allenai/

Molmo2-MultiImageQA

Molmo2-MultiImageQA is a multi-image extension of PixMo-AskModelAnything, an instruction-tuning dataset for vision-language models. It contains human-authored question-answer pairs over multiple images with long-form answers. Molmo2-MultiImageQA is part of the Molmo2 dataset collection and was used to train the Molmo2 family of models. Quick links:

📃 Paper 🎥 Blog with Videos

import datasets

train_dataset =… See the full description on the dataset page: https://huggingface.co/datasets/allenai/Molmo2-MultiImageQA.