: It could be part of a private dataset for video reasoning benchmarks like Video-TT or NA-VQA, which use long-form video to test narrative understanding.
