Just pushed out a new release of the multi-search Weka package, since I received a very useful pull request from Jan van Rijn. The change includes recording the various setups that have been evaluated during a run. This makes it much easier analyzing the setups than parsing the (optional) log file.