Enhancing Sequential Recommendations through Multi-Perspective Reflections and Iteration

2024-09-10

arXiv.org

AI Summary

Existing LLM-based approaches for sequential recommendation suffer from three critical limitations: entanglement of explicit and implicit user features, insufficient exploitation of collaborative signals, and inefficient reflective adaptation to dynamic user preferences. Method: We propose MoRE—a novel multi-perspective reflective generation framework that jointly models explicit (textual) features, implicit (behavioral) patterns, and collaborative (cross-user interaction) signals via an LLM-driven reflection mechanism; integrates a refining-and-iteration self-optimization pipeline; and employs a contextual bandit–guided meta-reflector for dynamic, personalized reflection strategy selection. Contribution/Results: MoRE is the first architecture to unify these three complementary perspectives in reflective sequential modeling. Evaluated on three real-world datasets, it significantly outperforms state-of-the-art methods in recommendation accuracy, while reducing training time and GPU memory consumption. Moreover, MoRE demonstrates superior robustness to preference drift and behavioral dynamics.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Search and Optimization: Learning to Search
• Reasoning under Uncertainty: Sequential Decision Making

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Sequence recommendation (SeqRec) aims to predict the next item a user will interact with by understanding user intentions and leveraging collaborative filtering information. Large language models (LLMs) have shown great promise in recommendation tasks through prompt-based, fixed reflection libraries, and fine-tuning techniques. However, these methods face challenges, including lack of supervision, inability to optimize reflection sources, inflexibility to diverse user needs, and high computational costs. Despite promising results, current studies primarily focus on reflections of users' explicit preferences (e.g., item titles) while neglecting implicit preferences (e.g., brands) and collaborative filtering information. This oversight hinders the capture of preference shifts and dynamic user behaviors. Additionally, existing approaches lack mechanisms for reflection evaluation and iteration, often leading to suboptimal recommendations. To address these issues, we propose the Mixture of REflectors (MoRE) framework, designed to model and learn dynamic user preferences in SeqRec. Specifically, MoRE introduces three reflectors for generating LLM-based reflections on explicit preferences, implicit preferences, and collaborative signals. Each reflector incorporates a self-improving strategy, termed refining-and-iteration, to evaluate and iteratively update reflections. Furthermore, a meta-reflector employs a contextual bandit algorithm to select the most suitable expert and corresponding reflections for each user's recommendation, effectively capturing dynamic preferences. Extensive experiments on three real-world datasets demonstrate that MoRE consistently outperforms state-of-the-art methods, requiring less training time and GPU memory compared to other LLM-based approaches in SeqRec.

Problem

• Decoupling explicit and implicit user features in recommendations
• Enhancing cross-user collaborative filtering signals utilization
• Improving inefficient reflection update strategies dynamically

Innovation

• Decouples explicit and implicit user interaction patterns
• Utilizes cross-user collaborative filtering signals
• Employs self-improving meta-reflector with dynamic selection