Adarsh Vatsa
PhD student, Computer Science · Stevens Institute of Technology
I'm interested in reasoning. How it works, where it breaks, and what it takes to make it hold up, in machines and in people. Most of my days go to language models and formal methods, and most of my evenings go to mathematics and whatever question won't leave me alone.
This site is a notebook. It holds essays, half-finished ideas and notes on what I'm reading, kept in public so I have to think them through properly.
Writing
- A notebook, in publicWhy this site exists, and the few rules it follows.
Research
I study how to make machine reasoning checkable. That means pairing language models with formal tools, so what a model produces can be verified instead of trusted, and then using the verifier's feedback to make the model better. My current thread applies this to access-control policy synthesis. The broader question is how learning and proof can work together.
interests language models · formal verification · reinforcement learning · deep learning · mathematics
Selected papers
- RAISE: Reinforcing Access Control Policy Synthesis in LLMs via Symbolic Evaluation
- AutoCedar: An Agentic Framework for Verifier-Guided Access Control Policy Synthesis
- Neurosymbolic Characterization for Reliable Access Control Policy Analysis
- Synthesizing Access Control Policies Using Large Language Models
About
I'm a PhD student at Stevens Institute of Technology in Hoboken, New Jersey. Email is the best way to reach me, and I'm always happy to hear from people thinking about similar things.
- emailavatsa@stevens.edu
- scholarGoogle Scholar
- githubadarsh-vatsa
- feedRSS