This repo contains the annotation data and evaluation code for the paper "MLVU: A Comprehensive Benchmark for Multi-Task Long Video Understanding".
🆕 7/28/2024: The data for the MLVU-Test set has been released (🤗 Link)! The test set includes 11 different tasks, featuring our newly added Sports Question Answering (SQA, single-detail LVU) and Tutorial… See the full description on the dataset page:
https://huggingface.co/datasets/MLVU/MVLU.