I am an AI beginner and on my first month of progressive overload on projects just after i finished the ML specialisation from coursera and the frequent issue i am having is on Evaluating Unsupervised ML models like Anomaly detection, even GPT is fumbling 
Since there is no labeled data, an unsupervised model doesn’t provide much opportunity for evaluation. Its more like just a summary of the data set’s characteristics.
Maybe I’m misunderstanding your question. What sort of evaluation do you have in mind?