Vulnerability of LLMs' Stated Beliefs? LLMs Belief Resistance Check Through Strategic Persuasive Conversation Interventions
This research tested how susceptible Large Language Models (LLMs) are to persuasion using strategic conversations. By analyzing different models and domains, the study found that LLM belief changes depend heavily on model size and the specific persuasive tactics used. It concludes that defenses must focus on interaction-aware strategies rather than just message content.