Harmonizing Semantic and Collaborative in LLMs: Reasoning-based Embedding Generator for Sequential Recommendation

2026-06-15

AI Summary

This work addresses the long-tail problem in sequential recommendation and the limitations of existing large language model (LLM)-based embedding approaches, which often neglect LLMs’ intrinsic reasoning capabilities and lack explicit collaborative signal guidance. To this end, the authors propose ReaEmb, a novel framework that uniquely integrates LLM internal reasoning with explicit collaborative signals through a two-stage training paradigm: first, latent reasoning–enhanced contrastive learning refines semantic representations; second, collaborative reward–driven reinforcement learning strengthens user–item interaction modeling. ReaEmb incorporates a dual-pass forward reasoning mechanism and an auxiliary attention module, significantly boosting the performance of diverse sequential recommendation models across three real-world datasets, thereby demonstrating its effectiveness and generalizability.

Abstract

Sequential Recommender Systems (SRS) predict the next item of interest based on users' interaction histories and have been widely deployed, but hindered by long-tail problem. Large Language Models (LLMs), with strong semantic understanding and reasoning capabilities, offer a promising way to enrich item semantics and have recently been used as embedding generators. However, two fundamental gaps remain. First, current LLM-based embedding methods fail to exploit the model's inner reasoning capacity. Second, existing methods often inject collaborative signals implicitly via supervised fine-tuning, lacking explicit guidance for collaborative embedding alignment. In this paper, we introduce ReaEmb, a novel framework that resolves both issues via a Latent Reasoning-enhanced Contrastive Learning (LRCL) stage and a Collaborative Reward Reinforcement Learning (CRRL) stage. LRCL exploits the LLMs' inner reasoning capacity through a two-pass forward process with an additional attention module. CRRL subsequently explicitly injects collaborative signals into the LLM via a tailored reinforcement learning. Extensive experiments on three real-world datasets demonstrate superior effectiveness of ReaEmb across multiple SRS models. To ease reproducibility, we release the code online.

Problem

• Sequential Recommendation
• Large Language Models
• Embedding Generation
• Collaborative Signals
• Reasoning Capacity

Innovation

• Reasoning-based Embedding
• Latent Reasoning-enhanced Contrastive Learning
• Collaborative Reward Reinforcement Learning
• Sequential Recommendation
• Large Language Models