Enhancing Retrieval Systems with Inference-Time Logical Reasoning
Felix Faltings, Wei Wei, Yujia Bao · 2025
Read Enhancing Retrieval Systems with Inference-Time Logical Reasoning by Felix Faltings, Wei Wei, Yujia Bao (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-2503-17860.txt.
Description
Felix Faltings, Wei Wei, and Yujia Bao teach search engines to reason with logic at query time, in a 2025 paper. Classic retrieval turns a query into a vector and ranks by closeness, which mangles requests with nots, ands, and ors. Their framework reads the query's logical skeleton first. Scores for the simple parts are then combined by the rules of logic into the final ranking.
No retraining of the embedding model is needed, and the speed stays practical. On synthetic and real benchmarks, across several models, the method beat plain vector search consistently. The gains concentrated exactly where logic bites: complex, compound questions. The authors frame query-time reasoning as a cheap upgrade for retrieval systems.
Catalog record
| CREATOR | Felix Faltings, Wei Wei, Yujia Bao |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-17860