PROVE-IT – Preview

PROVE-IT

Luca Marzari Technische Universität Wien

Funding program: ESPRIT

Grant DOI: 10.55776/ESP1944725

Funding amount: 349,732 €

Duration: 2025 – 2028 (3 years)

Discipline

Computer Sciences (80%) · Mathematics (20%)

Keywords

Artificial Intelligence Neural Network Verification Counterfactual Explanations Deep Reinforcement Learning Trustworthy AI Energy Systems

FWF Austrian
Science Fund

About the Project

🔍 PROVE-IT aims to advance the safety and transparency of AI systems by developing novel probabilistic verification methods and robust counterfactual explanations for sequential decision-making, with a particular focus on reinforcement learning in safety-critical domains such as energy systems. The goal is to provide scalable tools that not only certify system behavior with statistical guarantees but also explain decisions in a way that is meaningful and trustworthy for human operators.

Research Pillars

🔐 Probabilistic Verification

Developing scalable verification algorithms that provide formal statistical guarantees on the behavior of learned policies in safety-critical settings.

💡 Counterfactual Explanations

Generating robust, provably stable counterfactual explanations for sequential decisions, enabling meaningful human-interpretable feedback.

⚡ Energy Systems

Applying verification and explainability tools to real-world reinforcement learning scenarios in smart grid and energy management domains.

🤖 Safe Reinforcement Learning

Bridging formal verification with iterative decision-making to produce certified, transparent, and trustworthy autonomous agents.

People

Funding

Role	Name	Affiliation
Principal Investigator	Luca Marzari	TU Wien

This project is funded by the Austrian Science Fund (FWF) under the ESPRIT fellowship programme (Grant DOI: 10.55776/ESP1944725).