This is the official implementaion of paper 'Adaptive Keyframe Sampling for Long Video Understanding', which is accepted in CVPR 2025. Multimodal large language models (MLLMs) have enabled open-world ...