Currently, most of the study of Boolean formula learning is theoretically the modeling and derivation. Less people considers efficiency improvements in practical applications. Now more mature Boolean formula learning algorithms mainly use the query model, and the query model relies on external SMT tools to answer the query problems. Although Boolean formula learning algorithm can get the right result after a polynomial number of queries, reducing the number of queries can reduce the number of using SMT tools to answer the queries, namely to reduce the computing time of the queries. This paper took the smallest monotonous assignment vector of the monotone theory to reduce the number of queries in Boolean formula learning algorithm and improved the efficiency of the learning algorithm.