WideSeek-R1 Corpus

To train WideSeek-R1 efficiently, we deployed a suite of local search tools that the model can use during training.

This repository includes three components:

• wiki_corpus.jsonl: Serves as the model’s Search tool. Given a query, it returns the most relevant snippets.
• wiki_webpages.jsonl: Serves as the model’s Access tool. Given a specific URL, it returns the full webpage content.
• qdrant/: A local Qdrant vector database built by embedding wiki_corpus.jsonl. It enables efficient retrieval and acts as the core backend for the Search tool.

Both wiki_corpus.jsonl and wiki_webpages.jsonl are sourced from the ASearcher-Local-Knowledge dataset.

Acknowledgement

Thanks to ASearcher for providing a comprehensive, high-quality wiki corpus.

Citation

If you use this dataset in your research, please cite our paper:

@article{xu2026wideseek,
  title   = {WideSeek-R1: Exploring Width Scaling for Broad Information Seeking via Multi-Agent Reinforcement Learning},
  author  = {Xu, Zelai and Xu, Zhexuan and Zhang, Ruize and Zhu, Chunyang and Yu, Shi and Liu, Weilin and Zhang, Quanlu and Ding, Wenbo and Yu, Chao and Wang, Yu},
  journal = {arXiv preprint arXiv:2602.04634},
  year    = {2026},
}