Abstract
Vision-language models such as CLIP achieve strong zero-shot classification, yet under distribution shift, visual embeddings drift from fixed text embeddings. Training-free calibration avoids the per-sample optimization of prompt learning, but prior feature calibration gives each image the full bias of one hard cluster. We propose Domain Recentering with Confidence Calibration (DRC), a training-free method adapting CLIP from a set of unlabeled target images. DRC fits a Gaussian mixture once and subtracts from each embedding a posterior-weighted average of component means. It then removes residual class preference with a log-prior correction, estimating the prior from confidence-weighted predictions. Among compared methods, DRC achieves the highest average accuracy on cross-domain datasets, exceeding zero-shot CLIP by 4.13 and 5.07 points with ViT-B/16 and ResNet-50, with gains over CLIP also holding under ImageNet distribution shifts.
Keywords
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Seol, Y., Shin, J., Yoon, H., & Hwang, U. (2026). Domain Recentering and Confidence-Weighted Prior Calibration for Vision-Language Models. https://omanscience.com/en/articles/domain-recentering-and-confidence-weighted-prior-calibration-for-vision-language-models
MLA 9
Seol, Youngeun, et al. "Domain Recentering and Confidence-Weighted Prior Calibration for Vision-Language Models." https://omanscience.com/en/articles/domain-recentering-and-confidence-weighted-prior-calibration-for-vision-language-models.
Chicago (author–date)
Seol, Youngeun, Jimin Shin, Heeseo Yoon, and Uiwon Hwang. 2026. "Domain Recentering and Confidence-Weighted Prior Calibration for Vision-Language Models." https://omanscience.com/en/articles/domain-recentering-and-confidence-weighted-prior-calibration-for-vision-language-models.
Harvard
Seol, Y., Shin, J., Yoon, H. and Hwang, U. (2026) 'Domain Recentering and Confidence-Weighted Prior Calibration for Vision-Language Models', Available at: https://omanscience.com/en/articles/domain-recentering-and-confidence-weighted-prior-calibration-for-vision-language-models.
Vancouver
Seol Y, Shin J, Yoon H, Hwang U. Domain Recentering and Confidence-Weighted Prior Calibration for Vision-Language Models. https://omanscience.com/en/articles/domain-recentering-and-confidence-weighted-prior-calibration-for-vision-language-models
IEEE
Y. Seol, J. Shin, H. Yoon, and U. Hwang, "Domain Recentering and Confidence-Weighted Prior Calibration for Vision-Language Models," https://omanscience.com/en/articles/domain-recentering-and-confidence-weighted-prior-calibration-for-vision-language-models.