Surbhi Goel

Surbhi Goel

Assistant Professor

University of Pennsylvania, Philadelphia

I am the Magerman Term Assistant Professor of Computer and Information Science at University of Pennsylvania. I am associated with the theory group, the ASSET Center on safe, explainable, and trustworthy AI systems, and the Warren Center for network and data sciences. My research interests lie at the intersection of theoretical computer science and machine learning, with a focus on developing theoretical foundations for modern machine learning paradigms, particularly deep learning. My group’s research is genererously supported by Microsoft Research and OpenAI.

Prior to this, I was a postdoctoral researcher at Microsoft Research NYC in the Machine Learning group. I obtained my Ph.D. in the Computer Science department at the University of Texas at Austin where I was fortunate to be advised by Adam Klivans. My dissertation was awarded UTCS’s Bert Kay Dissertation award, and my Ph.D. research was generously supported by the JP Morgan AI Fellowship and fellowships from UT Austin. During my PhD, I visited IAS at Princeton and the Simons Institute for the Theory of Computing at UC Berkeley (supported by the Simons-Berkeley Research Fellowship). Before that, I received my Bachelors degree from Indian Institute of Technology (IIT) Delhi.

I am actively involved in service and outreach roles. Along with Nika Haghtalab and Ellen Vitercik, I co-founded Learning Theory Alliance (LeT‐All), a community building and mentorship initiative for the learning theory community. I am currently serving as the Office Hours co-chair for ICLR 2024, co-treasurer for the Association for Computational Learning Theory, and on the steering committee for the Association for Algorithmic Learning Theory. I am currently visiting the Simons Institute for the Theory of Computing at UC Berkeley in the role of a co‐organizer for the Special Year on Large Language Models.

Students: I currently advise Ezra Edelman (they/them). I am fortunate to also collaborate with several undergraduate, masters, and PhD students at UPenn who I do not directly advise.

Teaching: In Spring 2023 and 2024, I co-taught CIS 5200: Machine Learning with Eric Wong. In Fall 2023, I taught a special topics course CIS 7000: Foundations of Modern Machine Learning: Theory and Empirics. .

For prospective students who are interested in working with me: Please apply to the CIS PhD program and list me as a potential advisor. Unfortunately I will not be able to respond to individual emails from prospective PhD applicants at this time. If you are a current UPenn student looking to do an independent research project, send me an email with your CV, an overview of your research interests, and a brief description of 1-2 recent papers (not necessarily mine) you have read and enjoyed. I do not have any current opportunities for external students.

I started (and never updated) a blog: Unproven Algorithms.

Download my resumé.

Interests
  • Theory
  • Machine Learning
Education
  • PhD in Computer Science, 2020

    University of Texas at Austin

  • MS in Computer Science, 2019

    University of Texas at Austin

  • BTech in Computer Science and Engineering, 2015

    Indian Institute of Technology, Delhi

Recent Publications & Preprints

(2024). Logicbreaks: A Framework for Understanding Subversion of Rule-based Inference.

PDF Cite

(2024). Explicitly Encoding Structural Symmetry is Key to Length Generalization in Arithmetic Tasks.

PDF Cite

(2024). Tolerant Algorithms for Learning with Arbitrary Covariate Shift. NeurIPS 2024 [Spotlight].

PDF Cite

(2024). Stochastic Bandits with ReLU Neural Networks. ICML 2024.

PDF Cite

(2024). The Evolution of Statistical Induction Heads: In-Context Learning Markov Chains. NeurIPS 2024.

PDF Cite