R&D / Banking AI Evaluation Library / Datasets and sources
TAT-QA
Questions over tables and text from real financial reports, with answers and the numeric reasoning needed to reach them.
Open the publisher page: TAT-QA repository (opens in a new tab)
Dodone links to the publisher and does not host or copy this data. Read the licence on the publisher page before you use it.
About this source
- Publisher
- NExT++ (National University of Singapore)
- Context
- Financial reports with tables and text, English
- Version
- Not available (the test set answers were added in January 2024)
- Last checked
- 2026-11-03
- Labels or answers
- An answer for each question, with its scale and derivation where one is needed.
Facts stated by the publisher
- 16,552 questions over 2,757 contexts that mix a table and text.
Licence, access and privacy
- Licence
- Creative Commons Attribution 4.0 International (data); MIT License (code) (opens in a new tab) (Licence verified)
- Scope
- The README states CC BY 4.0 for the data set. The LICENSE file states MIT for the code. Use needs the citation the README gives.
- Access
- Open. Public GitHub repository.
- Privacy
- Privacy reviewed: low risk. Company financial reports. No customer records.
Limitations
- It tests general financial report reasoning, not banking operations or compliance decisions.
- Public reports may be in the training data of current models.
Suggested for these evaluations
- Financial-report numerical reasoning Direct fit
Why it helps: It has questions over financial report tables and text, with answers, scales and derivations.
What it does not validate: Banking operations, bank-internal reports and regulatory returns.