---
license: mit
task_categories:
- text-generation
language:
- en
tags:
- arxiv
- maths
- computer-science
- physics
size_categories:
- 10K<n<100K
---

<h1 align="center">arxiv-tex-corpus-medium (15GB)</h1>
<p align="center">
Medium-scale LaTeX corpus from arXiv (math, CS, physics, statistics)
</p>

📄 Paper: https://arxiv.org/abs/2602.17288

## 📚 Overview

**arxiv-tex-corpus-medium (15GB)** is a medium-sized version of the arXiv LaTeX corpus, containing structured LaTeX source content extracted from selected arXiv categories.

This dataset is restricted to the following categories:

* `math`
* `cs`
* `hep-th`
* `hep-ph`
* `quant-ph`
* `stat.ML`
* `stat.TH`

This version (~15GB) is intended for:

* Research experiments
* Model prototyping
* Ablation studies
* Resource-constrained training setups


## 🔗 Full Dataset

For the complete version (~80GB), see:

👉 [arxiv-tex-corpus-full](https://huggingface.co/datasets/anuj0456/arxiv-tex-corpus-full)

The 80GB version contains the full dataset without size reduction.


## 📦 Files

train.jsonl
val.jsonl
checkpoint.txt
```

### `train.jsonl`

Training split in JSON Lines format.

### `val.jsonl`

Validation split in JSON Lines format.

### `checkpoint.txt`

Internal processing checkpoint file used during dataset creation.
Not required for training.

## 🧾 Data Format

Each line in `train.jsonl` and `val.jsonl` is a JSON object.

Example schema:

```json
{
  "paper_id": "xxxx.xxxxx",
  "category": "cs",
  "latex": "\\documentclass{article} ..."
}
```

### Fields

* `paper_id` — arXiv identifier
* `category` — one of:

  * `math`
  * `cs`
  * `hep-th`
  * `hep-ph`
  * `quant-ph`
  * `stat.ML`
  * `stat.TH`
* `latex` — extracted LaTeX source content

## 📊 Dataset Characteristics

* Size: ~15GB
* Format: JSONL
* Categories: mathematics, computer science, theoretical physics, quantum physics, and statistics
* Derived from the same pipeline as the 80GB version
* Snapshot date: *(add date here)*

This version is a reduced subset of the full dataset.

## 🧹 Processing Pipeline

The dataset was constructed by:

1. Downloading arXiv source archives.
2. Extracting LaTeX source files.
3. Filtering to retain only specified categories.
4. Converting documents to structured JSONL format.
5. Creating train/validation splits.
6. Subsampling to create the 15GB medium version.

## ⚖️ Licensing

This dataset contains LaTeX source files from papers hosted on arXiv.

Each paper retains its original license as specified by its authors on arXiv.

Users are responsible for complying with the licensing terms of individual papers.

For license information:
[https://arxiv.org/help/license](https://arxiv.org/help/license)

## 🎯 Intended Use

* Medium-scale pretraining
* Fine-tuning on scientific domains
* Research in mathematical reasoning
* LaTeX modeling and generation

## 🚫 Limitations

* LaTeX sources may contain compilation errors.
* Some source bundles may be incomplete.
* Licensing varies per paper.
* This dataset is a subset of the full corpus.

## 📌 Citation

If you use this dataset, please cite:

1. The original arXiv papers
2. The arxiv-latex-corpus dataset

Example:

```bibtex
@dataset{arxiv_latex_corpus_2026,
  title = {arxiv-latex-corpus},
  year = {2026},
  publisher = {Hugging Face},
}
```

🤝 Acknowledgements
This dataset is derived from papers made publicly available by authors via arXiv. We thank the research community for openly sharing their work.