Preslav Nakov: Towards Truly Open, Language-Specific, Safe, Factual, and Specialized Large Language Models (MBZUAI, Abu Dhabi and HSE Faculty of Computer Science)
In this online-lecture Preslav Nakov, a professor at the Mohamed bin Zayed University of Artificial Intelligence:
– will discuss the urgent need for transparent open-source large language models (LLMs) and highlight MBZUAI's LLM360 initiative;
– will share experience in creating language-specific LLMs, such as Jais, an Arabic-centric model, and Nanda, an open Hindi LLM;
– will introduce Do-Not-Answer, a dataset for evaluating LLM safety, and present tools developed at MBZUAI to address factuality challenges: OpenFactCheck for fact-checking, LM-Polygraph for uncertainty prediction, and LLM-DetectAIve for detecting machine-generated text;
– and finally, will advocate for specialized models and showcase the variety of LLMs being developed at MBZUAI's Institute on Foundation Models (IFM).
When: 13 December, 17:00
Language: English