Large language models make it cheap to run influence operations whose messages never repeat, yet real platforms offer almost no ground truth for testing whether coordination detectors can find them. The problem is sharpest on Telegram, where crisis discourse is dense and moderation data is scarce.
The preprint (arXiv:2610.23944) introduces TeleTwin, a calibrated digital twin of Telegram crisis discourse. It is designed as a ground-truth testbed: researchers can inject known LLM-generated messages and then measure how well coordination detectors spot them, without the ambiguity of real-world labels.
Because this is a single preprint, the claims are not yet peer-reviewed. The value is in the proposed method: a controlled environment that mimics real Telegram discourse while providing the labels that live platforms lack.