large-traversaal/math500_urdu_final

# 📘 Math500 Urdu Final Dataset

**Dataset:** `large-traversaal/math500_urdu_final`
**Publisher:** Traversaal.ai
**Hosted on:** Hugging Face Datasets

A curated dataset of **500 mathematical problems** translated into **Urdu**, along with solutions and answers — designed for reasoning, multilingual understanding, and math-focused NLP research.

---

## 🧠 Overview

`math500_urdu_final` contains a set of challenging mathematical problems originally from the **MATH-500 subset** tasks (500 problems from the MATH benchmark), with full translations and solutions in Urdu. Each entry provides both English and Urdu representations of the problem, solution, and answer. ([Hugging Face][2])

This dataset is useful for:

* Benchmarking mathematical reasoning in multilingual large language models.
* Training or evaluating models on complex math problems in Urdu.
* Research in cross-lingual reasoning, translation quality, and multilingual education AI tools.

---

## 🗂 Dataset Structure

The dataset has **one split**:

| Split | Rows |
| ------ | ---- |
| `test` | 500 |

### Field Descriptions

Each record has the following fields:

| Field Name | Type | Description | |
| --------------- | -------- | ------------------------------------------------ | ------------------- |
| `problem` | `string` | Math problem in English. | |
| `solution` | `string` | Step-by-step solution in English (if available). | |
| `answer` | `string` | Final answer in English. | |
| `urdu_problem` | `string` | Math problem translated into Urdu. | |
| `urdu_solution` | `string` | Step-by-step solution in Urdu. | |
| `urdu_answer` | `string` | Final answer in Urdu. | |
| `subject` | `string` | Math subject area (e.g., Algebra, Geometry). | |
| `level` | `int64` | Difficulty level indicator (1–5). | |
| `unique_id` | `string` | Dataset-specific identifier. | ([Hugging Face][1]) |

---

## 🚀 How to Load

You can load this dataset using the Hugging Face `datasets` library:

```python
from datasets import load_dataset

dataset = load_dataset("large-traversaal/math500_urdu_final")
print(dataset["test"][0])
```

---

## 🔧 Potential Uses

* **Training multilingual reasoning models (Urdu).**
* **Evaluation of math problem-solving capabilities across languages.**
* **Fine-tuning Urdu LLMs for logical and symbolic reasoning.**
* **Curriculum-based educational applications in Urdu.**

## 📄 Citation

If you use this dataset in your research, please cite the **UrduBench paper**:

```bibtex
@misc{shafique2026urdubenchurdureasoningbenchmark,
title={UrduBench: An Urdu Reasoning Benchmark using Contextually Ensembled Translations with Human-in-the-Loop},
author={Muhammad Ali Shafique and Areej Mehboob and Layba Fiaz and Muhammad Usman Qadeer and Hamza Farooq},
year={2026},
eprint={2601.21000},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2601.21000}
}

# 📘 Math500乌尔语最终数据集（Math500 Urdu Final Dataset）

**数据集**：`large-traversaal/math500_urdu_final`
**发布方**：Traversaal.ai
**托管平台**：Hugging Face 数据集库（Hugging Face Datasets）

本数据集为经过精选的500道数学问题的乌尔语译本，附带完整解题过程与最终答案，专为推理能力、多语言理解研究以及面向数学的自然语言处理（Natural Language Processing, NLP）研究打造。

---

## 🧠 数据集概览

`math500_urdu_final`（Math500乌尔语最终数据集）包含一系列具有挑战性的数学问题，其原始数据来自**MATH-500子集**任务（即MATH基准测试（MATH benchmark）中的500道题目），并附带乌尔语全译本与解题过程。每条数据均提供问题、解题过程与答案的英文及乌尔语双版本。（[Hugging Face][2]）

本数据集可应用于以下场景：

* 对多语言大语言模型（Large Language Model, LLM）的数学推理能力进行基准测试
* 针对乌尔语复杂数学问题对模型进行训练或评估
* 开展跨语言推理、翻译质量评估以及多语言教育人工智能工具相关研究

---

## 🗂 数据集结构

本数据集仅包含一个划分：

| 划分名称 | 数据行数 |
| ------ | ---- |
| `test` | 500 |

### 字段说明

每条数据包含以下字段：

| 字段名 | 数据类型 | 字段说明 |
| --------------- | -------- | ------------------------------------------------ |
| `problem` | `string` | 英文数学问题。 |
| `solution` | `string` | 英文分步解题过程（如可获取）。 |
| `answer` | `string` | 英文最终答案。 |
| `urdu_problem` | `string` | 乌尔语译本数学问题。 |
| `urdu_solution` | `string` | 乌尔语分步解题过程。 |
| `urdu_answer` | `string` | 乌尔语最终答案。 |
| `subject` | `string` | 数学学科领域（例如代数、几何学）。 |
| `level` | `int64` | 难度等级标识（1~5级）。 |
| `unique_id` | `string` | 数据集专属标识符。（[Hugging Face][1]）

---

## 🚀 数据集加载方式

可通过Hugging Face的`datasets`库加载本数据集：
python
from datasets import load_dataset

dataset = load_dataset("large-traversaal/math500_urdu_final")
print(dataset["test"][0])

---

## 🔧 潜在应用场景

* **针对乌尔语训练多语言推理模型**
* **跨语言数学解题能力评估**
* **针对逻辑与符号推理对乌尔语大语言模型（Large Language Model, LLM）进行微调**
* **基于课程体系的乌尔语教育应用**

## 📄 引用规范

若您在研究中使用本数据集，请引用**UrduBench**相关论文：
bibtex
@misc{shafique2026urdubenchurdureasoningbenchmark,
title={UrduBench: An Urdu Reasoning Benchmark using Contextually Ensembled Translations with Human-in-the-Loop},
author={Muhammad Ali Shafique and Areej Mehboob and Layba Fiaz and Muhammad Usman Qadeer and Hamza Farooq},
year={2026},
eprint={2601.21000},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2601.21000}
}