SIEVE: Multimodal Dataset Pruning using Image-Captioning Models (CVPR 2024) - View it on GitHub