Simple is Better than Complex: A Representation-centric Perspective for Prompting-based Vision-Language FusionPublished in Transactions on Machine Learning Research (TMLR), 2026Share on Twitter Facebook LinkedIn Previous Next