Editors-in-Chief Neurosymbolic Artificial Intelligence Tarek R. Besold, Prof. Artur d'Avila Garcez, and Prof. Pascal Hitzler Dear Dr. Besold, Prof. d'Avila Garcez, and Prof. Hitzler, I am pleased to submit our manuscript titled “Towards Semantic Understanding of GNN Layers Embedding with Functional-Semantic Activation Mapping,” for consideration in the NeSy 2024 special issue of Neurosymbolic Artificial Intelligence. This paper extends our previous work with Functional-Semantic Activation Mapping (FSAM), introducing new insights into how varying the number of layers in Graph Neural Networks (GNNs) influences both performance and semantic alignment. Our key findings include: Layer Depth vs. Semantic Coherence: We demonstrate that additional layers may improve accuracy but often degrade semantic coherence, potentially leading to correct predictions based on misaligned representations. Neuron Specialization and Community Analysis: Deeper layers can reduce neuron specialization, resulting in overlapping communities and misclassifications within classes, signaling a loss of class-specific features. Importance of Assessing Beyond Accuracy: Our work underscores the need to assess GNNs with a balanced view of accuracy and semantic clarity to improve model interpretability. These contributions highlight the value of FSAM in diagnosing GNN behavior across different layer configurations, offering a practical framework to balance interpretability and performance in neurosymbolic AI. Thank you for considering our manuscript for the NeSy 2024 special issue. We hope that our findings will contribute meaningfully to the journal’s aim of advancing explainable and semantically coherent AI models. Sincerely, Kislay Raj Doctoral Researcher , School of Computing, Dublin City University, Ireland