1 paper · 1 filter
Myunsoo Kim, Seongwoong Shim, Byung-Jun Lee
False negatives pose a critical challenge in vision-language pretraining (VLP) due to the many-to-many correspondence between images and texts in large-scale datasets. These false…