文章采用日记法收集来自16个国民经济行业的真实工作场景下具有学习特征的搜索任务实例，结合任务概念框架对搜索任务进行表征，构建面向“搜索即学习”研究的多维任务集，并进一步剖析了真实工作场景下搜索任务的属性特征，对任务相关特征、参与者认知变化、行为及关系进行探讨，根据参与者的学历层次、完成任务的认知层次，以及在完成任务时所处的时间压力进行分组比较三组变量间的关系。研究发现：参与者的任务执行熟悉度与任务感兴趣程度，所需信息类型的数量显著影响认知变化，任务执行者特征对认知变化与行为的影响在不同情境下均存在显著差异。本研究开发的任务集可作为“搜索即学习”研究的公共研究资源，以及评估生成式人工智能技术复杂任务处理能力的评测集，为开发面向真实工作任务情境与高效的信息系统提供依据。

In this paper，the diary method is used to collect the search task examples with learning characteristics from 16 national economic industries in real work context，and the search task is characterized in combination with the task conceptual framework to construct a multidimensional task library for "search-as-learning" research，and further analyze the attributes and characteristics of search tasks in real work context. Task-related characteristics，cognitive changes and behaviors of participants，and relationships of them were discussed，and the relationships among the three groups of variables were compared according to the participants’ educational level，cognitive level of completing tasks，and time pressure of the participants when completing the task. The results showed that the participants' familiarity and interest in the task，and the amount of required information types significantly affected the cognitive changes，and the influence of the characteristics of the task performers on the cognitive changes and behaviors was significantly different in different contexts. The task library developed in this study can serve as a public research resource for "search as learning" research，as well as an evaluation set for evaluating the complex task processing ability of AIGC technology，providing a basis for developing efficient information systems for real-world work task contexts.