The generally used term that I'm familiar with for this is "referential transparency", that given the same inputs you'll get the same outputs every time. LLMs c

The generally used term that I'm familiar with for this is "referential transparency", that given the same inputs you'll get the same outputs every time. LLMs can be deterministic (referentially transparent) but almost none are, i.e. when they are given the exact same input they do not return the exact same output.