Search papers, labs, and topics across Lattice.
This study investigates the long-term dynamics of multi-turn conversations among large language models (LLMs) by examining whether they exhibit attractor-like behaviors in their interactions. By analyzing self-play and mixed-play dyadic debates across seven LLMs and twenty controversial topics, the authors identify model-specific attractors that influence conversational trajectories and stylistic choices. The findings reveal that certain models, such as Claude Haiku, serve as strong attractors, shaping the discourse of other models, which highlights the structured and asymmetric nature of influence in open-ended LLM interactions.
Open-ended LLM conversations reveal model-specific attractors that significantly shape discourse dynamics and influence partner behaviors.
Large language models (LLMs) are increasingly used in open-ended multi-agent settings, but the long-run dynamics of model--model interaction remain poorly understood. We study whether open-ended LLM discussions exhibit attractor-like behavior, i.e. topic-independent stable sets of behaviors which conversations settle into. Across 7 LLMs and 20 controversial topics, we compare self-play and mixed-play dyadic debates, tracking trajectories in representation space, discourse traits, and stances. We find self-play trajectories to be model-specific attractors that draw their conversation partners asymmetrically in mixed-play debates, influencing the other models'stylistic choices and behavior. For example, Claude Haiku is a strong attractor of other models in latent space, corresponding to other models taking on its traits like metacommentary, and models like GPT-4.1 nano are especially malleable. Our results suggest that open-ended LLM interactions are partially predictable from model-specific attractors, but shaped by structured and asymmetric partner influence. Overall, our analysis sheds some light on the complex behavior of open-ended multi-agent interaction, which we hope is helpful in designing, predicting, and monitoring autonomous agentic systems in the real world.
