One Token per Highly Selective Frame: Towards Extreme Compression for Long Video Understanding

Devs

One Token per Highly Selective Frame: Towards Extreme Compression for Long Video Understanding | Read Paper on Bytez