Ashutosh Trivedi

Associate Professor of Computer Science · University of Colorado Boulder

My research develops formal methods for reinforcement learning and trustworthy AI, with a focus on verification and accountability in high-stakes decision-making systems.

CV Google Scholar DBLP Email
Ashutosh Trivedi

Formal Methods for Reinforcement Learning

Foundations and algorithms for verifying, synthesizing, and reasoning about reinforcement-learning systems, including temporal objectives, recursion, and symbolic representations.

Trustworthy Reasoning with Learning and LLMs

Combining formal verification, symbolic reasoning, and learning to produce explanations and guarantees that are checkable, not just plausible.

AI, Software, and Accountability

Methods for auditing, testing, and debugging high-stakes software systems, with applications to fairness, legal compliance, and socio-technical decision-making.

Secure & Safe Cyber-Physical Systems

Rigorous methods for security, privacy, and safety of cyber-physical and learning-enabled systems, with applications to medical devices and critical infrastructure.

Selected publications

Stochastic Neural Simulation Relations for Control Transfer

🏆 DARPA Disruptive Idea Award
cps award
NeuS · 2025
PDF

Regular Reinforcement Learning

🏆 CAV Distinguished Paper Award
award formalrl trustworthyAI
CAV · 2024
PDF Details

Closure Certificates

cps theory
HSCC 2024 · 2024
PDF Details

Metamorphic Testing and Debugging of Tax Preparation Software

accountableSE trustworthyAI
ICSE-SEIS 2023 · 2023
PDF Details

Recursive Reinforcement Learning

formalrl theory trustworthyAI
NeurIPS · 2022
PDF

Omega-Regular Objectives in Model-Free Reinforcement Learning

formalrl theory trustuworthyAI
TACAS · 2019
PDF Details

Join the group

PhD students & postdocs

I’m recruiting PhD students and occasional postdocs interested in formal methods, reinforcement learning, cyber-physical systems, and AI accountability. Strong theoretical foundations and curiosity about real-world impact are especially welcome.
Current openings →

Recent news

Aug 2025
Teaching CSCI 5444 (Introduction to the Theory of Computation) this term.
Aug 2025
Our research on AI explainability in Sudoku was featured in CNET.
Aug 2025
New arXiv preprint with Maria Leonor Pacheco, Fabio Somenzi, and Dananjay Srinivas: Explaining Hitori Puzzles: Neurosymbolic Proof Staging for Sequential Decisions.
Aug 2025
Paper accepted at ASE 2025: Uncovering Discrimination Clusters: Quantifying and Explaining Systematic Fairness Violations.
Aug 2025
Paper accepted for presentation at CDC 2025: Objective Improvement Algorithm for Controller Synthesis in Uncertain Environments.

All news →