Wiki-CoR
收藏资源简介:
Wiki-CoR数据集是由亚马逊团队创建的,用于评论检索任务(CoR)。该数据集结合了组成检索和视觉问题回答的特点,旨在回答给定的问题并生成额外的文本响应,以提供关于视觉信息的进一步澄清和细节。数据集中的实体是图像或图像和文本数据的组合,用于生成答案。该数据集的设计考虑了人类学习过程中视觉和文本联合考察的优势,有望促进这一领域的研究。
The Wiki-CoR dataset was created by the Amazon team for the Comment Retrieval (CoR) task. It combines the characteristics of compositional retrieval and visual question answering, aiming to answer given questions and generate additional textual responses to provide further clarification and details about visual information. The entities in this dataset are images or combinations of image and text data, which are used for answer generation. The design of this dataset takes into account the advantages of joint visual and textual learning in human learning processes, and is expected to promote research in this field.

- 1UniCoRN: Unified Commented Retrieval Network with LMMs亚马逊 · 2025年



