In WEKA GUI go to Explorer, open your ARFF file and then go to Classify-->More options-->Output predictions-->Choose. There choose a format to see the classifications for your test set. Try first with PlainText, WEKA will output the predictions for each of the test instances. In the column "error" WEKA indicates with a + if an error is made in the classification.
I able to find the incorrect instances using this technique if there is no cross validation . But if we use k CV fold, then how can we find the exact incorrect instances since it is not showing all the records detail, only serial number starting from 1 for each fold is showing? Please help me to resolve this.