My understanding so far: You take an LLM and use it without thinking (That's what openjev does?) You leave out the text generation in the end and take the confidence score in the matrix before that phase That's it. Right? They gave it a mysterious marketing name.
