Training Tags Network
Tag Groups
These interactive visualizations are part of the study "Perpetual Misogyny: How Gendered Tropes Shape Text-To-Image AI Personalization". The research explores how user-driven content and tagging behavior on the model-sharing hub CivitAI has evolved over a two-year span.
The first network visualizes tag co-occurrence patterns—how users promote their models by grouping descriptive tags together. The second visualization maps semantic structures in the training data of over 11,000 LoRA models, revealing latent biases and conceptual clusters.