Hangoo Kang | University of Illinois Urbana Champaign

About Me

I hope you’re doing great! I am Hangoo Kang, a senior undergraduate student at the University of Illinois Urbana Champaign (UIUC), pursuing a B.S. in Computer Science.

Currently, I am working at the FOrmally Certified Automation and Learning (FOCAL) Lab, advised by Professor Gagandeep Singh on Trustworthy AI, Agentic AI, Adversarial RLHF training, and Programming Language on AI. I am also part of the Approximate and Resilient Computing (ARC) Lab where I am advised by Professor Sasa Misailovic. With guidance from Professor Misailovic, I am researching the potential of applying programming language principles to AI systems. Specifically, I have been developing a grammar-constrained decoding algorithm to ensure syntactical correctness in AI-generated outputs. Additionally, I am collaborating with Professor Minjia Zhang on AI safety, exploring the vulnerability of state-of-the-art language models to minimal noise on their input system.

Research Interests

Machine Learning: Safety, Robustness, Agents, Multimodal Models, Bias in AI
Programming Language: Grammar Guided Generation, Software Verification, Program Synthesis
Reinforcement Learning: RLHF, Adversarial Training

News

[May. 2023] Started as Undergrad RA at UIUC’s ARC Lab
[May. 2023] Started as Undergrad RA at UIUC’s FOCAL Lab

Publications

SynCode: LLM Generation with Grammar Augmentation

Shubham Ugare, Tarun Suresh, Hangoo Kang, Sasa Misailovic, Gagandeep Singh

PDF Code Under Review

Stochastic Monkeys at Play: Random Augmentations Cheaply Break LLM Safety Alignment

Jason Vega, Junsheng Huang*, Gaokai Zhang*, Hangoo Kang*, Minjia Zhang, Gagandeep Singh

PDF Code Under Review (* indicates equal contribution)