General
FETV
Released video-generation measurements for four models on 619 prompts, including individual human ratings and five automatic alignment metrics.
20,788items
4subjects
CC-BY-4.0 for the FETV dataset, as declared by the authors. Captured code retains its component-specific upstream licenses.license
generaldomain
textmodality
videomodality
item-level responses released
Response matrix
Loading response matrix…
lowhighUnobserved
Scale: Item-specific grading scales