Fetching the paper…

TESTA: Temporal-Spatial Token Aggregation for Long-form Video-Language Understanding · Around