Despite their wide-ranging benefits, LLM-based agents deployed in open environments can be exploited to produce manipulative material. In this study, we task LLMs with propaganda objectives and analyze their outputs using two domain-specific models: one that classifies text as propaganda or non-propaganda, and another that detects rhetorical techniques of propaganda (e.g., loaded language, appeals to fear, flag-waving, name-calling). Our findings show that, when prompted, LLMs exhibit propagandistic behaviors and use a variety of rhetorical techniques in doing so. We also explore mitigation via Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and ORPO (Odds Ratio Preference Optimization). We find that fine-tuning significantly reduces their tendency to generate such content, with ORPO proving most effective.

翻译：尽管基于LLM的智能体在开放环境中部署具有广泛益处，但可能被利用来产生操纵性内容。在本研究中，我们赋予LLM宣传目标，并使用两个领域特定模型分析其输出：一个将文本分类为宣传或非宣传，另一个检测宣传的修辞技巧（例如，情绪化语言、恐惧诉求、挥舞旗帜、人身攻击）。我们的发现表明，在被提示时，LLM表现出宣传行为并在此过程中运用多种修辞技巧。我们还探索了通过监督微调（SFT）、直接偏好优化（DPO）和ORPO（比值偏好优化）进行缓解。我们发现，微调显著降低了其生成此类内容的倾向，其中ORPO证明最为有效。