agentlans/nvidia-Nemotron-Science-Math

---
license: cc-by-4.0
task_categories:
- text-generation
- question-answering
language:
- en
tags:
- chain-of-thought
- reasoning
- math
- science
- nemotron
- nlp
- synthetic-data
size_categories:
- 100K<n<1M
---

# NVIDIA Nemotron Science and Math Reasoning

This is an unofficial, curated collection derived from NVIDIA's open-source Nemotron datasets. It is designed specifically to train language models in complex scientific and mathematical reasoning by providing structured chain-of-thought (CoT) examples.

To ensure efficiency, the shortest available CoT sequence was chosen for each question, filtering out redundant variations while preserving the core logical progression toward the final answer.

## Data Composition

| Source | Split/Type | Rows |
| :--- | :--- | ---: |
| [nvidia/Nemotron-Science-v1](https://huggingface.co/datasets/nvidia/Nemotron-Science-v1) | MCQ | 174&thinsp;154 |
| [nvidia/Nemotron-Math-v2](https://huggingface.co/datasets/nvidia/Nemotron-Math-v2) | RQA | 70&thinsp;054 |
| [nvidia/Nemotron-Post-Training-Dataset-v1](https://huggingface.co/datasets/nvidia/Nemotron-Post-Training-Dataset-v1) | stem | 66&thinsp;392 |
| [nvidia/Nemotron-Post-Training-Dataset-v1](https://huggingface.co/datasets/nvidia/Nemotron-Post-Training-Dataset-v1) | math | 58&thinsp;637 |
| [nvidia/Nemotron-Science-v1](https://huggingface.co/datasets/nvidia/Nemotron-Science-v1) | RQA | 52&thinsp;179 |

*Note: For the Post-Training Dataset splits, a maximum of 500&thinsp;000 rows per split was applied during processing.*

## Data Structure

Each entry is formatted as a JSON object.

* **`id`**: A unique identifier (UUID) or the original ID from the source dataset.
* **`question`**: The scientific or mathematical problem prompt.
* **`answer`**: The final concise answer.
* **`thought`**: The shortest chain-of-thought extracted from the original `<think>` tags.
* **`source`**: The original dataset and split origin.

### Example Entry

```json
{
"id": "4f184420-aea5-41e9-9a33-c398d5366895",
"question": "Solve the following problem... [Problem Details] ...",
"answer": "\\[\n\\boxed{-1.48\\text{eV}}\n\\]",
"thought": "We need to parse the problem... [Detailed reasoning steps] ...",
"source": "nvidia/Nemotron-Science-v1 RQA"
}
```

## Limitations

* **Subset Scope**: This represents only a portion of the original NVIDIA datasets.
* **Functional Constraints**: The dataset does not include multi-turn conversations, tool-calling capabilities, or visual-reasoning tasks.
* **Model Performance**: Users should be aware that LLMs may still exhibit hallucinations or errors in complex multi-step reasoning, regardless of training data quality.

## Licensing

This dataset is released under the **Creative Commons Attribution 4.0 (CC-BY-4.0)** license. Please ensure you provide proper attribution to the original creators at NVIDIA and cite this repository when utilizing this data for training or fine-tuning models.