Understanding LLMs through Generative Behavior, Successes and Shortcomings

Explore the generative capabilities of large language models (LLMs) in this insightful lecture presented by Swabha Swayamdipta from USC Information Sciences Institute. Delve into the research findings that shed light on LLMs' behavior, successes, and limitations. Discover how standard inference algorithms effectively work with LLMs due to their specific design. Examine recent successes and failures of LLMs across various tasks, including those requiring world knowledge, domain-specific expertise, linguistic capabilities, and awareness of human and social utility. Gain a complex yet partial understanding of LLMs and learn about potential future directions in the field. Benefit from the expertise of Swabha Swayamdipta, an Assistant Professor of Computer Science at USC, whose research focuses on natural language processing, machine learning, and the evaluation of generative language models.