The speaker interprets the representation-based results as direct evidence that the model knows it should not deceive the grader.

“using this we can like track yeah this sort of is is direct evidence that the model is aware that it shouldn't be doing this um and then when you run the when you run these vectors as sort of you get the cosine similarity between the vector and the representation over a big web corpus you know I think we use fine web um and the examples that it highlights in most in like for these vectors are just fascinating yeah they're examples of like”