Agentic retrieval method improves search with selective feature use
AGR: Agentic Generative Retrieval with Tool-Augmented Search
Information Retrieval
Summary
Search systems use lots of different clues or features to find good results, but current methods either use all features all the time or stick to fixed rules about how to use them. This can slow things down and add confusing information. The authors created a new approach called Agentic Generative Retrieval (AGR) that decides when and which features to use during the search, making the process smarter and more focused. Their method tests well on multiple benchmarks and even in real-world online experiments.
What this means in practice
- •For search engine developers: Build search systems that dynamically choose which features to use for each query to improve result quality and efficiency.
- •For recommendation system engineers: Incorporate on-demand feature acquisition into recommendation workflows to better tailor suggestions based on relevant data.
Authors
Ming Yang, Yaqi Yin, Zhibo Wang, Xiaoze Jiang, Jingwei Zhuo, Chao Zhang, Quanzhi Zhu, Weiguo Zheng, Liang Zeng, Yunfeng Zhao, Jiaqiang Liu, Kefeng Wang, Yupeng Huang, Kaiqiao Zhan, Xuanping Li
Abstract
Large-scale search and recommendation systems derive much of their retrieval quality from rich, heterogeneous features, yet existing generative retrieval (GR) methods struggle to exploit such feature ecosystems efficiently. Static feature injection either encodes all available signals or places them according to predetermined schemes, increasing serving cost and introducing irrelevant or conflicting evidence. More fundamentally, existing methods cannot adaptively determine when a feature is needed, which feature to select, or where it should influence hierarchical Semantic ID (SID) generation. We propose Agentic Generative Retrieval (AGR), a tool-augmented paradigm that interleaves SID generation with on-demand feature acquisition. Conditioned on the request and decoded SID prefix, AGR decides whether to invoke a tool, which feature to acquire, and where its observation should influence subsequent decoding, thereby steering generation toward feature-consistent SID branches. To supervise this behavior, we construct tool-augmented trajectories based on feature applicability and information gain along the SID hierarchy, and train the model through SID pretraining followed by tool-augmented fine-tuning. Experiments on an industrial dataset and three public benchmarks, together with an online A/B test, consistently demonstrate the effectiveness of AGR.