The first column of the REF files represents the time axis at 10ms steps, and the second column is the f0 in Hz. The corresponding .wav files are as described on the melody extraction wiki (44.1 kHz, MONO, etc).
Please note that the test set to be used in the evaluation, which will not include these files, will contain some segments in which the dominant melody is represented by a musical instrument other than voice. Additional example files from the 2004 test set may be found courtesy of the UPF MTG group here. However, please note that a different grid size was used for the 2004 competition.