HARK: Hierarchical Agentic Retrieval with Keyframing for Video Understanding (Student Abstract)
Abstract
Current video understanding models struggle with temporal reasoning and efficient processing while balancing detail preservation with computational efficiency. We propose a hierarchical memory system that segments videos into action and scene units, combined with question-aware agentic keyframe selection. Our method achieves 70.3% overall accuracy on VideoMME short video benchmarks.
Fields of Study & Subjects
Multimodal Machine Learning Applications
Human Pose and Action Recognition
Video Analysis and Summarization
Publication details
- Journal / source
- Underline Science Inc.
- Publication date
- 7 January 2026
- Language
- EN