Bio
Hossein A. (aka Saeed) is a PhD student in Foundation AI at the University College London (UCL), affiliated with the Web Intelligence Group (WI) at UCL Centre for Artificial Intelligence (UCL AI Center). His primary research interests are centred around natural language processing and information retrieval. His PhD research is particularly focused on the evaluation of large language models (LLMs), exploring the dynamic roles these models play as both evaluators and evaluatees. This dual perspective is critical for advancing the understanding and development of LLMs, as it addresses both their capabilities and limitations in various contexts. Before starting his PhD at UCL, Saeed was awarded an MSc in Computer Science from the University of Zanjan, where he worked on context awareness and responsibilities aspects of recommendation systems.
Research interests
Large Language Models (LLMs) like ChatGPT are advanced machine learning systems trained on vast, diverse datasets to perform a wide range of tasks. These models have shown impressive performance on new tasks by simply following given instructions, known as "prompts." However, current evaluation methods have limitations that hinder our understanding of their capabilities, particularly in self-evaluation scenarios. Saeed's research aims to develop new methods for evaluating the strengths and weaknesses of LLMs, enabling researchers in academia and industry to make informed decisions about the safety and utility of these models. Saeed will focus on benchmarking these models and identifying the limitations introduced when the same LLM is used as both the evaluator and the evaluatee.