---
license: cc-by-nc-4.0
tags:
- vector-search
- multi-vector-search
- nearest-neighbors
- text-embeddings
pretty_name: Multi-Vector HNSW Benchmark Datasets
size_categories:
- 10k<n<100k
language:
- en
---

## Multi-Vector HNSW Benchmark Datasets

This repository contains benchmark datasets used by the [Multi-Vector HNSW](https://github.com/habedi/multi-vector-hnsw) project.
The datasets are from [habedi/multi-vector-search-datasets](https://huggingface.co/datasets/habedi/multi-vector-search-datasets).
Each record includes a question ID and three distinct 768-dimensional vectors representing the `title`, `body`, and `tags` of the question.
The text embeddings were generated using the `all-mpnet-base-v2` text embedding model.

There are three datasets; each includes questions from a separate Q\&A community hosted on Stack Exchange.
Each dataset consists of three files:

* `train.json`: The data used to build the index.
* `test.json`: The query data used for searching the index.
* `neighbours.json`: The ground truth, containing the actual `k=100` nearest neighbors for each item in `test.json`.

| #  | Dataset        | Q\&A Community         | Num Vectors | Dimensions | Train Size | Test Size |
|----|----------------|-------------------|-------------|------------|------------|-----------|
| 1  | `se_cs_768`    | Computer Science  | 3           | 768        | 36,712     | 4,080     |
| 2  | `se_ds_768`    | Data Science      | 3           | 768        | 26,055        | 2,895     |
| 3  | `se_p_768`     | Political Science | 3           | 768        | 11,174     | 1,242     |