PRIME: Large Language Model Personalization with Cognitive Dual-Memory and Personalized Thought Process
This work addresses the problem of aligning LLM outputs with individual preferences for users and developers, offering a systematic approach that is incremental by building on existing cognitive models.
The paper tackles the lack of a unified theoretical framework for large language model personalization by integrating a cognitive dual-memory model, resulting in the PRIME framework that effectively captures dynamic personalization beyond popularity biases, as validated through extensive experiments.
Large language model (LLM) personalization aims to align model outputs with individuals' unique preferences and opinions. While recent efforts have implemented various personalization methods, a unified theoretical framework that can systematically understand the drivers of effective personalization is still lacking. In this work, we integrate the well-established cognitive dual-memory model into LLM personalization, by mirroring episodic memory to historical user engagements and semantic memory to long-term, evolving user beliefs. Specifically, we systematically investigate memory instantiations and introduce a unified framework, PRIME, using episodic and semantic memory mechanisms. We further augment PRIME with a novel personalized thinking capability inspired by the slow thinking strategy. Moreover, recognizing the absence of suitable benchmarks, we introduce a dataset using Change My View (CMV) from Reddit, specifically designed to evaluate long-context personalization. Extensive experiments validate PRIME's effectiveness across both long- and short-context scenarios. Further analysis confirms that PRIME effectively captures dynamic personalization beyond mere popularity biases.