Internalizing Multi-Agent Reasoning for Accurate and Efficient LLM-based Recommendation

TL;DR

This paper introduces STAR, a single-agent recommender system that internalizes multi-agent reasoning through a trajectory-driven distillation process, significantly improving accuracy and efficiency in real-time recommendations.

Contribution

The paper presents a novel framework that internalizes multi-agent reasoning into a single model, reducing latency while maintaining high reasoning accuracy in recommender systems.

Findings

STAR outperforms its teacher by 8.7% to 39.5% in accuracy.

The framework eliminates iterative latency, enabling real-time recommendations.

Extensive experiments validate the effectiveness of the proposed approach.

Abstract

Large Language Models (LLMs) are reshaping recommender systems by leveraging extensive world knowledge and semantic reasoning to interpret user intent. However, effectively integrating these capabilities with collaborative signals while avoiding prohibitive inference latency remains a critical bottleneck. To address this, we propose a trajectory-driven internalization framework to develop a Single-agent Trajectory-Aligned Recommender (STAR). Specifically, to internalize complex reasoning capabilities into a single efficient model, we first design a multi-agent teacher system capable of multi-turn tool usage and reflection. This teacher utilizes a Collaborative Signal Translation mechanism to explicitly convert latent behavioral patterns into descriptive natural language evidence to enhance reasoning accuracy. Subsequently, a trajectory-driven distillation pipeline transfers this agentic logic, including planning, tool usage, and self-reflection, into the compact STAR model. Extensive experiments demonstrate that STAR surpasses its teacher by 8.7% to 39.5% while eliminating iterative latency, paving the way for real-time, reasoning-enhanced recommendation.