[qudat]

So if they are using a pretrained model and the second llm scores all responses below the ok threshold what happens?