Helios-R-6M


# Helios-R-6M

> **Helios-R-6M** is a high-quality, compact reasoning dataset designed to strengthen multi-step problem solving across mathematics, computer science, and scientific inquiry. While the dataset covers a range of disciplines, math constitutes the largest share of examples and drives the reasoning complexity.

---

## Quick Start with Hugging Face Datasets🤗

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

```py
from datasets import load_dataset

dataset = load_dataset("prithivMLmods/Helios-R-6M", split="train")
```

---

## Overview

* **Total Samples**: \~6,186,188
* **Split**: `train` only
* **Languages**: English
* **Format**: Apache Arrow (auto-converted to Parquet)
* **License**: Apache-2.0
* **Tags**: `math`, `code`, `science`, `reasoning`, `thinking`, `cot`

## Highlights

* Focused on generating thoughtful, chain-of-thought solutions with logical progression and clarity.
* Includes both formal mathematical problems and abstract scientific or logical challenges.
* Suitable for model pretraining and fine-tuning in STEM and reasoning-intensive domains.

## Dataset Structure

Each sample in the dataset includes:

* **`problem`** (string): A question or prompt from math, science, logic, or coding.
* **`solution`** (string): A structured, step-by-step reasoning trace usually starting with `<think>`, simulating an instructive problem-solving mindset.

---

## Source & Derivation

Helios-R-6M is a refined and optimized blend derived from:

* [prithivMLmods/Poseidon-Reasoning-5M](https://huggingface.co/datasets/prithivMLmods/Poseidon-Reasoning-5M)
* [glaiveai/reasoning-v1-20m](https://huggingface.co/datasets/glaiveai/reasoning-v1-20m)
* [prithivMLmods/Open-Omega-Explora-2.5M](https://huggingface.co/datasets/prithivMLmods/Open-Omega-Explora-2.5M)
* Additional custom modular problems curated and contributed by **prithivMLmods**

This combination allows for both depth and breadth across technical and academic disciplines, with special attention to multi-domain reasoning performance.

## License

Apache License 2.0


# Helios-R-6M

> **Helios-R-6M** 是一款高质量、轻量化的推理数据集，旨在强化数学、计算机科学与科学探究领域的多步问题求解能力。尽管该数据集覆盖多门学科，但数学相关样本占比最高，且决定了数据集的推理复杂度。

---

## 基于Hugging Face Datasets🤗快速上手

bash
pip install -U datasets

python
from datasets import load_dataset

dataset = load_dataset("prithivMLmods/Helios-R-6M", split="train")

---

## 数据集概览

* **总样本量**：约6,186,188条
* **数据集拆分**：仅包含训练集（`train`）
* **语言**：英语
* **存储格式**：Apache Arrow格式（可自动转换为Parquet格式）
* **开源协议**：Apache-2.0
* **标签**：`math`（数学）、`code`（代码）、`science`（科学）、`reasoning`（推理）、`thinking`（思维）、`cot`（思维链，Chain-of-Thought）

## 数据集亮点

* 专注于生成具备逻辑递进与清晰表达的高质量思维链解决方案
* 涵盖正式数学问题与抽象科学、逻辑挑战两类题型
* 适用于科学、技术、工程与数学（STEM）领域以及推理密集型场景的大语言模型预训练与微调

## 数据集结构

该数据集的每条样本包含以下字段：

* **`problem`**（字符串类型）：来自数学、科学、逻辑或编程领域的问题或提示文本
* **`solution`**（字符串类型）：结构化的分步推理轨迹，通常以`<think>`开头，模拟具有教学意义的问题解决思维过程

---

## 数据集来源与衍生说明

Helios-R-6M 是一款经过优化与精炼的混合数据集，衍生自以下数据集：

* [prithivMLmods/Poseidon-Reasoning-5M](https://huggingface.co/datasets/prithivMLmods/Poseidon-Reasoning-5M)
* [glaiveai/reasoning-v1-20m](https://huggingface.co/datasets/glaiveai/reasoning-v1-20m)
* [prithivMLmods/Open-Omega-Explora-2.5M](https://huggingface.co/datasets/prithivMLmods/Open-Omega-Explora-2.5M)
* 由**prithivMLmods**整理并贡献的额外自定义模块化问题

该组合兼顾了技术与学术领域的深度与广度，特别针对多领域推理性能进行了优化。

## 开源协议

Apache License 2.0