VideoInstruct-100K
收藏arXiv2025-09-30 收录
官方服务：
资源简介：
该数据集包含了大约10万对视频指令，用于对VaQuitA模型进行微调，旨在提升模型将视频特征与文本指令对齐的能力，以此来增强模型在视频指令理解任务上的表现。
This dataset contains approximately 100,000 video-instruction pairs, which are utilized for fine-tuning the VaQuitA model. The objective is to improve the model's capability to align video features with textual instructions, thus enhancing its performance on video instruction understanding tasks.
提供机构：
MBZUAI
