AUC: 61.69 -> 87.97 | Improved |
| skip 85% of adaptation with CAS | TPT · ImageNet | accuracy: 68.88 -> 69.03 | Improved |
| skip 85% of adaptation with CAS | TPT · ImageNet | total inference time (h): 7.76 -> 2.55 | Improved |
| skip 85% of adaptation with CAS | ZERO · ImageNet | accuracy: 69.28 -> 69.32 | No clear effect |
| add CAS to calibration-oriented C-TPT | C-TPT · ImageNet and variants (avg) | AUC: 62.49 -> 85.67 | Improved |
| add similarity reweighting to consistency score (ablation) | TPT · Fine-grained datasets (avg) | AUC: 87.65 -> 87.97 | Improved |
| add similarity reweighting to consistency score (ablation) | ZERO · Fine-grained datasets (avg) | AUC: 90.82 -> 91.25 | Improved |
| add similarity reweighting to consistency score (ablation) | TPT · ImageNet and variants (avg) | AUC: 89.14 -> 89.42 | No clear effect |
| add consistency term to similarity score (ablation) | TPT · ImageNet and variants (avg) | AUC: 82.84 -> 89.42 | Improved |
| change AEP function to f(s)=2s (ablation) | TPT · ImageNet and variants (avg) | AEP: 62.19 -> 61.94 | 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 0.92)
- Realistic Test-Time Adaptation of Vision-Language Models (CVPR 2025 · similarity 0.91)
- Frustratingly Easy Test-Time Adaptation of Vision-Language Models (NeurIPS 2024 · similarity 0.91)
- Efficient Test-Time Adaptation of Vision-Language Models (CVPR 2024 · similarity 0.90)
- Efficient Test-time Adaptation through Candidate Verification and Divergence Shifts (NeurIPS 2026 · similarity 0.90)
- DOTA: Distributional Test-time Adaptation of Vision-Language Models (NeurIPS 2025 · similarity 0.89)
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.
