rank1-run-files: Pre-computed Run Files for Reranking Evaluation

📄 Paper | 🚀 GitHub Repository

This dataset contains pre-computed run files used by the rank1 family of models on various retrieval benchmarks. These files are what were used for top-k rereranking and also include the re-annotated DL19 qrels. These files are needed to download to reproduce our results.

Benchmarks Included

The dataset includes run files for the following benchmarks:

• BEIR (multiple datasets including NFCorpus, SciFact, etc.)
• NevIR
• TREC-DL 2019
• BRIGHT

Associated Models and Resources

Resource	Description
rank1-7b	Base rank1 model (7B parameters)
rank1-14b	Larger rank1 variant (14B parameters)
rank1-32b	Largest rank1 variant (32B parameters)
rank1-mistral-2501-24b	Mistral-based rank1 variant (24B parameters)
rank1-llama3-8b	Llama 3.1-based rank1 variant (8B parameters)
rank1-r1-msmarco	All R1 output examples from MS MARCO
rank1-training-data	Training data used for rank1 models

Citation

If you use these run files in your research, please cite:

@misc{weller2025rank1testtimecomputereranking,
      title={Rank1: Test-Time Compute for Reranking in Information Retrieval}, 
      author={Orion Weller and Kathryn Ricci and Eugene Yang and Andrew Yates and Dawn Lawrie and Benjamin Van Durme},
      year={2025},
      eprint={2502.18418},
      archivePrefix={arXiv},
      primaryClass={cs.IR},
      url={https://arxiv.org/abs/2502.18418}, 
}

License

MIT License