Detecting Hallucinations

This repository is part of my thesis, titled "Investigating Hallucinations in AI Text Generation Using Semantic Tools", where is presented a refined framework for detecting hallucinations in language models.

The project focuses on enhancing the framework for hallucination detection by introducing modifications in the entailment step. Specifically, we replace the Entailment LLM with a fine-tuned Transformer model for Natural Language Inference (NLI). This adjustment aims to improve the accuracy and reliability of hallucination detection in language models.

We recommend using a virtual environment, like Conda, to install the necessary packages. We use Python 3.12.5 and Pytorch 12.4.

1. Clone the repository:

   git clone https://github.com/tassos37000/dah.git
   cd dah

2. Install Pytorch following the instructions from the official website.

3. Install the required Python packages:

   pip install -r requirements.txt

4. Open and execute the notebook:

   jupyter notebook

To use the code, simply run the notebook.

This project is licensed under the GPL v3 License.