Revised comparison of modal theorem provers

The following pictures show the performance of the Logics Workbench, KRIS*, KsatLISP, the translation approach using SPASS, and FaCT on the parameter setting PS13 (N=6, M=1, K=3, D=1, P=0).
Percentile graphs for LWB* on PS13
Percentile graphs for KRIS* on PS13 Percentile graphs for KsatLISP on PS13
Percentile graphs for TA* on PS13 Percentile graphs for FaCT on PS13
Again we can improve the performance of some procedures by allowing subformula reordering on the input formulae. The following pricture shows the performance of KsatLISP with sorting on the parameter setting PS13.

Percentile graphs for KsatLISP on PS13
It is interesting to compare the percentile graphs for KsatLISP without and with sorting on (N=6, M=1, K=3, D=1, P=0) with the median CPU time graphs persented in Giunchiglia, Giunchiglia, Sebastiani, and Tachella (1998). They consider only the parameter setting (N=4, M=1, K=3, D=1, P=0) where a dramatic improvement can be seen after enabling sorting. However, for sets of random formulae generated with N>4 the improvement with sorting is less dramatic.


Top: Contents | Previous: Guidelines for generating more challenging problems | Next: Semantic versus syntactic branching