I will present cost minimizing search technique, which is a search-based QBF
solving technique for general format QBFs. Unlike the most QBF solvers, it does
not require its input format to be prenex and CNF. The search engine calculates
i) truth probability, ii) expected search space, and iii) cost for each node and
explores the search space in the 'minimal-cost-first' manner. I compare CMS with
a baseline solver that does brute-force search directly, and with QuBE, sKizzo,
and Quantor, the three state-of-the-art QBF solvers. Experiments show that CMS
is 1-4 orders of magnitude faster than the baseline solver, and it is 1-2 orders
of magnitude faster than the other three solvers in one-third of the free-format
instances available in QBF 1.0 test suite from the QBFLIB.