FISSION improves bot detection through label augmentation
September 23, 2026
FISSION generates training labels for bot detection by splitting account activity into positively labeled sub-accounts. This method enables models to learn embeddings that better group coordinated influence operations and Wikipedia sockpuppets.
HOW THIS AFFECTS YOU
●
builderYou can use label augmentation to train more robust detection models when ground-truth labels are scarce.
●
policyThis provides a technical path for identifying more sophisticated coordinated influence operations.