Reasoning-enhanced Query Understanding through Decomposition and Interpretation
Yunfei Zhong, Jun Yang, Yixing Fan, Jiafeng Guo, Lixin Su, Maarten de Rijke, Ruqing Zhang, Dawei Yin, Xueqi Cheng · 2025
Read Reasoning-enhanced Query Understanding through Decomposition and Interpretation by Yunfei Zhong, Jun Yang, Yixing Fan, Jiafeng Guo, Lixin Su, Maarten de Rijke, Ruqing Zhang, Dawei Yin, Xueqi Cheng (2025) 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-2509-06544.txt.
Description
Long, tangled search queries are broken apart and explained in a 2025 information retrieval paper led by Yunfei Zhong, with six co-authors. Search engines must guess user intent, and language models help, but most studies use short keyword queries. AI-driven search now brings complex, multi-part questions that are far less studied.
The method, ReDI, works in three stages. A complex query is split into focused sub-queries, each sub-query is enriched with a written interpretation of its meaning, and documents found for each part are merged into one ranking. The team also built a large collection of real, tangled questions drawn from a commercial search service, and distilled a big reasoning model's skill into smaller ones. On public benchmarks, ReDI beat strong rivals in both keyword-based and meaning-based retrieval.
Catalog record
| CREATOR | Yunfei Zhong, Jun Yang, Yixing Fan, Jiafeng Guo, Lixin Su, Maarten de Rijke, Ruqing Zhang, Dawei Yin, Xueqi Cheng |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 48 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-06544