paper
Attractor States Emerge in Multi-Turn LLM Conversations
paperactiveprovisional
attractor-states-emerge-in-multi-turn-llm-conversations-9019bd9f·1 events·first seen 18h agoAliases: Attractor States Emerge in Multi-Turn LLM Conversations
Co-occurring entities
More like this (12)
From Tokens to States: LLMs as a Special Case of World Models and the Continuous Path BeyondAccuracy and Satisfaction in Multi-Turn LLM Dialogues for NFR AssessmentAgentic Chain-of-Thought Steering for Efficient and Controllable LLM ReasoningA sleep-like consolidation mechanism for LLMsFollow the Latent Roadmap: Navigating Revocable Decoding for Diffusion LLMs with Anchor TokensRevising Context, Shifting Simulated Stance: Auditing LLM-Based Stance Simulation in Online Discussionslong-context LLMsMulti-Component LLM AgentAttention Amnesia in Hybrid LLMs: When CoT Fine-Tuning Breaks Long-Range Recall, and How to Fix ItExpRL: Exploratory RL for LLM Mid-TrainingDo LLM Embedding Spaces Recover Expert Structure?multi-turn language models
Recent events (1)
Attractor states emerge in multi-turn LLM conversations, with asymmetric model influence
A new arXiv preprint studies long-run dynamics in multi-agent LLM conversations across 7 models and 20 controversial topics, finding that self-play trajectories form model-specific attractor states that asymmetrically influence conversation partners in mixed-play debates. Claude Haiku is identified as a strong attractor that pulls other models toward its stylistic traits (e.g., metacommentary), while GPT-4.1 nano is found to be especially malleable. The results suggest open-ended LLM interactions are partially predictable from model-specific attractors, with implications for designing and monitoring autonomous agentic systems.