OmniThoughtV is a large-scale multimodal long-chain-of-thought dataset distilled from the FineVision dataset using Alibaba Cloud's AI platform (PAI) distillation toolkit, EasyDistill. This dataset establishes a transparent and reproducible data distillation pipeline, enabling efficient construction of multimodal reasoning chains of thought. Fine-tuning smaller models with this dataset effectively endows them with stronger reasoning capabilities and enhances… See the full description on the dataset page:
https://huggingface.co/datasets/alibaba-pai/OmniThoughtV_Raw_1.8M.