Researcher, Amazon
1 paper at NeurIPS 2025
Enhancing long video understanding via extreme compression by reducing each selected frame to a single token.