RuFinQA-Source is the sanitized reference dataset used to generate the RuFinQA-CoT benchmark (https://huggingface.co/datasets/RusNLPWorld/RuFinQA-CoT). It contains 4,418 company-year observations from 931 Russian companies registered in the Republic of Tatarstan for the period 2021–2025.

All raw data was extracted from the State Information Resource of Accounting (Financial) Statements (BFO) of the Federal Tax Service of Russia (https://bo.nalog.ru/) — the sole data source used in this project.

The dataset is published for reproducibility: users can re-run the benchmark generator pipeline on this reference to obtain the same benchmark questions.

Curated by: Mullosharaf K. Arabov
Shared by: Mullosharaf K. Arabov
Language(s) (NLP): Russian (ru)
License: CC BY 4.0 (see Legal Notice below)
Dataset size: 4,418 rows × 524 columns, single split (train)

Legal Notice

Copyright and Ownership. The author of this dataset card is Mullosharaf K. Arabov. However, the author is NOT the copyright holder of the underlying financial data. The raw accounting (financial) statements were obtained from a single public source: the State Information Resource of Accounting (Financial) Statements (BFO) of the Federal Tax Service of the Russian Federation (https://bo.nalog.ru/), which is published openly under the legislation of the Russian Federation.

What belongs to the author: the parser, the anonymization pipeline, the metadata, the dataset card.

What does NOT belong to the author: the raw financial figures reported by the companies. These are used here solely for research and benchmarking purposes.

All company identifiers (INN) and names have been anonymized using a deterministic mapping with a fixed random seed (42), the same mapping used in the RuFinQA-CoT benchmark.

If you are a rights holder of any underlying financial statement and believe it should not appear in this dataset, please contact the author at cool.araby@gmail.com for prompt removal.

Dataset Sources

Repository: https://huggingface.co/datasets/RusNLPWorld/RuFinQA-Source
Source data (single source): https://bo.nalog.ru/ — State Information Resource of Accounting (Financial) Statements (BFO), operated by the Federal Tax Service of Russia
Companion benchmark: https://huggingface.co/datasets/RusNLPWorld/RuFinQA-CoT

Uses

Direct Use

Reproducing the RuFinQA-CoT benchmark generation pipeline.
Research on financial statement analysis with real Russian RAS data.
Training and evaluating tabular models on Russian accounting data.
Benchmarking financial ratio computation, extraction, and anomaly detection.

Out-of-Scope Use

Investment or financial advice. This dataset is a research artifact, not a source for real-world investment decisions.
Company profiling or identification. Identifiers are anonymized; re-identification attempts are prohibited.
Extrapolation to the full Russian economy. The dataset covers a single region (Tatarstan, 2021–2025).

Dataset Structure

Data Fields

Row = one company-year observation. Columns fall into four groups.

Identification

Field	Type	Description
`inn`	string	Anonymized 10-digit identifier (synthetic, seed=42)
`report_year`	int	Reporting year (2021–2025)
`source_year`	int	Year from source filename
`company_name`	string	Anonymized company name (synthetic)
`okved`	string	Russian activity code (OKVED2)
`unit_source`	string	Original reporting unit (e.g., `тыс. руб`)
`unit_target`	string	Target unit (`million_rub`)

Financial statements (in million RUB)

Prefix	Statement	Lags
`bal_*`	Balance sheet (RAS line codes)	`_t1`, `_t2` (two prior years)
`inc_*`	Income statement	`_t1`
`cf_*`	Cash flow statement	`_t1`

Derived ratios

Field	Formula
`roa`	Net Profit / Total Assets
`roe`	Net Profit / Shareholders' Equity
`current_ratio`	Current Assets / Current Liabilities
`margin_oper`	Operating Profit / Revenue

Removed fields

For privacy, the following fields were removed before publication:

signer_name, signer_role (personal data)
address, kpp, okpo (identification)
auditor_name, auditor_inn, auditor_ogrn (third-party identifiers)
date_formation, extraction_number, audit_mandatory, approval_mandatory, approval_body (service metadata)

Data Splits

Single split (train) with 4,418 observations. The companion benchmark RuFinQA-CoT uses a filtered subset of 3,504 observations (drop of rows with missing inc_2110 or bal_1600).

Statistics

Distribution by Year

Year	Observations	Share
2021	839	19.0%
2022	884	20.0%
2023	898	20.3%
2024	902	20.4%
2025	895	20.3%
Total	4,418	100.0%

Top-10 OKVED Codes

OKVED	Observations	Share
`41.20`	165	3.7%
`42.11`	99	2.2%
`06.10.1`	96	2.2%
`46.71.2`	83	1.9%
`49.41.1`	75	1.7%
`09.10.9`	74	1.7%
`45.11.2`	73	1.7%
`46.90`	65	1.5%
`45.20.1`	56	1.3%
`43.21`	52	1.2%

Unique OKVED codes: 294.

