Quoted By:
Wizard seems to be pretty good at differentiating between characters with cards containing multiple character description. I haven't gotten any mixup of each character characteristic or description in any generation. Seem to be almost on par or slightly bellow GPT-4 Turbo in terms of understanding instruction and context. Though it suffers the same problem of not recognizing "Do Not" instruction like GPT 3.5.