Records found: 3
Hugging Face open-sourced FineVision — a 24M-sample multimodal dataset that boosts VLM performance across benchmarks while keeping data leakage minimal.
'Meta CLIP 2 trains CLIP models from scratch on native worldwide image-text pairs and shows that scaling metadata, curation, and model capacity reduces the curse of multilinguality while improving zero-shot and multilingual benchmark results.'
Yandex has launched Alchemist, a compact supervised fine-tuning dataset that significantly improves text-to-image model quality by selecting high-impact image-text pairs using a novel model-guided approach.