IntentQA

该数据集名为IntentQA，是一个专注于意图推理的多选题视觉问答（VQA）数据集，它基于NeXT-QA构建而成，旨在测试模型在推理问题上的表现。该数据集包含了与因果和时间相关问题相关的视频，以及相应构造的问题和答案。其规模源自NeXT-QA，所属的任务类型为视频问答。

This dataset, named IntentQA, is a multiple-choice visual question answering (VQA) dataset focused on intentional reasoning. It is constructed based on NeXT-QA and designed to evaluate model performance on reasoning-related tasks. The dataset contains videos related to causal and temporal questions, along with correspondingly formulated questions and answers. Its scale is derived from NeXT-QA, and its affiliated task type falls under video question answering.

搜集汇总

背景与挑战

背景概述

IntentQA是一个专注于视频意图推理的数据集，包含原始视频、预计算特征和详细的QA注释，用于研究视频问答中的逻辑推理能力。该数据集在ICCV2023论文中提出，并提供了多个模型的性能基准。