CG-Bench
收藏官方服务：
资源简介：
CG-Bench is a new benchmark for clue-grounded question answering (QA) in long videos, addressing the limitations of existing benchmarks that focus on short videos and rely on multiple-choice questions (MCQs), which allow models to answer by elimination rather than true understanding. It includes 1,219 manually curated videos across 14 primary, 171 secondary, and 638 tertiary categories, making it the largest benchmark for long video analysis. The dataset contains 12,129 QA pa
CG-Bench是一款面向长视频场景下线索锚定问答（clue-grounded question answering, QA）的全新基准测试集，旨在解决现有基准测试的局限性：现有基准测试仅聚焦短视频，且依赖多项选择题（multiple-choice questions, MCQs）的作答形式，此类测试允许模型通过排除法得出答案，而非实现对视频内容的真正理解。该数据集收录了1219条经人工整理的视频，涵盖14个主类别、171个次类别与638个三级类别，是目前规模最大的长视频分析基准测试集。该数据集共包含12129条QA相关内容
提供机构：
lykcl0428创建时间：
2025-02-12
搜集汇总
数据集介绍
以上内容由遇见数据集搜集并总结生成
