---
language: en
license: apache-2.0
viewer: false
---

# Dataset card for ArXiv Astronomy Similarity

This dataset is a random sample of [ArXiv abstracts](https://huggingface.co/NeuML/txtai-arxiv) labeled as `astro-ph`. It is intended for training vector similarity models.

It also has a [BEIR](https://github.com/beir-cellar/beir)-compatible version of the `test` split that can be used to measure the accuracy of vector models trained with this data.