Model Validation means feeding a model an input and checking how well the model output conforms to “reality”. In this case that could be as simple as feeding WLv2 a lot of custom workouts and seeing how it score those workouts compared to human scoring. Or as complex as having folks do “free form” workouts and having WLv2 score those and have humans “validate” the scoring.
The problem in this case is the the further from a planned workout the test workouts get (think a group ride) the harder it is to come up with the “reality” score that you are comparing WLv2’s score against