Dataset Viewer
The dataset viewer is not available for this subset.
Job manager crashed while running this job (missing heartbeats).
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Dataset Card for UnarXive BM25 Index
This repository contains the BM25 sparse index built from the UnarXive 2024 dataset, which includes over 2.3 million arXiv papers (1991–2024) with rich metadata, structured full text, and linked citations.
The index was created using paper abstracts and is designed for fast keyword-based retrieval in scientific information retrieval applications.
What’s Inside
- BM25 index (e.g., Faiss-compatible or pyserini-compatible, depending on format)
- Indexed fields:
paper_id,title,abstract - Based on: ines-besrour/unarxive_2024
Use Cases
- Efficient retrieval in scholarly QA or RAG pipelines
- Search engines over academic papers
- Benchmarking sparse retrieval methods
- Downloads last month
- 19