Instructions to use SparseCL/UAE-SparseCL-msmarco with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use SparseCL/UAE-SparseCL-msmarco with Transformers:
# Load model directly from transformers import AutoTokenizer, our_BertForCL tokenizer = AutoTokenizer.from_pretrained("SparseCL/UAE-SparseCL-msmarco") model = our_BertForCL.from_pretrained("SparseCL/UAE-SparseCL-msmarco", device_map="auto") - Notebooks
- Google Colab
- Kaggle
File size: 673 Bytes
eae6b19 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 | ---
license: mit
---
# UAE-SparseCL-msmarco
This repository contains a SparseCL model based on `WhereIsAI/UAE-Large-V1` and trained on the MS MARCO dataset.
SparseCL is designed for contradiction retrieval.
For implementation and usage details, please see the [SparseCL repository](https://github.com/xuhaike/SparseCL).
## Citation
Please cite our paper if you use this model:
```bibtex
@inproceedings{xu2025contradiction,
title={Contradiction Retrieval via Contrastive Learning with Sparsity},
author={Xu, Haike and Lin, Zongyu and Chang, Kai-Wei and Sun, Yizhou and Indyk, Piotr},
booktitle={International Conference on Machine Learning},
year={2025}
}
``` |