For each of the benchmark categories, do your calculations anywhere you like. Only the final numbers need to be submitted in the QBox_benchmark file.
For the other tasks, submit the files specified in the tasks.
One for each task with the version number(s) that have the problem. An example is given for the syntax of this is given in the second milestone task.
One for each task explaining the analysis as to how you went about finding the faulty versions and the numbers to compare to the benchmark version.
Some of the tasks have additional files to be submitted, like direction in which you see the issue.
Do the above manually.
The final task is to create scripts to run these checks automatically. One script for each issue.
This is the automation part.
The log file will be used to check some things but you need not do anything specific to enhance the log file.