A post claims a more than 100-fold question-answering speedup for VideoJev versus Thinking.
A post introduces VideoJev as a model for video understanding and interaction and says its 4B and 35B variants have been open-sourced. It claims a more than 100-fold QA speedup versus Thinking, plus 6.7- and 10.8-point mIoU gains over 4B and MoE backbones, respectively.
56.3K
2 Sources, first seen ago
@arankomatsuzakiVideoJev: System One Model for Video Understanding and Interaction - >100x QA speedup vs. Thinking - +6.7 / +10.8 pts mIoU gain over 4B / MoE backbones - Opensourced 4B and 35B variants
@_akhaliqVideoJev System One Model for Video Understanding and Interaction paper: https://huggingface.co/papers/2610.18615@arankomatsuzakiVideoJev: System One Model for Video Understanding and Interaction - >100x QA speedup vs. Thinking - +6.7 / +10.8 pts mIoU gain over 4B / MoE backbones - Opensourced 4B and 35B variants@_akhaliqVideoJev System One Model for Video Understanding and Interaction paper: https://huggingface.co/papers/2610.18615—
Not ranked yet
—
Not ranked yet
