Explore cutting-edge research in natural language reasoning through this insightful lecture by Rachel Rudinger from the University of Maryland. Delve into three key areas that challenge assumptions in NLP systems: Defeasible Inference, which allows models to update prior conclusions based on new evidence; a re-examination of partial-input baselines and their implications for model reasoning abilities; and an analysis of social stereotypes in generative reasoning models. Gain valuable insights into the complexities of language processing, common-sense reasoning, and the ethical considerations in AI development. Learn from Rudinger's expertise in computational semantics and her work at prestigious institutions like Johns Hopkins University and the Allen Institute for AI.