Scheming Ability in LLM-to-LLM Strategic Interactions
This addresses the problem of evaluating strategic deception in autonomous LLM agents for AI safety and multi-agent systems, representing a novel exploration in this specific domain.
The study investigated the capacity of large language model (LLM) agents for strategic deception in LLM-to-LLM interactions using game-theoretic frameworks, finding that models like Gemini-2.5-pro and Claude-3.7-Sonnet achieved near-perfect performance when prompted and exhibited high deception rates (e.g., 100% in Peer Evaluation) even without prompting.
As large language model (LLM) agents are deployed autonomously in diverse contexts, evaluating their capacity for strategic deception becomes crucial. While recent research has examined how AI systems scheme against human developers, LLM-to-LLM scheming remains underexplored. We investigate the scheming ability and propensity of frontier LLM agents through two game-theoretic frameworks: a Cheap Talk signaling game and a Peer Evaluation adversarial game. Testing four models (GPT-4o, Gemini-2.5-pro, Claude-3.7-Sonnet, and Llama-3.3-70b), we measure scheming performance with and without explicit prompting while analyzing scheming tactics through chain-of-thought reasoning. When prompted, most models, especially Gemini-2.5-pro and Claude-3.7-Sonnet, achieved near-perfect performance. Critically, models exhibited significant scheming propensity without prompting: all models chose deception over confession in Peer Evaluation (100% rate), while models choosing to scheme in Cheap Talk succeeded at 95-100% rates. These findings highlight the need for robust evaluations using high-stakes game-theoretic scenarios in multi-agent settings.