Najla Alhamdan

تخصص الذكاء الاصطناعي · دفعة 2023

The Bias Within: Evaluating the Generation and Classification Dichotomy in Large Language Models

This project investigates bias detection in text generation by large language models (LLMs), focusing on GPT-3.5. The team generated a dataset of 800 storiesin English and Arabic with varying levels of bias. The study revealed a disparity between an LLMs' generative and classification capabilities, emphasizing the need for task-specific training for better bias analysis.

اطّلع عليه في تقرير البرنامج (PDF)