{"ID":2841699,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2511.11255","arxiv_id":"2511.11255","title":"Align$^3$GR: Unified Multi-Level Alignment for LLM-based Generative Recommendation","abstract":"Large Language Models (LLMs) demonstrate significant advantages in leveraging structured world knowledge and multi-step reasoning capabilities. However, fundamental challenges arise when transforming LLMs into real-world recommender systems due to semantic and behavioral misalignment. To bridge this gap, we propose Align$^3$GR, a novel framework that unifies token-level, behavior modeling-level, and preference-level alignment. Our approach introduces: Dual tokenization fusing user-item semantic and collaborative signals. Enhanced behavior modeling with bidirectional semantic alignment. Progressive DPO strategy combining self-play (SP-DPO) and real-world feedback (RF-DPO) for dynamic preference adaptation. Experiments show Align$^3$GR outperforms the SOTA baseline by +17.8% in Recall@10 and +20.2% in NDCG@10 on the public dataset, with significant gains in online A/B tests and full-scale deployment on an industrial large-scale recommendation platform.","short_abstract":"Large Language Models (LLMs) demonstrate significant advantages in leveraging structured world knowledge and multi-step reasoning capabilities. However, fundamental challenges arise when transforming LLMs into real-world recommender systems due to semantic and behavioral misalignment. To bridge this gap, we propose Ali...","url_abs":"https://arxiv.org/abs/2511.11255","url_pdf":"https://arxiv.org/pdf/2511.11255v2","authors":"[\"Wencai Ye\",\"Mingjie Sun\",\"Shuhang Chen\",\"Wenjin Wu\",\"Peng Jiang\"]","published":"2025-11-14T12:52:43Z","proceeding":"cs.IR","tasks":"[\"cs.IR\"]","methods":"[\"Large Language Model\",\"Language Model\"]","has_code":false}