Paper summary
Paper summary: A Very Big Video Reasoning Suite
VBVR is a video-reasoning dataset of 200 curated tasks and over a million clips, about 1,000× larger than earlier ones, with a rule-based benchmark used for an early study of how video reasoning scales.
- Video models have mostly been judged on visual quality; this dataset targets reasoning about continuity, interaction and causality.
- The benchmark uses rule-based, human-aligned scorers instead of model-based judges, for reproducible results.