News

2024

• 2024.12: I will be at ICSDS 2024 to give an invited talk titled “Mind the Filtration: E-processes vs. P-processes at Stopping Times.”
  • This talk will be based on my recent work on Combining Evidence Across Filtrations Using Adjusters.
• 2024.10: I will be at INFORMS 2024 to present my recent work (in NeurIPS'23) on Counterfactually Comparing Abstaining Classifiers. Hope to see you there!
  • Session Title: “TC57 - Causal Inference”
  • Time & Place: Tuesday, 10/22, 1:45 PM - 2:00 PM @ Summit - Terrace Suite 1
• 2024.09: I am on the 2024–2025 academic job market for faculty positions in Statistics, Data Science, Machine Learning, and Operations Research.
  • Please refer to my CV for more details.
• 2024.08: I was at JSM 2024 to present our recent work on Combining Evidence Across Filtrations Using Adjusters.
  • Slides are available here.
• 2024.07: Our recent preprint, titled The Geometry of Categorical and Hierarchical Concepts in LLMs, just won the Best Paper Award at the ICML 2024 Workshop on Mechanistic Interpretability!
  • Building upon our recent ICML paper, we formalize the representations of categorical concepts in LLMs as convex polytopes and further show that hierarchical relations between concepts are represented as orthogonality.
  • Joint work with Kiho Park, Yibo Jiang, and Victor Veitch.
• 2024.06: A substantially revised version of our paper, Combining Evidence Across Filtrations Using Adjusters, is now on arXiv.
  • The updated preprint includes new results, including a formal characterization of adjusters as “e-lifters”.
  • Joint work with Aaditya Ramdas.
• 2024.05: Our paper on The Linear Representation Hypothesis and the Geometry of LLMs is accepted to ICML 2024!
  • This is an exciting foundational work on how we can formalize different notions of linear representations in LLMs (e.g., linear probes and steering vectors) and unify those notions by identifying an inner product that encodes semantic structure.
  • Joint work with Kiho Park and Victor Veitch.
• 2024.02: Our new preprint, titled Combining Evidence Across Filtrations Using Adjusters, is now available on arXiv.
  • We address an intriguing challenge for sequential inference that arises when combining e-processes constructed on different information sets.
  • Joint work with Aaditya Ramdas.

2023

• 2023.11: Our new preprint (my first one at UChicago DSI), titled The Linear Representation Hypothesis and the Geometry of LLMs, is now available on arXiv.
  • Joint work with Kiho Park and Victor Veitch.
  • The paper is also accepted for oral and poster presentations at the NeurIPS 2023 Workshop on Causal Representation Learning (CRL).
• 2023.09: Our recent paper, titled Counterfactually Comparing Abstaining Classifiers, is accepted to NeurIPS 2023!
  • This work develops an interesting connection between abstaining classifiers, causal inference, and black-box evaluation in ML.
  • Joint work with Aditya Gangrade and Aaditya Ramdas.
• 2023.09: I am excited to start my new position as a Postdoctoral Scholar at the UChicago Data Science Institute! My mentor is Victor Veitch.

• 2023.07: Our paper on Comparing Sequential Forecasters is now accepted for publication in Operations Research!
  • This is joint work with Aaditya Ramdas.
• 2023.06: I defended my Ph.D. Thesis at Carnegie Mellon University!
  • Here are the slides from my thesis defense.
  • I am grateful to my committee members: Aaditya Ramdas (Chair), Aarti Singh, Edward Kennedy, Johanna Ziegel, and Alexander D’Amour.