RusFinQABenchmark — Evaluation Results

This dataset contains evaluation results for 8 open-weight large language models on the RuFinQA benchmark.

📊 Overview

• Total evaluated records: 8,100
• Models: 8
• Domains: 17
• Topics: 172
• Levels: 3

🤖 Models Evaluated

Model	Records
llama3.2:3b	1,100
phi4-mini:3.8b	1,000
qwen2.5:7b-instruct	1,000
mistral:7b-instruct	1,000
deepseek-r1:7b	1,000
gemma3:4b	1,000
llama3.1:8b	1,000
aya-expanse:8b	1,000

---

📖 Data Sources, Licensing & Legal Notice

Data Origin

This dataset contains model-generated outputs and evaluation metrics produced by running open-weight large language models on the RuFinQA benchmark. The underlying questions and gold solutions are derived from the RuFinQA dataset.

Ownership & Rights

• The evaluation results, metrics, and model generations are released under the MIT License.
• The underlying benchmark questions and gold solutions are subject to the original licensing terms of RuFinQA.
• We do not claim ownership of the model outputs or the original financial texts used in the benchmark.

Notice‑and‑Takedown Policy

We respect intellectual property rights. If you are a copyright owner and believe that your content appears in this dataset without proper authorization, please contact us. We will promptly remove the disputed entries upon verification.

📧 Contact for takedown requests: marabov@kpfu.ru
⏱️ Response time: Within 14 business days.

---

📈 Key Performance Metrics (aggregated)

Metric	Mean	Std
final_answer_match	0.62	0.37
recall	0.71	0.29
precision	0.68	0.31
bertscore	0.83	0.11
rouge1	0.58	0.22
rougeL	0.54	0.23

---

📚 Domain Distribution

Domain (RU)	Domain (EN)	Records
Ценные бумаги	Securities	835
Финансовое регулирование	Financial Regulation	659
Налоги	Taxation / Taxes	555
Аннуитеты и вклады	Annuities and Deposits	508
Финансовые рынки	Financial Markets	507
Личные финансы	Personal Finance	504
Процентные ставки	Interest Rates	475
Кредиты и займы	Loans and Borrowing	475
ESG и устойчивое финансирование	ESG and Sustainable Finance	459
Крипто-финансы	Crypto Finance	459
Слияния и поглощения (M&A)	Mergers and Acquisitions (M&A)	456
Финансовые коэффициенты	Financial Ratios	456
Управление рисками	Risk Management	448
Амортизация	Depreciation / Amortization	400
Инвестиционные проекты	Investment Projects	360
Страхование и актуарные расчёты	Insurance and Actuarial Calculations	272
Корпоративные финансы	Corporate Finance	272

---

📊 Level Distribution

Level	Records
Intermediate	3,582
Basic	2,610
Advanced	1,908

---

📝 Data Structure

Each record contains:

Field	Type	Description
id	string	Task identifier
level	string	Basic / Intermediate / Advanced
domain	string	Financial domain
topic	string	Specific topic
model	string	Model name
question	string	Question (Russian)
solution	string	Gold solution
steps	list	Gold reasoning steps
final_answer	float	Correct answer
model_generation	string	Raw model output
recall	float	Hard recall
precision	float	Hard precision
final_answer_match	int	Correct final answer (0/1)
fuzzy_*	float	Fuzzy metrics
soft_*	float	Soft metrics
dtw_*	float	DTW metrics
bertscore	float	BERTScore
rouge*	float	ROUGE scores

---

🚀 Usage

from datasets import load_dataset

dataset = load_dataset("arabovs-ai-lab/RusFinQABenchmark", split="train")
print(dataset[0])

Example Record

{
  "id": "arith_COMP_0001_2025_roa",
  "level": "Intermediate",
  "domain": "Финансовые коэффициенты",
  "topic": "Рентабельность активов (ROA)",
  "model": "llama3.2:3b",
  "question": "Рассчитай рентабельность активов (ROA) для компании...",
  "solution": "ROA = 44.691 / 633.696 = 0.0705 (7.05%)",
  "steps": [...],
  "final_answer": 0.0705,
  "model_generation": "ROA = 44.69 / 633.70 = 0.0705",
  "recall": 0.92,
  "precision": 0.88,
  "final_answer_match": 1,
  "bertscore": 0.91,
  "rouge1": 0.84
}

Analyzing Results

import pandas as pd
from datasets import load_dataset

dataset = load_dataset("arabovs-ai-lab/RusFinQABenchmark", split="train")

# Convert to DataFrame
df = pd.DataFrame(dataset)

# Calculate accuracy per model
model_acc = df.groupby('model')['final_answer_match'].mean().sort_values(ascending=False)
print(model_acc)

# Filter by domain
df_esg = df[df['domain'] == 'ESG и устойчивое финансирование']
print(f"ESG domain accuracy: {df_esg['final_answer_match'].mean():.3f}")

---

📄 License

MIT License — applies to evaluation results, metrics, and metadata in this dataset. The underlying benchmark content is subject to the original RuFinQA licensing terms.

---

📚 Citation

If you use this evaluation dataset, please cite the original RuFinQA paper:

@misc{rufinqa2025,
  author = {Arabov, Mullosharaf K.},
  title = {RuFinQA: A Massive Multi-Task Reasoning Benchmark for Russian Financial Report Understanding},
  year = {2025},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/arabovs-ai-lab/RuFinQA}
}

---

👤 Author

Mullosharaf K. Arabov
ORCID: 0000-0003-2525-1183
PhD in Physics and Mathematics, Associate Professor
Department of Data Analysis and Programming Technologies
Kazan (Volga Region) Federal University
📧 marabov@kpfu.ru

---

🔗 Links

• 📊 Dataset: https://huggingface.co/datasets/arabovs-ai-lab/RusFinQABenchmark
• 📊 Main RuFinQA Dataset: https://huggingface.co/datasets/arabovs-ai-lab/RuFinQA
• 💻 Generator code: [GitHub]
• 📄 Paper: https://arxiv.org/abs/2607.01388

---

Generated automatically from RuFinQA evaluation pipeline.