IPA Lab
Interactive Programming
& Analysis Lab
We see the elements of play and exploration in activities of software and data engineering as an essential part of developing larger computational systems.
Right now, we're wondering…
-
Can LLM agents autonomously compromise enterprise networks, and how should such capabilities be governed?
-
How effective are AI agents at repairing bugs in large industrial codebases?
-
Which strategies do LLM agents follow when solving tasks, and how can we make them visible?
-
How can developers analyze and debug probabilistic programs?
-
Can we recover the grammars implicit in hand-written parsers?
-
What role should software models play in AI-driven software engineering?
approach
- 01Empirical Understanding. We want to understand the underlying structures and artifacts that govern exploration and experimentation in computational tasks.
- 02System Building. We design, build, and evaluate interventions in the form of systems that either automate or augment human ability to better deal with these structures and artifacts.
themes
teaching
latest papers
- ATLAS: Discovering Agent Strategies through LLM-Guided Abstraction and Automata Learning
- Adversarial Bug Reports as a Security Risk in Language Model-Based Automated Program Repair
- Can LLMs Hack Enterprise Networks? - Replicated Computational Results (RCR) Report
- Can LLMs Hack Enterprise Networks? Autonomous Assumed Breach Penetration-Testing Active Directory Networks
people
students
We're currently advising as many students as we can while still giving each of them the time and attention they deserve, so we aren't taking on new theses or projects right now. We don't have any open positions at the moment either.