MMR-V is a new video QA benchmark requiring long-range, multi-frame reasoning, on which the best AI model scores 52.5% versus 86% for humans.
Research of intelligent home secu- rity surveillance system based on zigbee,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
MMR-V: What's Left Unsaid? A Benchmark for Multimodal Deep Reasoning in Videos
MMR-V is a new video QA benchmark requiring long-range, multi-frame reasoning, on which the best AI model scores 52.5% versus 86% for humans.