In my experience it’s generally accurate when summarizing content provided in the prompt context. Where it can run into trouble is “recalling” (if you can call

In my experience it’s generally accurate when summarizing content provided in the prompt context. Where it can run into trouble is “recalling” (if you can call it that) content that it was trained on.