Fine-Tuning Large Language Models for Question Answering

Description

This project implements a BERT-based Question Answering system using the SQuAD 2.0 dataset. It fine-tunes pre-trained language models to accurately answer questions based on provided contexts.

Technologies/Concepts Used

• BERT
• SQuAD 2.0
• Fine-Tuning
• Natural Language Processing

Performance

• F1 Score: 76.70
• Exact Match (EM) Score: 73.85

Usage

1. Install required packages (transformers, torch, tqdm).
2. Mount Google Drive and download SQuAD 2.0 dataset.
3. Preprocess the data, train the model, and evaluate performance.
4. Use the trained model to answer questions on new contexts.