Know You Before You Speak: User-State Modeling for LLM Personalization in Multi-Turn Conversation
arXiv:2605.24647v1 Announce Type: new Abstract: Personalized dialogue requires more than recalling explicit user histories: systems also need to infer hidden user states that evolve through interaction and shape appropriate response strategies. Existing memory- and profile-based methods primarily reuse observable user information, offering limited support for modeling user-state dynamics or selecting actions based on how they shape future user states. We propose PUMA (Prospective User-state Modeling for Action selection), a framework grounded in the Free Energy Principle (FEP) that formulates
The proliferation of large language models (LLMs) and their integration into user-facing applications highlights the critical need for sophisticated personalization to improve user experience and efficacy.
Advanced user-state modeling is crucial for the seamless adoption and utility of AI systems, moving beyond simple recall to infer dynamic user needs and influence future interactions meaningfully.
This research introduces a framework that allows LLMs to proactively anticipate user states and select actions accordingly, representing a significant step towards truly intelligent and adaptive AI agents.
- · AI developers
- · Customer service platforms
- · Personalized learning systems
- · UX designers
- · Generic chatbot providers
- · Rule-based AI systems
LLMs will become significantly more capable of maintaining coherent and personalized multi-turn conversations.
Improved personalization will accelerate the adoption of AI agents across various sectors, collapsing certain white-collar workflows.
The ability to predict and influence user states could lead to more sophisticated forms of influence and potentially ethical concerns regarding manipulation.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL