titleDREAM Technical Report
authorsBin Zhang, Bowen Zheng, Chao Yi, Chengyu Lai, Dian Chen, Dimin Wang, Gaoyang Guo, Jialin Zhu, Jian Wu, Jing Yu, Jiuning Lin, Lingqing Zhang, Lingyun Zheng, Mao Zhang, Mingming Pan, Ruiquan Lan, Shuai Zhong, Wen Chen, Wendong Zhang, Xiaodong Zhu, Xuan Chen, Xunke Xi, Yifan Lu, Yiheng Wang, Yue Zeng, Yujie Luo, Yuning Jiang, Zhe Hu, Zhibo Xiao, Zihong Huang, Binbin Cao, Bo Zheng, Danning Wang, Dixuan Wang, Ge Fan, Haixia Wu, Han Zhu, Hao Fang, Haoming Chen, Huiping Chu, Jian Wang, Jianjun Wu, Jiawei Wu, Jiaxin Yu, Jingwen Liu, Jinzhe Shan, Kai Meng, Kai Zhang, Keqin Xu, Kewei Zhu, Lang Tian, Leihui Chen, Li Chen, Licheng Xu, Lide Xiao, Ruitong Zhang, Shiyao Peng, Silu Zhou, Tao Wang, Wei Shi, Wenjun Yang, Xiang Chen, Xiang Gao, Xiao Ren, Xu Liu, Xuwen Wang, Yang Li, Yeqiu Yang, Yi Hu, Yichen Yuan, Yinnan Song, Yipeng Yu, Yuan Liu, Yunqi Gao, Zhiliang Huang, Zhujin Gao, Zongyuan Wu
date2026-08-13
venuearXiv (preprint, not peer reviewed)
idarXiv:2608.09408
affiliationalibaba
upvotes5 (Hugging Face Daily Papers, 2026-11-01)
linksabs pdf hf
abstractIndustrial recommender systems commonly use cascaded retrieval, ranking, and re-ranking pipelines. Although efficient, these pipelines fragment information and objectives across modules, rely on rigid rules, and have limited awareness of real-time intent, leaving session-level shifts among browsing, comparison, and purchase insufficiently addressed. We present DREAM (Developing Recommender Engine with Agentic Methods), an autonomous optimization control architecture that adds a perception-aware, orchestrable, and auditable policy layer atop existing pipelines without replacing them. DREAM has two core components. First, a three-tier Intent Engine fuses on-device signals into structured L0/L1/L2 intent representations; its edge-cloud trigger chain reduces reporting volume to approximately 8.7%. Second, a Meta Engine uses a MetaModel for layered M1-to-M2-to-M3 reasoning: intent summarization, strategy planning informed by Strategy Memory, and parameter translation. It dispatches the resulting parameters through a unified outlet with safety guardrails. A Reward Dual Loop continuously optimizes both components by combining offline simulation for strategy-space exploration with online feedback for outcome calibration, forming a cycle of generation, execution, evaluation, and experience accumulation. Large-scale A/B tests on Taobao's homepage feed show that re-ranking control alone improves IPV by 2.06%, Core IPV by 2.39%, and GMV by 0.88%. Extending control to fine ranking raises these gains to 2.71%, 3.06%, and 1.31%, respectively, while consistently improving PV by more than 1%. These gains require neither replacement of pipeline models nor compromise of serving stability, supporting agentic meta-control as a viable paradigm for industrial recommendation.

cite

@misc{zhang2026dream,
  title={DREAM Technical Report},
  author={Bin Zhang and Bowen Zheng and Chao Yi and Chengyu Lai and Dian Chen and Dimin Wang and Gaoyang Guo and Jialin Zhu and Jian Wu and Jing Yu and Jiuning Lin and Lingqing Zhang and Lingyun Zheng and Mao Zhang and Mingming Pan and Ruiquan Lan and Shuai Zhong and Wen Chen and Wendong Zhang and Xiaodong Zhu and Xuan Chen and Xunke Xi and Yifan Lu and Yiheng Wang and Yue Zeng and Yujie Luo and Yuning Jiang and Zhe Hu and Zhibo Xiao and Zihong Huang and Binbin Cao and Bo Zheng and Danning Wang and Dixuan Wang and Ge Fan and Haixia Wu and Han Zhu and Hao Fang and Haoming Chen and Huiping Chu and Jian Wang and Jianjun Wu and Jiawei Wu and Jiaxin Yu and Jingwen Liu and Jinzhe Shan and Kai Meng and Kai Zhang and Keqin Xu and Kewei Zhu and Lang Tian and Leihui Chen and Li Chen and Licheng Xu and Lide Xiao and Ruitong Zhang and Shiyao Peng and Silu Zhou and Tao Wang and Wei Shi and Wenjun Yang and Xiang Chen and Xiang Gao and Xiao Ren and Xu Liu and Xuwen Wang and Yang Li and Yeqiu Yang and Yi Hu and Yichen Yuan and Yinnan Song and Yipeng Yu and Yuan Liu and Yunqi Gao and Zhiliang Huang and Zhujin Gao and Zongyuan Wu},
  year={2026},
  eprint={2608.09408},
  archivePrefix={arXiv},
  url={https://arxiv.org/abs/2608.09408}
}