---
license: mit
language:
- en
library_name: transformers
tags:
- text-generation
- information-retrieval
- ranking
- reranking
- blockrank
- mistral
base_model: mistralai/Mistral-7B-Instruct-v0.3
datasets:
- quicktensor/blockrank-msmarco-train-10p
metrics:
- ndcg
- mrr
---

# BlockRank-Mistral-7B: Scalable In-context Ranking with Generative Models

[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/nilesh2797/BlockRank/blob/main/quickstart.ipynb)

**BlockRank-Mistral-7B** is a fine-tuned version of [Mistral-7B-Instruct-v0.3](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.3) optimized for efficient in-context document ranking. It implements BlockRank, a method that makes LLMs efficient and scalable for ranking by aligning their internal attention mechanisms with the structure of the ranking task.

<p align="center">
  <img src="https://raw.githubusercontent.com/nilesh2797/BlockRank/main/assets/blockrank_diagram.png" alt="BlockRank Architecture" width="600"/>
</p>

### Key Features

- **Linear Complexity Attention**: Structured sparse attention reduces complexity from O(n²) to O(n)
- **2-4× Faster Inference**: Attention-based scoring eliminates autoregressive decoding
- **Auxiliary Contrastive Loss**: Mid-layer contrastive objective improves relevance signals
- **Strong Zero-shot Generalization**: SOTA performance on BEIR benchmarks

## Citation

If you use this model, please cite:

```bibtex
@article{gupta2025blockrank,
  title={Scalable In-context Ranking with Generative Models},
  author={Gupta, Nilesh and You, Chong and Bhojanapalli, Srinadh and Kumar, Sanjiv and Dhillon, Inderjit and Yu, Felix},
  journal={arXiv preprint arXiv:2510.05396},
  year={2025}
}
```

## Model Card Contact

For questions or issues, please open an issue on [GitHub](https://github.com/nilesh2797/BlockRank/issues).

## Additional Resources

- **Paper**: [arXiv:2510.05396](https://arxiv.org/abs/2510.05396)
- **Code**: [GitHub Repository](https://github.com/nilesh2797/BlockRank)
- **Dataset**: [HuggingFace Dataset](https://huggingface.co/datasets/nilesh2797/icr-msmarco-10p-train)
- **Demo**: [Colab Notebook](https://colab.research.google.com/github/nilesh2797/BlockRank/blob/main/examples/quickstart.ipynb)

## License

This model is released under the MIT License. See [LICENSE](https://github.com/nilesh2797/BlockRank/blob/main/LICENSE) for details.