[CfP] Workshop on Logical Methods for Neural Network Analysis (LogicNN) @ FLoC 2026

[Nicolas Troquard]

The Workshop on Logical Methods for Neural Network Analysis (LogicNN) @ FLoC 2026 welcomes contributions that demonstrate how logical frameworks can address core challenges in neural‑network analysis.

** Invited speaker **
Martin Grohe, RWTH Aachen University

** Invited tutorial speaker **
Omri Isac, Hebrew University of Jerusalem
"Verification of DNNs with Marabou"

See details at https://perso.ens-lyon.fr/francois.schwarzentruber/research/events/logicnn2026/ .

# Important dates

- Paper submission: April 27, 2026
- Author notification: May 29, 2026
- Camera-ready submission: June 23, 2026
- Date of workshop: July 25, 2026 (full-day)

# Description

Neural networks power critical applications, from autonomous vehicles to medical diagnostics, but their opacity pose significant safety and trust challenges. Formal logical frameworks offer a rigorous path to understand, validate, and trust these models. This workshop aims to bridge the gap between logic and neural network analysis.

We invite contributions that demonstrate how logical frameworks can address core challenges in neural‑network analysis. Topics of interest include the expressivity of logical formalisms for neural networks, the formal verification of safety‑critical properties, the development of interpretation mechanisms grounded in logic, and the assessment of the computational complexity of model behavior.

- Expressivity and translation: Can logical formalisms accurately represent neural network behavior? We seek submissions exploring translation of neural architectures (CNNs, GNNs, RNNs, transformers) into logical specifications or vice versa.
- Safety verification: Can logic provide formal guarantees of neuron networks safety-critical properties?
- Logical interpretation: Can logic help building human-understandable explanations grounded in logic?
- Decidability and computational complexity: Can logic help answer foundational questions about the tractability of neural network analysis?

# Submissions

Submissions can be in three categories:

- Regular paper: at least 10 pages + bibliography
- Short paper: 5-9 pages + bibliography
- Abstract of already published paper: 1-3 pages + bibliography

The proceedings will be submitted to CEUR-WS.org for online publication.