“…We now attempt to establish that augmenting k -induction with auxiliary invariants from KIPDR improves its overall effectiveness: In the previous experiment, we presented CPAchecker-CTIGAR, which uses an adaptation of PDR as its verification engine. When we compare the results of CPAchecker-CTIGAR from Table 2 to the results from evaluations [4,5] of other techniques for previous versions of the same benchmark set, we see that neither of the two CTIGAR implementations is competitive. For example, running the k -induction configuration of CPAchecker without auxiliary-invariant generation on our benchmark set, we obtain 1 239 correct proofs and 836 correct alarms, as shown in the fourth column of Table 2 (KI). In general, however, the strength of PDR is considered to be its capability for generating safety invariants, so that it is more interesting to analyze its usefulness as an invariant generator.…”