Context-Based Biased LLM (CB-LLM): Case of LLM for Palestine

Context-Based Biased LLM (CB-LLM): Case of LLM for Palestine

In a world filled with competing narratives and conflicting information, it is crucial to develop technology that supports social justice and personalized education. Our team has built a Context-Based Biased Large Language Model (CB-LLM) designed to present factual and accurate information on specific areas of interest, particularly focusing on the case of Palestine.

Data Overview

The CB-LLM has been trained on a comprehensive dataset comprising various sources related to Palestine, including:

• Palestine Chronicles - Articles: 8,788 articles (62,457 chunks)
• Palestine Chronicles - News: 14,418 articles (36,372 chunks)
• Electronic Intifada - News: 2,646 articles (20,805 chunks)
• Electronic Intifada - Opinion: 1,024 articles (11,237 chunks)
• Electronic Intifada - Review: 364 articles (3,018 chunks)
• Electronic Intifada - Blog: 6,529 articles (55,266 chunks)
• Palestine Book Awards: ~250 articles (171,274 chunks)

Cut-off Date

The data used for training the CB-LLM is current as of September 30, 2023.

Demo

Experience the CB-LLM in action by visiting the demo.

This AI-driven project is a significant step towards using technology to ensure that important narratives are presented accurately and justly, contributing to the broader cause of social justice.