Key Financial Indicators (million RUB)

Column	Count	Mean	Median	Min	Max
`bal_1600`	4,187	7,099	1,155	0	1,493,118
`bal_1300`	4,176	3,691	242	-44,230	992,131
`bal_1200`	4,027	3,403	864	0	868,128
`bal_1500`	3,973	2,485	549	0	451,922
`inc_2110`	4,177	6,832	2,158	0	1,563,778
`inc_2400`	4,225	591	41	-48,540	251,375

The distributions are heavily right-skewed: median assets ≈ 1.2 billion RUB, mean ≈ 7.1 billion RUB, max ≈ 1.5 trillion RUB.

Derived Ratios

Ratio	Count	Mean	Median	Min	Max
`roa`	4,110	7.3731	0.0459	-66.9296	29917.1642
`roe`	4,105	51.9094	0.2267	-176.7500	200445.0000
`current_ratio`	3,970	2.4962	1.3626	0.0210	326.1300
`margin_oper`	4,015	-0.2136	0.0365	-728.3775	1.0000

Note on extreme outliers: the raw ratios include extreme values (e.g., ROA = 29,917) arising from companies with near-zero denominators (total assets < 0.01 million RUB). The RuFinQA-CoT benchmark applies a data-quality guard (DATA_MIN_ABS = 0.01 million RUB for balance-sheet denominators) to exclude such cases. The reference dataset preserves these rows for full transparency.

Dataset Creation

Source Data

Single source. All raw data was obtained from the State Information Resource of Accounting (Financial) Statements (BFO) of the Federal Tax Service of Russia (https://bo.nalog.ru/).
Download. 4,418 Excel files covering companies registered in the Republic of Tatarstan for the period 2021–2025.
Extraction. A custom parser (BFODataProcessor) extracted numeric values from standardized RAS report line codes, preserving the year-over-year lag structure. All monetary values were converted to millions of Russian rubles.
Anonymization. All company INNs and names were replaced using a deterministic mapping with fixed seed = 42.
PII removal. Signer names, addresses, KPP, OKPO, and auditor identifiers were dropped.
Final size. 4,418 company-year observations covering 931 unique companies.

Who are the source data producers?

The source data producers are the Russian companies themselves, which are legally required to publish their annual accounting statements. The BFO resource aggregates and publishes these as open data. The author of this dataset did not create, modify, or control the source financial figures.

Bias, Risks, and Limitations

Bias

Geographic bias. All companies are registered in the Republic of Tatarstan.
Sector bias. Skewed toward construction (41.20), wholesale trade (46.*), and extractive industries (06.10.1, 09.10.9).
Temporal bias. Coverage limited to 2021–2025.
Size bias. Small and medium enterprises dominate; very large companies are underrepresented.

Risks

Extreme outliers in derived ratios. Raw roa, roe, and current_ratio include values that are mathematically valid but practically meaningless (division by near-zero denominators). Users should apply their own data-quality filters.
Re-identification risk. Although INNs and names are anonymized, the underlying financial figures remain publicly available. Users must not attempt to re-identify companies.
Not investment-grade data. The dataset is a research artifact.

Limitations

524 columns, many of which are 100% missing (e.g., bal_1216, cf_3161). These are preserved for schema completeness but carry no information.
Only RAS standards. No IFRS, US GAAP, or other accounting frameworks.
Single region. Not representative of the full Russian economy.

Recommendations

Apply data-quality filters (|denominator| ≥ 0.01 million RUB) before computing ratios.
Use median, not mean, when reporting central tendencies.
Report the FCS fallback rate (117 rows in RuFinQA-CoT) when using derived indicators.

Citation

@misc{rufinqa_source_2026,
  author       = {Arabov, Mullosharaf K.},
  title        = {RuFinQA-Source: Reference Financial Statements Dataset for the RuFinQA-CoT Benchmark},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/RusNLPWorld/RuFinQA-Source}}
}

Glossary

BFO — Бухгалтерская (финансовая) отчётность: State Information Resource of Accounting (Financial) Statements, operated by the Federal Tax Service of Russia.
RAS / РСБУ — Russian Accounting Standards.
INN — Russian tax identification number.
OKVED — Russian National Classifier of Types of Economic Activity.
ROA / ROE / CR — Return on Assets / Return on Equity / Current Ratio.

Dataset Card Authors

Mullosharaf K. Arabov — Kazan Federal University, Institute of Computational Mathematics and Information Technologies.

Dataset Card Contact

Email: cool.araby@gmail.com Affiliation: Kazan Federal University, Kazan, Russia

Downloads last month: 13

Total file size:

7.06 MB