Research

Currently working in the Jacob Andreas lab on optimizing language models to collaborate with humans directly via RL.

In the past, I worked in the Isola lab on in context learning and inner optimization in transformers, and before that in the Tegmark and Solar-Lezama labs, all at MIT.

Papers:

• Breakpoint: Scalable evaluation of system-level reasoning in LLM code agents. Kaivalya Hariharan*, Uzay Girit*, Atticus Wang, Jacob Andreas. CoLM, 2025. A methodology to synthetically generate software tasks of arbitrary difficulty, with a detailed analysis of agent behavior and static measures of difficulty.
• The Quantization Model of Neural Scaling. Eric J. Michaud, Ziming Liu, Uzay Girit, Max Tegmark. NeurIPS, 2023. Modeling capability emergence in terms of structure in the task distribution of language.
• Lower Data Diversity Accelerates Training: Case Studies in Synthetic Tasks. In Submission. Suhas Kotha*, Uzay Girit*, Tanishq Kumar*, Gauran Ghosal, Aditi Raghunathan. Investigating an ICL generalization phenomenon where memorization acts as a pathway to generalization.
• Between the Bars: Gradient-based Jailbreaks are Bugs that induce Features. Kaivalya Hariharan*, Uzay Girit*. Accepted to NeurIPS ATTRIB 2024, Redteaming Adversaries. An analysis of structure and patterns in language model adversaries.

Software

• new things coming soon
• Archivy, popular self-hostable and extensible knowledge management software.
• Espial, an engine for automated organization and discovery of personal knowledge. See the demo here.
• Dust, where I worked in summer 2023 with Stanislas Polu when only 4 people were there. I optimized parts of the Rust backend and built a new power user AI collaboration product
• AdiosCorona, a general resource on COVID guidelines I built with a group of French scientists, which delivered information to millions of people during the pandemic.

See GitHub for more.