GenRec: A Preference-Oriented Generative Framework for Large-Scale Recommendation Conference: Proceedings of the 49th International ACM SIGIR Conference on Research and Development in Information Retrieval; July 20–24, 2026; Melbourne, VIC, Australia Proceedings of the 49th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR ’26), July 20–24, 2026, Melbourne, VIC, Australia DOI: 10.1145/3805712.3808437 ISBN: 979-8-4007-2599-9/2026/07 CCS: Information systems Recommender systems

Yanyan Zou, Junbo Qi, Lunsong Huang, Yu Li, Kewei Xu, Jiahao Gao, Binglei Zhao, Xuanhua Yang, Sulong Xu, Shengjie Li · 2026

Read GenRec: A Preference-Oriented Generative Framework for Large-Scale Recommendation Conference: Proceedings of the 49th International ACM SIGIR Conference on Research and Development in Information Retrieval; July 20–24, 2026; Melbourne, VIC, Australia Proceedings of the 49th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR ’26), July 20–24, 2026, Melbourne, VIC, Australia DOI: 10.1145/3805712.3808437 ISBN: 979-8-4007-2599-9/2026/07 CCS: Information systems Recommender systems by Yanyan Zou, Junbo Qi, Lunsong Huang, Yu Li, Kewei Xu, Jiahao Gao, Binglei Zhao, Xuanhua Yang, Sulong Xu, Shengjie Li (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-14878.txt.

Description

Zou and nine others present GenRec, a preference generative recommendation system on the JD App in one decoder-only design. Page-wise next-token prediction supervises pages over items, resolving pointwise training ambiguity. A linear token merger compresses semantic IDs in prompts while keeping full decoding, halving input length with negligible accuracy loss. GRPO-SR pairs group-relative optimization with NLL regularization, using rewards from a reward model plus a gate against reward hacking. A/B tests showed 9.5 percent more clicks and 8.7 percent more transactions than the old pipeline.

Catalog record

CREATORYanyan Zou, Junbo Qi, Lunsong Huang, Yu Li, Kewei Xu, Jiahao Gao, Binglei Zhao, Xuanhua Yang, Sulong Xu, Shengjie Li
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE22912
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.14878

Browse the full library · About the Tower of Babel