The detailed statistics of this RQ are listed as follows. For both selection strategies, we offered 5 trials and calculated the average number of verifier calls. Note that we only consider those test cases successfully handled by SpecGen. This is because multiple uncontrollable factors causing SpecGen to fail (for instance, unexpected errors within OpenJML) also affect the number of verifier calls, making the statistics on failed test cases meaningless.
We can observe that the heuristic selection strategy effectively improves the efficiency of SpecGen compared to the random selection strategy.