Our lab investigates formal languages and automata theory. The theory of computation is the branch of computer science
and mathematics that deals with
whether or not a problem can be solved [computability]
how a problem can be solved [complexity]
using algorithms.
Today, our lab leverages these rigorous mathematical foundations to address challenges in deep learning and
LLMs. By bridging theoretical computer science and modern empirical AI,
we aim to make AI systems safer, more interpretable and more efficient.
We are mainly interested in
Formal Languages and Algorithms Simon's CongruenceCartesian TreeRegex EngineStringology
examining structural properties of various models of computations such as finite-state automata or pushdown automata
designing efficient algorithms for practical problems and establishing its correctness
improving grammar inference, pattern mining, and regular expression synthesis
Watermarking and Safety JailbreakingSteeringDetoxification
developing probability-balanced and linguistics-aware watermarking techniques for LLMs and diffusion language models
steering language models for safe generation, detoxification and mitigating algorithmic bottlenecks
designing algorithms for detecting hate speech or offensive text
LLMs for Code and Reasoning Agentic AITC-generation
applying theoretical principles to code generation
developing frameworks for automated test case generation, code time-complexity prediction, and fault localization
evaluating LLM reasoning on PSPACE-complete regex problems