ImageJev
A task collection for evaluating the decision-making capabilities of vision models, requiring models to output explicit judgments or action choices based on image content rather than merely performing recognition and description.
Timeline (last 90 days)
Fovea-4B achieves leading performance on the ImageJev task.
ImageJev is a task collection for measuring whether a model completes decision-type outputs based on image content.
Fovea-4B's optimization objective is concentrated on decision-type tasks such as ImageJev, unlike general-purpose vision-language models.
All Facts (3)
Fovea-4B achieves leading performance on the ImageJev task.
50%UnverifiedImageJev is a task collection for measuring whether a model completes decision-type outputs based on image content.
50%UnverifiedFovea-4B's optimization objective is concentrated on decision-type tasks such as ImageJev, unlike general-purpose vision-language models.
50%