Pathological diagnosis is the gold standard of tumor diagnosis and the cornerstone of clinical treatment. Artificial intelligence (AI) has made significant progress in detecting tumor tissues and tumor cells, which contributes to accurately, efficiently and quantitatively identifying tumor cells and/or tumor characteristics, leading to improved efficiency of pathologists and making up for the shortage of pathologists. The premise of pathological AI is efficient and accurate labeling, which is to outline the tumor cells of various types and different degrees of differentiation. To promote the standardization and data quality control of labeling, experts of oncology, pathology, electronic information science and other fields jointly discussed the pathological data set construction and data quality control for solid tumor, and thus an expert group was formed for a future expert consensus. Our group is dedicated to the construction of the AI-based standardized pathological data set for solid tumor. This paper introduces the primary opinions reached by our group in the process of tumor cell labeling from multiple aspects, including specimen source, labeling team, labeling rules, labeling process, quality control, and solutions for difficult cases.