PhD Defense Completed 🎓

I am happy to share that I have successfully defended my PhD thesis, titled “Learning to Guide Automated Reasoning: A GNN-Based Framework.”

Defense Details

• Date: 6 May 2025
• Time: 13:15
• Location: Room 10132 Häggsalen, Ångström

---

• Opponent: Prof. Stephan Schulz, DHBW Stuttgart, Germany
• Supervisor:
  • Prof. Philipp Rümmer, Uppsala University, Sweden & University of Regensburg, Germany
  • Prof. Yi Wang, Uppsala University, Sweden
  • Research Scientist Marc Brockschmidt, Google DeepMind, UK
• Chair: Prof. Bengt Jonsson, Uppsala University, Sweden
• Committee:
  • Prof. Mihaela Sighireanu, ENS Paris-Saclay, France
  • Reader Konstantin Korovin, University of Manchester, U.K.
  • Prof. Christian Rohner, Uppsala University, Sweden
  • Prof. Tobias Wrigstad, Uppsala University, Sweden

---

Short Introduction

Symbolic solvers rely on hand‑crafted heuristics that often fail to generalize. This thesis replaces or augments them with learned guidance: Graph Neural Networks (GNNs) trained on graph representations of Constrained Horn Clauses and word equations. Integrated into a CHC solver and a word‑equation solver, the models guide clause and branch selection and deliver consistent speedups, with practical techniques (caching, hybrid heuristics, selective queries) to control overhead. The results point to scalable, data‑driven heuristics for formal methods and motivate extensions to models, datasets, and the word‑equation pipeline.

• Online record in DiVA, Slides