LARQL treats neural network weights as a graph database to enable complex structural queries. The tool allows researchers to isolate specific weight patterns without iterating through entire tensors. LARQL simplifies the analysis of model internals. This provides a more precise way to audit LLM weights during interpretability research.