endurasolution/ron-math-dataset

---
license: apache-2.0
task_categories:
- text-generation
- table-question-answering
language:
- en
tags:
- math
- mathematics
- reasoning
- synthetic
- instruction-tuning
- openron
size_categories:
- 100M<n<1B
---

# OPENRON Math Instruction Dataset

A **massive-scale mathematical reasoning dataset** developed by **OPENRON**, designed for training and evaluating high-performance large language models (LLMs) on mathematical instruction following and reasoning tasks.

---

## Dataset Overview

The **OPENRON Math Instruction Dataset** contains high-quality, synthetic mathematical instruction–response pairs generated at scale.
It is specifically curated to support **reasoning-focused training**, including structured problem solving with step-by-step logic.

Each data sample follows a standardized schema and is suitable for supervised fine-tuning (SFT), instruction tuning, and reasoning research.

---

## Key Features

- **Large Scale**: Available in multiple sizes ranging from **1M to 500M samples**
- **Reasoning-Oriented**: Includes detailed **Chain-of-Thought (CoT)** reasoning
- **LaTeX Support**: Mathematical expressions are written in LaTeX for clarity and correctness
- **Structured Format**: JSONL with clearly separated fields
- **Model-Friendly**: Optimized for modern reasoning-capable LLMs

---

## Dataset Structure

Each JSONL entry contains the following fields:

```json
{
"instruction": "Problem description or question",
"input": "Optional additional context (may be empty)",
"output": "Detailed solution with step-by-step reasoning"
}