Models and dataset for the work: "Advancing Compositional Awareness in CLIP with Efficient Fine-Tuning"