Creating Neuro-Symbolic Models for Boosting Interpretability and Reasoning Capabilities of AI
Modern AI systems can recognize images, understand language, and solve complex tasks, but they often operate as “black boxes,” making it difficult to understand why they make certain decisions. This lack of transparency is a major challenge in high-stakes domains like healthcare, transportation, and public safety.
In this talk, Dr. Parth Padalkar will show how symbolic reasoning and logic can be combined with neural networks to produce clear, rule-based explanations of model decisions. He will demonstrate how compact logic-based rules can be extracted from CNNs and Vision Transformers to reveal what the model has learned, without sacrificing accuracy. The later part of the talk will also focus on improving reasoning abilities of LLMs using logic programs.
About the Speaker
Dr. Parth Padalkar is an Assistant Professor of Computer Science at Texas State University. He received his Ph.D. in Computer Science from The University of Texas at Dallas in 2026. His research lies broadly in Artificial Intelligence, with interests in neuro-symbolic AI, interpretable machine learning, knowledge representation and reasoning and computer vision. His work explores how learning-based AI systems can be integrated with symbolic representations and reasoning to improve interpretability and reasoning abilities. His broader research goal is to develop AI systems that can learn from data while also reasoning with structured knowledge, explaining their decisions, and adapting reliably when confronted with new situations.
His research has appeared in venues including AAAI, the International Conference on Logic Programming (ICLP), and the International Conference on Neuro-Symbolic Learning and Reasoning (NeSy). He has also served as a program committee member or reviewer for major AI venues including AAAI, NeurIPS, and NeSy
Help spread the word! Download and share flyer.
