SIEVE: Multimodal Dataset Pruning using Image-Captioning Models - View it on GitHub