FinanceMath

## FinanceMath
[**🌐 Homepage**](https://financemath-acl2024.github.io/) | [**🤗 Dataset**](https://huggingface.co/datasets/yale-nlp/FinanceMath) | [**📖 arXiv**](https://arxiv.org/abs/2311.09797) | [**GitHub**](https://github.com/yale-nlp/FinanceMath)

The data and code for the paper [FinanceMath: Knowledge-Intensive Math Reasoning in Finance Domains](https://arxiv.org/abs/2311.09797).
**FinanceMath** is a knowledge-intensive dataset focused on mathematical reasoning within the domain of finance. It requires the model to comprehend specialized financial terminology and to interpret tabular data presented in the questions.

<p align="center">
<img src="overview.png" width="80%">
</p>

## FinanceMath Dataset
All the data examples were divided into two subsets: *validation* and *test*.

- **validation**: 200 examples used for model development, validation, or for those with limited computing resources.
- **test**: 1000 examples for standard evaluation. We will not publicly release the annotated solution and answer for the test set.

You can download this dataset by the following command:

```python
from datasets import load_dataset

dataset = load_dataset("yale-nlp/FinanceMath")

# print the first example on the validation set
print(dataset["validation"][0])

# print the first example on the test set
print(dataset["test"][0])
```

The dataset is provided in json format and contains the following attributes:

```
{
"question_id": [string] The question id,
"question": [string] The question text,
"tables": [list] List of Markdown-format tables associated with the question,
"python_solution": [string] Python-format and executable solution by financial experts. The code is written in a clear and executable format, with well-named variables and a detailed explanation,
"ground_truth": [float] Executed result of `python solution`, rounded to three decimal places,
"topic": [string] The related financial area of the question,
}
```

## Contact
For any issues or questions, kindly email us at: Yilun Zhao ([e-mail]).

## Citation

If you use the **FinanceMath** dataset in your work, please kindly cite the paper:

```
@misc{zhao2024financemath,
title={FinanceMath: Knowledge-Intensive Math Reasoning in Finance Domains},
author={Yilun Zhao and Hongjun Liu and Yitao Long and Rui Zhang and Chen Zhao and Arman Cohan},
year={2024},
eprint={2311.09797},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2311.09797},
}
```

## FinanceMath
[**🌐 项目主页**](https://financemath-acl2024.github.io/) | [**🤗 数据集仓库**](https://huggingface.co/datasets/yale-nlp/FinanceMath) | [**📖 arXiv 论文页**](https://arxiv.org/abs/2311.09797) | [**GitHub 仓库**](https://github.com/yale-nlp/FinanceMath)

本数据集与代码配套于论文《FinanceMath：金融领域内的知识密集型数学推理》（https://arxiv.org/abs/2311.09797）。
**FinanceMath** 是一款聚焦金融领域数学推理的知识密集型数据集，要求模型理解专业金融术语，并解读题目中给出的表格数据。

<p align="center">
<img src="overview.png" width="80%">
</p>

## FinanceMath 数据集
所有数据样本被划分为两个子集：*验证集（validation）* 和 *测试集（test）*。
- **验证集（validation）**：包含200个样本，用于模型开发、验证，或供计算资源有限的使用者测试。
- **测试集（test）**：包含1000个样本，用于标准评测。我们不会公开测试集的标注解答与答案。

您可通过以下命令下载该数据集：
python
from datasets import load_dataset

dataset = load_dataset("yale-nlp/FinanceMath")

# 打印验证集的第一条样本
print(dataset["validation"][0])

# 打印测试集的第一条样本
print(dataset["test"][0])

该数据集以JSON格式提供，包含以下字段：

{
"question_id": [字符串类型] 问题编号，
"question": [字符串类型] 问题文本，
"tables": [列表类型] 与该问题关联的Markdown格式表格列表，
"python_solution": [字符串类型] 由金融专家编写的Python格式可执行解答。代码格式清晰可运行，变量命名规范且附带详细解释，
"ground_truth": [浮点类型] `python_solution` 的执行结果，保留三位小数，
"topic": [字符串类型] 该问题所属的金融领域主题，
}

## 联系方式
如有任何问题或疑问，请致信联系：赵轶伦（Yilun Zhao），邮箱：[e-mail]。

## 引用
若您在研究中使用 **FinanceMath** 数据集，请引用如下论文：

@misc{zhao2024financemath,
title={FinanceMath: Knowledge-Intensive Math Reasoning in Finance Domains},
author={Yilun Zhao and Hongjun Liu and Yitao Long and Rui Zhang and Chen Zhao and Arman Cohan},
year={2024},
eprint={2311.09797},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2311.09797},
}