PersonaMem v2: Towards Personalized Intelligence via Learning Implicit User Personas and Agentic Memory 📅 PersonaMem v3 is comming soon! 🚨 The paper is now released. View the full paper here and codebase here. Personalization is becoming the next milestone of artificial super intelligence. AI cannot always satisfy every user, especially on tasks with subjective goals, but personalization offers a path toward pluralistic alignment . PersonaMem v2 is the new state of the art LLM personalization dataset focusing on implicit personas in LLMs , where user–chatbot conversations implicitly indicate user preferences. For example, a user might unintentionally reveal a seasonal allergy in their email content while only asking a chatbot to refine the wording of that email . Our goal is to mimic realistic user personas and users' long form conversation histories with chatbots, in order to study how well AI systems can memorize and infer these implicit signals and understand the users they are interacting with, and therefore provide personalized responses over time to enhance user experience. For questions, please reach out to Bowen Jiang (Lauren) at bwjiang@seas.upenn.edu, or submit an issue…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy