---
license: apache-2.0
language:
- en
task_categories:
- question-answering
- text-generation
tags:
- math
- arithmetic
- synthetic
- short-answer
- instruction-dataset
size_categories:
- 10K<n<50K
---

# Nomi Math Dataset (20K)

## 📌 Overview
**Nomi Math Dataset** is a clean, synthetic dataset designed for training or fine-tuning small language models on **basic arithmetic**.

The dataset focuses on **short, direct answers** and intentionally avoids:
- step-by-step explanations
- reasoning chains
- meta text (e.g. "Step 1", "Explanation")

This makes it ideal for **instruction alignment** and fixing arithmetic accuracy in small models.

---

## 🎯 Tasks
- Arithmetic question answering
- Instruction following (direct answers only)

---

## ➗ Covered Operations
- Addition (`+`)
- Subtraction (`-`)
- Multiplication (`×`)
- Division (`÷`) with integer results only

Examples:
- `What is 8 times 8? → 64`
- `What is 10 divided by 2? → 5`

---

## 🧱 Data Format
Each sample is a single text field: