CoPersona: Collaborative Persona Graphs for Robust LLM Personalization

2026-07-01

AI Summary

Personalization with large language models is often hindered by sparse user histories and uneven coverage of preferences, leading to inaccurate inference of weakly observed attributes and fragile generalization. To address this, this work proposes a multi-level persona graph framework that decouples user representations into aspect-specific subspaces aligned with preference dimensions. By integrating non-parametric peer retrieval with parametric graph neural network reasoning, the method achieves aspect-level alignment among behaviorally similar users. It explicitly models preference heterogeneity and leverages multi-path graph structures to enhance signal utilization under data sparsity. Experiments demonstrate that the approach significantly outperforms strong baselines across multiple domains and model scales, substantially improving the robustness of personalization in real-world sparse-data scenarios.

Abstract

Real-world LLM personalization is often constrained by sparse and skewed user histories: most users provide only a handful of interactions, while even frequent users' logs capture an incomplete and biased view of their preferences. As a result, weakly observed user attributes are difficult to infer, leading to brittle personalization when test-time requests shift toward under-supported facets. Motivated by this limitation, we present CoPersona, a graph-based collaborative personalization framework that completes sparse user profiles by borrowing signals from behaviorally similar peers. However, directly transferring signals is difficult because uneven facet coverage introduces bias into interaction histories, obscuring user similarity in the unstructured global space. To address this issue, CoPersona decomposes interaction histories into multiple facet-level representations and explicitly models peer-to-peer, facet-level alignment through a multiplex persona graph. To effectively leverage peer information at inference time, we employ a dual-branch architecture that combines non-parametric peer retrieval with parametric graph reasoning. Experiments across multiple domains and model scales demonstrate consistent improvements over strong baselines, validating CoPersona as an effective approach for robust LLM personalization.

Problem

• LLM personalization
• sparse user histories
• biased preferences
• user attribute inference
• robust personalization

Innovation

• collaborative personalization
• persona graph
• facet-level alignment
• sparse user profiles
• graph-based reasoning