How can we monitor, in real time, whether one uncertain prospect has any upside over another? To answer this question, we develop a novel family of sequential, anytime-valid tests for stochastic dominance (SD; also known as stochastic order…
In anytime-valid sequential inference, it is known that any admissible procedure must be based on e-processes, which are composite generalizations of test martingales that quantify the accumulated evidence against a composite null hypothesi…
(Figure courtesy: Kiho Park & Todd Nief) This paper concerns the question of how AI systems encode semantic structure into the geometric structure of their representation spaces. The motivating observation of this paper is that the natural
Understanding how semantic meaning is encoded in the representation spaces of large language models is a fundamental problem in interpretability. In this paper, we study the two foundational questions in this area. First, how are categorica…
This is a gentle, (mostly) non-technical introduction to game-theoretic statistics, e-values, and my recent paper with Aaditya Ramdas on “Combining Evidence Across Filtrations”. Originally posted in the UChicago Data Science Institute Blog.…
Informally, the “linear representation hypothesis” is the idea that high-level concepts are represented linearly as directions in some representation space. In this paper, we address two closely related questions: What does “linear represen…
(Scoring Rule = Brier. P = FiveThirtyEight; Q = Vegas-Odds.com.) Consider two forecasters, each making a single prediction for a sequence of events over time. We ask a relatively basic question: how might we compare these forecasters, eithe…
I have been an instructor for the following courses at INSEAD. Last year, I served as a postdoc mentor for the Data Science Clinic: a project-based course where students take on data science projects originating from real-world clients. I g…
Asterisks (*) denote equal contribution. “The Geometry of Categorical and Hierarchical Concepts in LLMs”K. Park, Y. J. Choe, Y. Jiang, V. VeitchInternational Conference on Learning Representations (ICLR), 2025 “The Linear Representation Hyp…
This is a gentle, (mostly) non-technical introduction to game-theoretic statistics, e-values, and my recent paper with Aaditya Ramdas on “Combining Evidence Across Filtrations Using Adjusters”. Originally posted in the UChicago Data Science…
Dr. Yo Joong “YJ” Choe is an Assistant Professor of Decision Sciences at INSEAD Singapore. He received a joint Ph.D. in Statistics and Machine Learning from Carnegie Mellon University in 2023, and he was a DSI Faraco Family Postdoctoral Fel…
In anytime-valid sequential inference, it is known that any admissible procedure must be based on e-processes, which are composite generalizations of test martingales that quantify the accumulated evidence against a composite null hypothesi…