Announcement
MF² benchmark aims to test vision-language models’ grasp of movie narratives
Manos Zaranis says MF² was accepted at EMNLP 2026 in Budapest and targets comprehension beyond shallow recall.
TLDR
Project lead Manos Zaranis introduced MF² (Movie Facts and Fibs) as a benchmark for whether vision-language models understand movie narratives, not just recall details. He says it was accepted at EMNLP 2026 in Budapest. A collaborator says the project grew out of an ELLIS workshop where participants watched classic cinema, and credits Zaranis with leading a project with 20-plus co-authors.
Combined views
—
3 Sources, first seen ago
