Katie Kang

I am a PhD student in the Berkeley Aritifical Intelligence Research (BAIR) lab at UC Berkeley, advised by Sergey Levine and Claire Tomlin. Previously, I also did my undergrad at Berkeley.

Email / Google Scholar

Research

I am broadly interested in better understanding the inner workings of deep learning. My research recently has focused on characterizing and controlling how LLMs generalize.

What Do Learning Dynamics Reveal About Generalization in LLM Reasoning?
Katie Kang, Amrith Setlur, Dibya Ghosh, Jacob Steinhardt, Claire Tomlin, Sergey Levine, Aviral Kumar
In Submission
arxiv / code

Unfamiliar Finetuning Examples Control How Language Models Hallucinate
Katie Kang, Eric Wallace, Claire Tomlin, Aviral Kumar, Sergey Levine
NAACL 2025
arxiv / code

Deep Neural Networks Tend To Extrapolate Predictably
Katie Kang, Amrith Setlur, Claire Tomlin, Sergey Levine
ICLR 2024
arxiv / code

Multi-Task Imitation Learning for Linear Dynamical Systems
Thomas Zhang*, Katie Kang*, Bruce Lee*, Claire Tomlin, Sergey Levine, Stephen Tu, Nikolai Matni
L4DC 2023
arxiv

Lyapunov Density Models: Constraining Distribution Shift in Learning-Based Control
Katie Kang, Paula Gradu, Jason Choi, Michael Janner, Claire Tomlin, Sergey Levine
ICML 2022
arxiv / website / talk / blog post

Hierarchically Integrated Models: Learning to Navigate from Heterogeneous Robots
Katie Kang, Gregory Kahn, Sergey Levine
CoRL 2021
NeurIPS Robot Learning Workshop 2020 [Best Paper Award]
arxiv / video / website

Generalization through Simulation: Integrating Simulated and Real Data into Deep Reinforcement Learning for Vision-Based Autonomous Flight
Katie Kang*, Suneel Belkhale*, Gregory Kahn*, Pieter Abbeel, Sergey Levine
ICRA 2019
arxiv / video / code / news

Teaching

EE 126 (Probability and Random Processes), Fall 2019, Spring 2019, Fall 2018

CS 70 (Discrete Math and Probability), Summer 2018

website template