General
VHELM
Original VHELM vision-language requests, image inputs, recorded outputs and release-declared task grades.
93,127items
112subjects
The captured HELM code is Apache-2.0. Underlying visual datasets retain their own source licenses; the public run exports do not supply a single replacement license for all inputs.license
vision_languagedomain
generaldomain
imagemodality
textmodality
item-level responses released
Response matrix
Loading response matrix…
lowhighUnobserved
Scale: Item-specific grading scales