Multimodal dataset

The dataset contains 4-turn dialogs, each featuring a sequence of question-answer pairs designed to explore the content, context, and relationships within a given image. It is well-suited for training

What’s in this dataset:

University-level math problems with step-by-step solutions* (* natural language step-by-step explanations can be added by request)
Format: LaTeX formulas with comments
Multiple solutions provided for ~10%   of the problems