SPRInG: Continual LLM Personalization via Selective Parametric Adaptation and Retrieval-Interpolated Generation

TL;DR

SPRInG is a semi-parametric framework for continual personalization of large language models, effectively adapting to evolving user preferences while avoiding catastrophic forgetting through selective updates and retrieval-augmented generation.

Contribution

It introduces a novel drift-driven selective adaptation method and a retrieval-interpolated generation approach for continual LLM personalization.

Findings

Outperforms existing baselines on personalized generation benchmarks.

Effectively adapts to user preference drift without catastrophic forgetting.

Demonstrates robustness in real-world continual personalization scenarios.

Abstract

Personalizing Large Language Models typically relies on static retrieval or one-time adaptation, assuming user preferences remain invariant over time. However, real-world interactions are dynamic, where user interests continuously evolve, posing a challenge for models to adapt to preference drift without catastrophic forgetting. Standard continual learning approaches often struggle in this context, as they indiscriminately update on noisy interaction streams, failing to distinguish genuine preference shifts from transient contexts. To address this, we introduce SPRInG, a novel semi-parametric framework designed for effective continual personalization. During training, SPRInG employs drift-driven selective adaptation, which utilizes a likelihood-based scoring function to identify high-novelty interactions. This allows the model to selectively update the user-specific adapter on drift signals while preserving hard-to-learn residuals in a replay buffer. During inference, we apply strict relevance gating and fuse parametric knowledge with retrieved history via logit interpolation. Experiments on the long-form personalized generation benchmark demonstrate that SPRInG outperforms existing baselines, validating its robustness for real-world continual personalization.