Align³GR: Unified Multi-Level Alignment for LLM-based Generative Recommendation
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (19) · 16154-16162
DOI: 10.1609/aaai.v40i19.38651 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Wencai Ye und 4 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref).
Abstract (Originalsprache)
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 recommendation systems due to semantic and behavioral misalignment. To bridge this gap, we propose Align³GR, 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 Align³GR 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.
Zitieren
APA
Ye, W., Sun, M., Chen, S., Wu, W., & Jiang, P. (2026). Align³GR: Unified Multi-Level Alignment for LLM-based Generative Recommendation. Proceedings of the AAAI Conference on Artificial Intelligence, 40(19), 16154-16162. https://doi.org/10.1609/aaai.v40i19.38651
Vancouver
Ye W, Sun M, Chen S, Wu W, Jiang P. Align³GR: Unified Multi-Level Alignment for LLM-based Generative Recommendation. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(19):16154-16162. doi:10.1609/aaai.v40i19.38651
BibTeX
@article{Ye2026,
author = {Ye, Wencai and Sun, Mingjie and Chen, Shuhang and Wu, Wenjin and Jiang, Peng},
title = {Align³GR: Unified Multi-Level Alignment for LLM-based Generative Recommendation},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {19},
pages = {16154--16162},
doi = {10.1609/aaai.v40i19.38651}
}
EUVIMED