This repository contains our work on Natural Language Inference (NLI) using the Stanford Natural Language Inference (SNLI) Dataset. Our model demonstrates an impressive AUC of 89.3%, which is significantly higher than the hazard level of 33.3%. Although the state of the art currently stands at 93%, our model showcases the potential to bridge the gap and contribute valuable insights to the NLI research community.
Our approach to NLI tackles various linguistic challenges and reduces the impact of common pitfalls associated with current techniques. The repository contains all the code, data, and resources needed to replicate our experiments and analyze our results.
- Python 3.8 or higher
- PyTorch 1.9.0 or higher
- Transformers 4.9.2 or higher
- tqdm 4.62.0 or higher
- Clone the repository:
git clone https://github.com/Pse1234/NLI.git
- Install the required packages:
pip install -r requirements.txt
To train the model, use the following command:
python model.py
Our model achieves an accuracy of 89.3% on the SNLI Dataset. For a detailed analysis of our methodology, experiments, and results, please refer to our paper.
We welcome contributions from the research community to help improve our NLI model. Please feel free to open issues, submit pull requests, or reach out to us directly.
This project is licensed under the MIT License - see the LICENSE file for details.