← 研踪 PaperTrailarXiv:2611.15161 · v1
CineMME: From Dialogue Grounding to Narrative Reasoning in Multimodal Video Understanding
首次提交 2026-11-11 15:58
版本更新 2026-11-11 15:58
雷达观测 2026-11-13 11:11
摘要
Narrative video understanding requires more than recognizing visible events: models must extract evidence from dialogue, bind it to speakers, and use it to infer narrative meaning. Existing video benchmarks evaluate aspects of this problem, including video question answering, audio-visual reasoning, and temporal grounding, but rarely test whether these abilities form a coherent dialogue-to-plot evidence chain within the same clips. We introduce \textbf{CineMME}, a human-verified benchmark designed to diagnose dialogue-grounded plot understanding, supported by approximately $1,600$ expert person-hours of annotation. CineMME contains $975$ curated short narrative video clips and two linked tracks: \textit{CineMME-Grounding}, with $13,554$ timestamped speech segments and $33,083$ speaker face boxes for temporal dialogue localization, transcription, and active-speaker grounding; and \textit{CineMME-Reasoning}, with $1,600$ multiple-choice and $316$ open-ended questions spanning perception, audio, narrative, social, and scene-presentation understanding. Our main evaluation compares $20$ MLLMs on reasoning and $9$ omni-modal models on grounding. The results reveal a pronounced who-what-when decoupling: Gemini-3 Pro localizes dialogue in time ($68.34\%$ tIoU) and transcribes it with a WER of $28.99\%$, yet achieves only $7.74\%$ vIoU for active-speaker grounding while achieving $67.66\%$ MCQ accuracy on questions with extractable answers. Modality ablations further show that vision is not uniformly beneficial: it improves performance on perception-heavy questions but can hurt dialogue-centric reasoning. CineMME provides a compact diagnostic testbed for evaluating how multimodal video models connect dialogue, speakers, visual evidence, and narrative meaning.
作者备注：Accepted to NeurIPS 2026 (Evaluations & Datasets Track). Project page: https://cinemme.github.io/
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-11 15:58 · 最新观测版本
CineMME: From Dialogue Grounding to Narrative Reasoning in Multimodal Video Understanding
Narrative video understanding requires more than recognizing visible events: models must extract evidence from dialogue, bind it to speakers, and use it to infer narrative meaning. Existing video benchmarks evaluate aspects of this problem, including video question answering, audio-visual reasoning, and temporal grounding, but rarely test whether these abilities form a coherent dialogue-to-plot evidence chain within the same clips. We introduce \textbf{CineMME}, a human-verified benchmark designed to diagnose dialogue-grounded plot understanding, supported by approximately $1,600$ expert person-hours of annotation. CineMME contains $975$ curated short narrative video clips and two linked tracks: \textit{CineMME-Grounding}, with $13,554$ timestamped speech segments and $33,083$ speaker face boxes for temporal dialogue localization, transcription, and active-speaker grounding; and \textit{CineMME-Reasoning}, with $1,600$ multiple-choice and $316$ open-ended questions spanning perception, audio, narrative, social, and scene-presentation understanding. Our main evaluation compares $20$ MLLMs on reasoning and $9$ omni-modal models on grounding. The results reveal a pronounced who-what-when decoupling: Gemini-3 Pro localizes dialogue in time ($68.34\%$ tIoU) and transcribes it with a WER of $28.99\%$, yet achieves only $7.74\%$ vIoU for active-speaker grounding while achieving $67.66\%$ MCQ accuracy on questions with extractable answers. Modality ablations further show that vision is not uniformly beneficial: it improves performance on perception-heavy questions but can hurt dialogue-centric reasoning. CineMME provides a compact diagnostic testbed for evaluating how multimodal video models connect dialogue, speakers, visual evidence, and narrative meaning.
查看此版本解读 →查看此版本原文 ↗