datasetdocvqa.org
docvqa.org
Access: open (login). Document VQA datasets
datasethuggingface.co
huggingface.co
Access: research-only. Video-text dataset
datasetmmmu-benchmark.github.io
mmmu-benchmark.github.io
Access: open. Multimodal reasoning benchmark
datasetlaion.ai
laion.ai
Still an index of URLs/alt-text; reconstructed images have separate rights considerations.