TrustAIRLab/Moltbook

这是一个名为Moltbook Dataset with Annotations的数据集，包含从Moltbook收集的超过40,000个帖子和12,000个子版块（submolts）数据。数据集中的44,376个帖子由GPT-5.2标注了9个内容类别和5个毒性级别。内容类别包括身份、技术、社交、经济、观点、推广、政治、垃圾邮件和其他。毒性级别从安全到恶意分为5个等级。此外，数据集还包含12,209个子版块的信息，如名称、显示名称、描述、订阅者数量、创建时间和最后活动时间等。

This is the annotated dataset of over 40K posts and 12K submolts collected from Moltbook presented in the paper ["Humans welcome to observe": A First Look at the Agent Social Network Moltbook]. Specifically, it contains 44,376 GPT-5.2-annotated posts (9 content categories and 5 toxicity levels) and 12209 submolts collected from Moltbook. The content categories include Identity, Technology, Socializing, Economics, Viewpoint, Promotion, Politics, Spam, and Others. The toxicity levels range from Safe to Malicious, categorized into 5 levels. The dataset also includes information about submolts such as name, display name, description, subscriber count, created at, and last activity at.