On Reasoning LLMs - Myths, Merits, and How to Move Forward

Attend this keynote presentation from KDD 2025 where Dan Roth, Eduardo D. Glandt Distinguished Professor at the University of Pennsylvania and Chief AI Scientist at Oracle, examines the current state and future potential of reasoning capabilities in Large Language Models. Explore the myths surrounding LLM reasoning abilities while discovering their genuine merits and limitations in complex problem-solving scenarios. Learn from Roth's extensive experience leading scientific efforts behind AWS's first-generation Generative AI products, including Titan Models, Amazon Q, and Bedrock, as he provides insights into how the field can advance beyond current reasoning paradigms. Gain perspective from a John McCarthy Award recipient and Fellow of AAAS, ACM, AAAI, and ACL on the theoretical and practical challenges of developing more sophisticated reasoning systems. Understand the intersection of natural language processing, machine learning, and knowledge representation as applied to modern AI systems, drawing from decades of research and industry experience in cognitive analytics and machine learning applications.