File size: 1,191 Bytes
4632959
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
---
dataset: unarxive_E5
title: UnarXive 2024 - Dense Vector Index (e5)
license: mit
tags:
  - dense-retrieval
  - e5
  - rag
  - scholarly-nlp
  - scientific-papers
  - vector-index
  - unarxive
pretty_name: UnarXive E5 Index
size_categories:
  - 1M<n<10M
source_datasets:
  - unarxive_2024
dataset_type: embeddings
---

# Dataset Card for UnarXive E5 Index

This repository contains a **dense vector index** built using the [e5-large-v2](https://huggingface.co/intfloat/e5-large-v2) model on the **abstracts** of over 2.3 million arXiv papers in the [UnarXive 2024](https://huggingface.co/datasets/ines-besrour/unarxive_2024) dataset.

It is designed for **neural retrieval**, especially in **retrieval-augmented generation (RAG)** pipelines, scientific QA systems, and dense ranking models.

## What's Inside

- Dense embeddings of paper abstracts (`e5-large-v2`)
- Index format: FAISS (flat or HNSW depending on version)
- Each vector is associated with `paper_id`, `title`, and `abstract`

## Use Cases

- Dense retrieval in scholarly QA or chatbots
- Scientific search engines with semantic understanding
- RAG systems over structured academic corpora
- Paper similarity or clustering