Open-Omega-Forge-1M


# **Open-Omega-Forge-1M**

> Open-Omega-Forge-1M is a carefully curated and optimized collection derived from multiple high-quality datasets, specifically designed to enhance reasoning capabilities across mathematical, scientific, and coding domains. This dataset represents a focused subset that maintains the quality and diversity of reasoning patterns while providing a more manageable size for training and evaluation. A high-quality, compact reasoning dataset designed for mathematics, code, and science applications, with mathematics playing a major role in the dataset composition.

> Mixture of Mathematics, Coding, and Science.

---

## Overview

- **Dataset Name:** Open-Omega-Forge-1M
- **Shaped & Curated by:** PrithivMLMods
- **Size:** 1M examples (~2.17GB in train split [partial])
- **Formats:** `.arrow`, Parquet
- **Languages:** English
- **License:** Apache-2.0

## Key Features

- **Compact and Clean:** Focuses on concise, high-signal problems with clear solutions.
- **Interdisciplinary:** Integrates math, programming, and scientific reasoning tasks.
- **Optimization:** Derived from several leading open datasets, further filtered and enhanced for quality and diversity.
- **Cutoff Emphasis:** Majority of the dataset is math-centric, supporting broader STEM modeling needs.

---

## Quick Start with Hugging Face Datasets🤗

```py
pip install -U datasets
```

```py
from datasets import load_dataset

dataset = load_dataset("prithivMLmods/Open-Omega-Forge-1M", split="train")
```

## Dataset Structure

Each entry contains:
- **problem:** The problem statement (text)
- **solution:** The step-by-step, reasoning-focused solution

Example schema:

| Column | Type | Description |
|----------|--------|----------------------------|
| problem | string | Description of the problem |
| solution | string | Reasoning-based solution |

## Response format:

```py
<think>
-- reasoning trace --
</think>
-- answer --
```

---

## Data Sources

Open-Omega-Forge-1M is a carefully curated and optimized derivative collection, sourced from the following open datasets:

- Curated and blended modular dataset from [PrithivMLmods](https://huggingface.co/prithivMLmods). [others]
- XenArcAI/MathX-5M
- nvidia/OpenCodeReasoning
- nvidia/OpenScience
- OpenMathReasoning

> [!note]
The collection also includes custom modular data from prithivMLmods, meticulously blended to ensure quality, balance, and diversity.

## Applications

Open-Omega-Forge-1M is ideal for:
- Training and evaluation of large language models (LLMs) in STEM fields
- Benchmarking math and multi-step reasoning models
- Research in mathematical problem solving, algorithmic code understanding, and scientific reasoning

---

## Citation

If you use this dataset, please cite:

```bitex
Open-Omega-Forge-1M by prithivMLmods
Derived and curated from:
- Curated and blended modular dataset from prithivMLmods. [others]
- XenArcAI/MathX-5M
- nvidia/OpenCodeReasoning
- nvidia/OpenScience
- OpenMathReasoning
```

## License

This dataset is distributed under the Apache-2.0 License. Please review the license terms of all referenced underlying datasets before use.

# **Open-Omega-Forge-1M**


> Open-Omega-Forge-1M 是一款经精心甄选与优化的数据集集合，源自多个高质量开源数据集，专为提升数学、科学与编码领域的推理能力而打造。本数据集为精选子集，既保留了推理模式的质量与多样性，又将规模控制在便于训练与评估的范围内。这是一款高质量、轻量化的推理数据集，面向数学、代码与科学应用，其中数学内容占主要组成部分。

> 涵盖数学、编码与科学领域

---

## 数据集概览

- **数据集名称：** Open-Omega-Forge-1M
- **整理与甄选方：** PrithivMLMods
- **规模：** 100万条样本（训练拆分集[部分]约2.17GB）
- **存储格式：** `.arrow`、Parquet
- **语言：** 英语
- **许可证：** Apache-2.0

## 核心特性

- **轻量化且简洁：** 聚焦于表述精炼、高价值且附带清晰解答的问题
- **跨学科性：** 整合数学、编程与科学推理任务
- **优化处理：** 源自多个主流开源数据集，经进一步筛选以提升质量与多样性
- **侧重数学占比：** 数据集主体以数学内容为核心，可满足更广泛的STEM（科学、技术、工程、数学）建模需求

---

## Hugging Face Datasets 🤗 快速上手

py
pip install -U datasets

py
from datasets import load_dataset

dataset = load_dataset("prithivMLmods/Open-Omega-Forge-1M", split="train")

## 数据集结构

每条样本包含以下字段：
- **problem：** 问题描述（文本格式）
- **solution：** 基于推理的分步解答

示例字段结构：

| 字段名 | 类型 | 说明 |
|----------|--------|----------------------------|
| problem | 字符串 | 问题描述 |
| solution | 字符串 | 基于推理的解答 |

## 响应格式：

py
<think>
-- 推理过程 --
</think>
-- 回答内容 --

---

## 数据来源

Open-Omega-Forge-1M 是一款经精心筛选与优化的衍生数据集，其数据源自以下开源数据集：

- 由[PrithivMLMods](https://huggingface.co/prithivMLmods)构建的甄选混合模块化数据集。[其他来源]
- XenArcAI/MathX-5M
- nvidia/OpenCodeReasoning
- nvidia/OpenScience
- OpenMathReasoning

> 【注】本数据集还包含PrithivMLMods定制的模块化数据，经精心混合以确保数据质量、平衡性与多样性。

## 应用场景

Open-Omega-Forge-1M 适用于以下场景：
- 科学、技术、工程、数学（STEM）领域大语言模型(LLM)的训练与评估
- 数学与多步推理模型的基准测试
- 数学问题求解、算法代码理解与科学推理相关研究

---

## 引用方式

若使用本数据集，请按以下方式引用：

bitex
Open-Omega-Forge-1M by prithivMLmods
Derived and curated from:
- Curated and blended modular dataset from prithivMLmods. [others]
- XenArcAI/MathX-5M
- nvidia/OpenCodeReasoning
- nvidia/OpenScience
- OpenMathReasoning

## 许可证

本数据集采用Apache-2.0许可证分发。使用前请查阅所有引用的底层数据集的许可证条款。