Qilin
收藏资源简介:
Qilin数据集是由清华大学和小红书公司合作创建的多模态信息检索数据集。该数据集包含来自小红书平台上超过3000万月活跃用户的APP级信息发现会话,涵盖了搜索、DQA(深度查询回答)和推荐等多种场景。Qilin特色在于收集了丰富的APP级上下文信号和真实用户反馈,有助于深入分析用户状态转换、回访和查询重构等行为,以建模用户满意度和长期留存。数据集中的用户会话包含文本和图像内容,适用于多种检索场景下的复杂查询意图理解和会话级跨模态匹配任务。
The Qilin Dataset is a multimodal information retrieval dataset co-created by Tsinghua University and Xiaohongshu. This dataset contains app-level information discovery sessions from over 30 million monthly active users on the Xiaohongshu platform, covering multiple scenarios including search, Deep Query Answering (DQA), and recommendation. A distinctive feature of the Qilin Dataset is its collection of rich app-level contextual signals and real user feedback, which supports in-depth analysis of user behaviors such as state transitions, revisit behaviors, and query reformulation, enabling the modeling of user satisfaction and long-term retention. The user sessions in the dataset include both text and image content, making it applicable for complex query intent understanding and session-level cross-modal matching tasks across various retrieval scenarios.




