Financial Transaction Retrieval and Contextual Evidence for Knowledge-Grounded Reasoning CCS: Information systems Information retrieval CCS: Information systems Data mining
Artem Sakhno, Daniil Tomilov, Yuliana Shakhvalieva, Inessa Fedorova, Daria Ruzanova, Omar Zoloev, Andrey Savchenko, Maksim Makarenko · 2026
Read Financial Transaction Retrieval and Contextual Evidence for Knowledge-Grounded Reasoning CCS: Information systems Information retrieval CCS: Information systems Data mining by Artem Sakhno, Daniil Tomilov, Yuliana Shakhvalieva, Inessa Fedorova, Daria Ruzanova, Omar Zoloev, Andrey Savchenko, Maksim Makarenko (2026) 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-2603-15459.txt.
Description
Sakhno and co-authors present FinTRACE, a retrieval-first system that turns raw bank transactions into evidence for downstream reasoning. General language models cope poorly with time-spread tabular records. FinTRACE converts transactions into reusable features, runs rule-based detectors, and stores the signals in a behavioral knowledge base. On public and industry tests, zero-shot churn scores doubled, and few-shot scores rose as well. Instruction tuning on the retrieved patterns then gave language models state-of-the-art results on transaction analytics tasks.
Catalog record
| CREATOR | Artem Sakhno, Daniil Tomilov, Yuliana Shakhvalieva, Inessa Fedorova, Daria Ruzanova, Omar Zoloev, Andrey Savchenko, Maksim Makarenko |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 20846 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.15459