ProVLA: Compositional Image Search with Progressive Vision-Language Alignment and Multimodal Fusion | IEEE Conference Publication | IEEE Xplore