I have not seen many work on explainable AI regarding large language models. I remember many very nice visualizations and visual analysis tools trying to comprehend, what the network „is seeing“ (eg. in the realm of image classification) or doing