[shubhamintech]

to me i solely evaluate on the way they write prompts and how elaborate they are in terms of explaining it and the patience when the model messed up