Multi-agents based User Values Mining for Recommendation

2025-05-02

AI Summary

Existing recommender systems rely heavily on dynamic user behavior modeling, leading to short-term, unstable recommendations that fail to capture stable long-term preferences. To address this, we propose ZOOM—a novel zero-shot, multi-LLM collaborative framework for value elicitation, the first of its kind to extract deep, stable user values directly from historical interactions without manual annotation. ZOOM introduces an evaluator-supervisor dual-agent mechanism to mitigate hallucination and employs semantic-preserving text summarization to overcome context-length limitations. It establishes an interpretable value extraction pipeline integrating zero-shot reasoning, multi-agent coordination, and advanced prompt engineering. Extensive experiments on two benchmark datasets demonstrate significant improvements in Recall@10 and NDCG@10. Moreover, ZOOM is model-agnostic and consistently enhances recommendation consistency, personalization, and robustness across diverse recommendation backbones, validating the generalizable benefits of value-aware embedding.

Technology Category

• Data Mining & Knowledge Management: Recommender Systems
• Humans and AI: Learning Human Values and Preferences
• Machine Learning: Large Multimodal Models (LMMs)

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
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Recommender systems have rapidly evolved and become integral to many online services. However, existing systems sometimes produce unstable and unsatisfactory recommendations that fail to align with users' fundamental and long-term preferences. This is because they primarily focus on extracting shallow and short-term interests from user behavior data, which is inherently dynamic and challenging to model. Unlike these transient interests, user values are more stable and play a crucial role in shaping user behaviors, such as purchasing items and consuming content. Incorporating user values into recommender systems can help stabilize recommendation performance and ensure results better reflect users' latent preferences. However, acquiring user values is typically difficult and costly. To address this challenge, we leverage the strong language understanding, zero-shot inference, and generalization capabilities of Large Language Models (LLMs) to extract user values from users' historical interactions. Unfortunately, direct extraction using LLMs presents several challenges such as length constraints and hallucination. To overcome these issues, we propose ZOOM, a zero-shot multi-LLM collaborative framework for effective and accurate user value extraction. In ZOOM, we apply text summarization techniques to condense item content while preserving essential meaning. To mitigate hallucinations, ZOOM introduces two specialized agent roles: evaluators and supervisors, to collaboratively generate accurate user values. Extensive experiments on two widely used recommendation datasets with two state-of-the-art recommendation models demonstrate the effectiveness and generalization of our framework in automatic user value mining and recommendation performance improvement.

Problem

• Extract stable user values from dynamic behavior data
• Overcome LLM limitations in direct value extraction
• Improve recommendation alignment with long-term preferences

Innovation

• Leveraging LLMs for user value extraction
• ZOOM framework with multi-agent collaboration
• Text summarization to condense item content