🧪 Multi-model · Multilingual · Full provenance

Controlled AI news stimuli,
built for research.

RogueGPT is a modular, model-agnostic framework for generating diverse, reproducible AI news stimuli under controlled experimental conditions - connecting static text generation with agent-based simulation of misinformation diffusion.

View on GitHub Read the Paper RogueGPT controlled AI stimulus generation pipeline for fake news research

From static generation to agentic campaigns

An extensible architecture across LLM families, languages, styles and formats.

🤖

Multi-Model

Model-agnostic across GPT-4, LLaMA, Mistral and DeepSeek - not locked to a single generator like GROVER.

🌍

Multilingual & Controllable

Generate across languages, journalistic styles and content formats with parameter-rich control.

🕸️

Agent-Based Simulation

Standardizes the "seeding" phase of misinformation diffusion models for reproducible experiments.

🔗

Full Provenance

Every stimulus is tracked end-to-end, linking generation to human-perception data in JudgeGPT.

Built for four communities: misinformation researchers needing controlled stimuli, red-teamers stress-testing detectors, agent-based simulation researchers needing reproducible seeds, and human-factors researchers studying perception.

Citation

RogueGPT is published in the Journal of Open Source Software. If you use it, please cite:

@article{loth2026roguegptjoss,
    author    = {Loth, Alexander and Kappes, Martin and Pahl, Marc-Oliver},
    title     = {{RogueGPT}: A Controlled Stimulus Generation Framework for
                 News Authenticity Research},
    journal   = {Journal of Open Source Software},
    year      = {2026},
    volume    = {11},
    number    = {125},
    pages     = {11219},
    publisher = {The Open Journal},
    doi       = {10.21105/joss.11219},
    url       = {https://doi.org/10.21105/joss.11219}
}

Controlled AI news stimuli,built for research.