Vision-Language Foundation Models With Hierarchical Motion Memory for Long Video Question Answering and Event Localization. JDIAS 2026, 1 (1).