Integrating Machine Learning and Knowledge Representation for Trustworthy AI

As AI models scale, they have become powerful at prediction. However, they still often struggle to reason consistently and transparently with symbolic knowledge and constraints, which can lead to factual errors or unpredictable behavior that are especially harmful in high-stakes domains such as biomedicine and healthcare. In this talk, I will present our recent neuro-symbolic approaches that directly integrate explicit symbolic knowledge, including ontologies, semantic constraints, and argumentation graphs, into modern AI systems (e.g., LLMs). These integrations can occur at different stages, such as representation, training objectives, or inference stages, allowing models to better align with domain-specific knowledge and requirements.

About Dr. Xiong

Bo Xiong is a Postdoctoral Researcher at Stanford University. He received his PhD in Computer Science (summa cum laude) from the University of Stuttgart and the International Max Planck Research School for Intelligent Systems (IMPRS-IS) in Germany. His research lies at the intersection of machine learning and knowledge representation, focusing on building neuro-symbolic models that integrate deep learning and large language models with symbolic knowledge and reasoning. He has published in leading AI venues such as NeurIPS, ICML, ICLR, KDD, ACL, etc., and his work has been recognized with the Best Student Paper Award at ISWC 2022, the 2025 SWSA Distinguished Dissertation Award, and the 2024 German Informatics Dissertation Award.