LARQL treats neural network weights as a graph database to enable complex structural queries. This tool allows researchers to isolate specific weight patterns without manual tensor slicing. It simplifies the analysis of model internals. Practitioners can now audit weight distributions using a familiar query language instead of writing custom Python scripts.