Dataset Groups Activity Stream Agqa: A Benchmark for Compositional Spatio-Temporal Reasoning The AGQA benchmark is a visual dataset comprising 192M hand-crafted questions about 9.6K videos from the Charades dataset. BibTex: @dataset{Madeleine_Grunde-McLaughlin_and_Ranjay_Krishna_and_Maaneesh_Agrawala_2025, abstract = {The AGQA benchmark is a visual dataset comprising 192M hand-crafted questions about 9.6K videos from the Charades dataset.}, author = {Madeleine Grunde-McLaughlin and Ranjay Krishna and Maaneesh Agrawala}, doi = {10.57702/a64c5emu}, institution = {No Organization}, keyword = {'Charades dataset', 'compositional spatio-temporal reasoning', 'video question answering'}, month = {jan}, publisher = {TIB}, title = {Agqa: A Benchmark for Compositional Spatio-Temporal Reasoning}, url = {https://service.tib.eu/ldmservice/dataset/agqa--a-benchmark-for-compositional-spatio-temporal-reasoning}, year = {2025} }
