Minerva

Minerva is a benchmark for complex video reasoning, evaluating models on multi-step reasoning over long and information-dense video content.

id minerva · max 1 · 4 models reported