MARS Technical Report
Modern industrial recommender systems use multi-stage cascades--retrieval, pre-ranking, ranking, and re-ranking--to narrow millions of candidates to a personalized list. While each stage deploys sophisticated models, pipeline-level decisions on channel activation, quotas, model selection, and score fusion remain engineer-tuned and user-agnostic within each scenario. This report presents MARS (Multi-stage Agentic Recommendation System), extending personalization from individual-stage predictions to full pipeline orchestration. A near-line agent built on fine-tuned Qwen3-0.6B generates per-user configurations--a seven-dimensional action covering pipeline mode, retrieval quotas, model selection, fusion weights, and reranker settings--from a constrained action space. Configurations are validated, cached, and consumed asynchronously; the cascade never waits for LLM inference. Schema checks, fallbacks, and a kill switch ensure serving safety. The policy is trained in three stages: Confidence-aware SFT distills schema-valid generation from a teacher model; Format-Anchored Preference Optimization (FAPO) learns action preferences via ActionMask with posterior weighting; Group Relative Policy Optimization (GRPO) refines the policy through Sandbox replay with a frozen reward model, without exposing exploratory actions to live traffic. Two-week A/B tests across five Taobao scenarios with over ten million daily users show positive gains on all key metrics. The largest reach +5.22% V2P and +7.34% GMV, with positive CTR lifts on both click surfaces. About 20% of requests use a lightweight Minimal mode, and deployment shows 12 ms lower mean serving latency versus production.
Authors
- Yong Duan (ORCID: https://orcid.org/0000-0003-3793-5099)
- Chaoqun Hou (ORCID: https://orcid.org/0009-0001-5010-0021)
- Fengze Jiang
- Xiannian Hu (ORCID: https://orcid.org/0009-0001-3296-2867)
- Junhong Lou (ORCID: https://orcid.org/0009-0009-0001-2494)
- Zi Zhuang
- Lv Shao (ORCID: https://orcid.org/0009-0008-5389-6476)
- Longxiang Xiong (ORCID: https://orcid.org/0009-0008-5377-6896)
- Xinyuan Qi
- Jiaqi Zheng
- Tong Liu
- Yu Zhu
- Qixin Li
- Cheng Guo
- Zongwei Wang
- Zixin Zhao
- Ping Liu
Publication Details
- Journal
- arXiv (Cornell University)
- Published
- 2026-11-18
- DOI
- https://doi.org/10.48550/arxiv.2611.24006
- Primary Topic
- Recommender Systems and Techniques
- Type
- preprint