Keywords: Vision-Language Models, Cross-Modal Retrieval, Representation Learning, Prostate Cancer Grading, Digital Pathology
TL;DR: A straightforward and effective approach for balancing classification accuracy and vision-text alignment within cross-modal vision models.
Abstract: Despite the promising capabilities of vision-language models (VLMs) in diverse tasks, recent studies reveal that they struggle with the fundamental task of image classification. In this study, we explore leveraging state-of-the-art task-specific classification models as a foundation for VLMs, aiming to preserve strong classification performance. Specifically, we assess the impact of contrastive tuning to enable cross-modal retrieval capabilities on a Vision Transformer (ViT) model trained for multi-label classification on natural images and a Hierarchical Vision Transformer (H-ViT) trained for prostate cancer grading in Whole-Slide Images (WSIs). Our results demonstrate that contrastive fine-tuning creates a clear trade-off: classification accuracy rapidly deteriorates toward zero as vision-text alignment improves. By balancing task-specific and contrastive objectives in the loss function during fine-tuning, we achieve competitive slide-level retrieval performance while maintaining classification accuracy. Our code is available on https://github.com/DIAGNijmegen/tradeoff_classification_alignment.git.
Primary Subject Area: Transfer Learning and Domain Adaptation
Secondary Subject Area: Unsupervised Learning and Representation Learning
Paper Type: Methodological Development
Registration Requirement: Yes
Midl Latex Submission Checklist: Ensure no LaTeX errors during compilation., Created a single midl25_NNN.zip file with midl25_NNN.tex, midl25_NNN.bib, all necessary figures and files., Includes \documentclass{midl}, \jmlryear{2025}, \jmlrworkshop, \jmlrvolume, \editors, and correct \bibliography command., Did not override options of the hyperref package, Did not use the times package., All authors and co-authors are correctly listed with proper spelling and avoid Unicode characters., Author and institution details are de-anonymized where needed. All author names, affiliations, and paper title are correctly spelled and capitalized in the biography section., References must use the .bib file. Did not override the bibliographystyle defined in midl.cls. Did not use \begin{thebibliography} directly to insert references., Tables and figures do not overflow margins; avoid using \scalebox; used \resizebox when needed., Included all necessary figures and removed *unused* files in the zip archive., Removed special formatting, visual annotations, and highlights used during rebuttal., All special characters in the paper and .bib file use LaTeX commands (e.g., \'e for é)., Appendices and supplementary material are included in the same PDF after references., Main paper does not exceed 9 pages; acknowledgements, references, and appendix start on page 10 or later.
Submission Number: 202
Go to MIDL 2025 Conference homepage