LOGICNN 2026
Logical Methods for Neural Network Analysis 2026

Table of Contents

• Preface
  Summary: There were 9 papers submitted for peer-review to this workshop: four regular papers, two short papers, three posters. All 9 papers were accepted for this volume.

Invited Talk

• Logic and the Power of Recurrent Graph Neural Networks - Abstract (invited paper)
  Martin Grohe 1

Invited Tutorial

• Verification of DNNs with Marabou - Abstract (invited paper) 2
  Omri Isac

Session 1 - Expressivity

• Expressive Power of Graph Transformers via Logic (poster paper) 3–7
  Veeti Ahvonen , Maurice Funk , Damian Heiman , Antti Kuusisto , Carsten Lutz
• Towards Understanding the Expressive Power of GNNs with Global Readout (poster paper) 8–13
  Maurice Funk , Daumantas Kojelis
• An Algebraic Characterization of Local Weisfeiler–Leman (short paper) 14–22
  Yi Chen , Eugenia Ternovska
• Words and Temporal Graphs: Comparing the Expressive Power of State Space Models and Recurrent Neural Networks (short paper) 23–30
  Eric Alsmann , Lowejatan Noori , Martin Lange

Session 2 - Verification

• Towards Continuous Constraint Programming for Sound Neural Network Verification 31–43
  Yi-Nung Tsao , Pierre Talbot
• IsaGrad: Verified Automatic Differentiation over Computational Graphs in Imperative HOL 44–67
  Alex Raymond Hyman , Filip Smola , Mark Chevallier , Jacques D. Fleuriot

Session 3 - From Neural Networks to Logic

• Neural Networks as Fuzzy Logic Formulas 68–103
  Damian Heiman , Antti Kuusisto , Esko Turunen
• Neural Networks into Łukasiewicz Logic, with Applications to Formal Verification (poster paper) 104–109
  Sandro Preto , Marcelo Finger
• From MLPs to Logic: An End-to-End Neuro-Symbolic Compilation Pipeline for Explainable Clinical Classification 110–121
  Mashael Alsowail , Hsi-Ming Ho , Ian Mackie