Latent Inter-User Difference Modeling for LLM Personalization

2025-07-28

AI Summary

Existing personalized LLM approaches rely on explicit language prompts, limiting their ability to capture deep inter-user differences. This work proposes a latent-space user modeling framework: first, user embeddings are constructed via contrastive learning to explicitly model behavioral relativity among users; second, difference-aware soft prompts are generated, and task-relevant features are compressed and filtered using a sparse autoencoder; finally, a lightweight personalized module is injected into a frozen LLM. Crucially, this is the first method to model user heterogeneity in the latent space—rather than at the prompt level—thereby avoiding prompt-engineering biases and enhancing representation generalizability. Evaluated on personalized review generation, the method achieves state-of-the-art performance across BLEU, ROUGE, and human evaluation metrics, demonstrating both effectiveness and robustness.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Prompt Engineering / Prompting
• Humans and AI: Intelligent User Interfaces

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

Large language models (LLMs) are increasingly integrated into users' daily lives, leading to a growing demand for personalized outputs. Previous work focuses on leveraging a user's own history, overlooking inter-user differences that are crucial for effective personalization. While recent work has attempted to model such differences, the reliance on language-based prompts often hampers the effective extraction of meaningful distinctions. To address these issues, we propose Difference-aware Embedding-based Personalization (DEP), a framework that models inter-user differences in the latent space instead of relying on language prompts. DEP constructs soft prompts by contrasting a user's embedding with those of peers who engaged with similar content, highlighting relative behavioral signals. A sparse autoencoder then filters and compresses both user-specific and difference-aware embeddings, preserving only task-relevant features before injecting them into a frozen LLM. Experiments on personalized review generation show that DEP consistently outperforms baseline methods across multiple metrics. Our code is available at https://github.com/SnowCharmQ/DEP.

Problem

• Modeling inter-user differences for LLM personalization
• Overcoming reliance on language prompts for user distinction
• Enhancing personalized outputs via latent space embeddings

Innovation

• Latent space modeling for user differences
• Contrastive embeddings highlight behavioral signals
• Sparse autoencoder filters task-relevant features