In a Databricks environment, you're analyzing query performance improvements. After several runs of a complex query on a large dataset, you notice a significant reduction in latency. What feature of Databricks is most likely contributing to this decrease in query execution time?