Benchmark RadarNARU
NARU evaluates multimodal models on Japanese long-form video understanding across narrative evolution (character evolution, sequential flow, plot progression, thematic development) and cultural nuance (aizuchi, reading the air, subtext, cultural context, sentiment) via multiple-choice QA.
- Released
- 2026-08-13
- Readiness
- Runnable
- Primary field
- General AI
Why it matters
NARU fills a gap in video QA benchmarks by jointly testing long-range narrative tracking and culturally grounded reasoning, which are absent in existing short-horizon or English-centric benchmarks. It provides a rigorous, reproducible protocol for measuring model capabilities in high-context media, supporting model development and product evaluation.
Motivation
Long-form video understanding encompasses tasks that go beyond retrieving isolated events, including tracking an evolving narrative and interpreting social meaning that may remain implicit.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.