Yizhuo Xu, Chaojian Yu, Yuanjie Shao, Tongliang Liu, Qinmu Peng, Xinge You · ACM International Conference on Multimedia (ACM MM) 2026 · 2026
DOI: 10.1145/3767308.3836191
Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).
Composed Image Retrieval (CIR) aims to retrieve target images using reference images and modification texts. Current CIR methods rely on fine-tuning vision-language pre-trained models. However, we find these approaches commonly suffer from severe overfitting, posing challenges for CIR with limited triplet data. To understand this issue, we present a systematic study of overfitting in VLP-based CIR, revealing a significant and previously overlooked generalization gap across models and datasets. Motivated by these findings, we introduce WRF4CIR, a Weight-Regularized Fine-tuning network for CIR. Specifically, during fine-tuning, we apply adversarial perturbations to the model weights for regularization, where perturbations are generated in the opposite direction of gradient descent. Intuitively, WRF4CIR increases the difficulty of fitting the training data, which helps mitigate overfitting in CIR under limited triplet supervision. Extensive experiments on benchmark datasets demonstrate that WRF4CIR significantly narrows the generalization gap and achieves substantial improvements over existing methods.
No comments yet — start the discussion below.