Reasoning distillation improves recommendations by using multiple teacher models
Multi-Teacher Reasoning Distillation for LLM Recommendation with Self-Exploration
Information RetrievalArtificial Intelligence
Summary
Recommender systems suggest things you might like based on your past choices, but sometimes they struggle to understand why you like them. This paper shows how to teach smaller recommendation systems to think like large language models by looking at multiple example explanations from different teachers. The system also learns to tell which explanations actually help improve recommendations and focuses on those. This method helps make recommendations smarter and more useful.
What this means in practice
- •For recommendation system engineers: Build recommendation systems that generate better suggestions by learning from multiple AI teachers and focusing on reasoning paths that truly help improve recommendations.
- •For ecommerce platform developers: Enhance product recommendations by integrating lightweight models distilled from large language models that adaptively select useful reasoning to personalize suggestions.
Authors
Liangbo Ning, Siyuan Liang, Yi Zhou, Yifeng Zhang, Qing Li, Wenqi Fan, Dacheng Tao
Abstract
Reasoning-augmented recommender systems (RecSys) enhance user preference modeling by generating reasoning trajectories over historical interactions before making recommendations. Knowledge distillation provides a practical paradigm for transferring such reasoning capabilities from powerful large language models to lightweight RecSys. However, existing methods typically rely on a single fixed teacher, remain constrained by the teacher knowledge ceiling, and indiscriminately distill teacher-generated trajectories without assessing their actual utility for downstream recommendation. To address these limitations, we propose MUSE (Multi-Teacher Utility-guided Selection with Self-Exploration), an adaptive reasoning distillation framework that jointly improves reasoning generation and reasoning utilization. MUSE first constructs a diverse reasoning pool using multiple teacher LLMs and progressively augments it via student self-exploration, enabling the student to leverage the strengths of different teachers and gradually explore reasoning strategies beyond teacher imitation. We then develop Rank-Gain Utility Assessment (RGUA), which evaluates each candidate according to the normalized full-catalog rank improvement of the ground-truth item and identifies the most and least recommendation-beneficial trajectories. Based on these utility-aware trajectory pairs, we introduce dual-level contrastive learning objectives that guide the student generator to prefer beneficial reasoning while training the recommendation encoder to more effectively exploit recommendation-relevant preference evidence. Extensive experiments on three real-world datasets demonstrate the effectiveness and robustness of the proposed framework.