Concurrent Streaming, Viewer Transfers, and Audience Loyalty in a Creator Ecosystem: A Minute-Level Longitudinal Study

The document is an arXiv feed with several papers; the most security-relevant contributions are: (1) a controlled IRB-approved study of reinforcement-learning (RL)–powered adaptive social bots that dynamically evade detection, comparing human detectors, state-of-the-art AI detectors (including ML and LLMs), and hybrid human–AI ensembles. Key finding: hybrid aggregation of human reports with AI predictions (and retraining protocols using human supervision) outperforms humans alone and AI alone, with implications for defenses against adaptive covert social influence operations. (2) Analysis of a