FediData: Multi-Modal Fediverse (Mastodon) Dataset Analyzed
A Mastodon (Fediverse) dataset of 725,282 posts from 64,345 users across 493 instances, collected by Gao et al. for CIKM 2025. Each record contains post text, image URLs, and following/follower edge relationships with a 12,548-user manual bot/human/unclassifiable label subset. The dataset is notable for its multilingual corpus with 92.8% image attachment rate and 4-way bot classification scheme across a decentralized federated network. Intended for research into bot detection, decentralized social network structure, and multi-modal content analysis.
Fediverse
Mastodon