Learning Dynamic User Personas from Implicit Interaction Streams via Iterative Refinement

Haifeng Wu

Personalizing large language models (LLMs) to individual users is essential for improving user experience, yet existing approaches typically rely on explicit preference supervision…