ChuanMeng's picture
Add task category, GitHub link and sample usage (#1)
c1e781a
metadata
license: mit
task_categories:
  - text-retrieval
tags:
  - retrieval-augmented-generation
  - deep-research
  - search

Passage Corpus (Pyserini Format) for the BrowseComp-Plus Dataset

This repository provides the passage corpus for the BrowseComp-Plus dataset in a format compatible with Pyserini, used in the paper Revisiting Text Ranking in Deep Research.

The corpus consists of 2,772,255 passages.

Unlike the version available at
https://huggingface.co/datasets/grill-lab/browsecomp-plus-passage-corpus, this release follows the Pyserini JSONL data format.

Each item contains two fields:

  • id: the unique passage identifier.
  • contents: the concatenation of the source document title and the passage text.

This format is directly compatible with Pyserini BM25.

Sample Usage

As described in the official repository, you can build a Lucene index using Pyserini with the following command:

python -m pyserini.index.lucene \
--collection JsonCollection \
--input /path/to/downloaded/data/ \
--index ./indexes/index.bm25.passage \
--generator DefaultLuceneDocumentGenerator \
--threads 16 \
--storePositions --storeDocvectors --storeRaw

Contact

If you have any questions or suggestions, please contact:

Citation

If you find this work useful, please cite:

@article{meng2026revisiting,
  title={Revisiting Text Ranking in Deep Research},
  author={Meng, Chuan and Ou, Litu and MacAvaney, Sean and Dalton, Jeff},
  journal={arXiv preprint arXiv:2602.21456},
  year={2026}
}