Yeah, that’s a pretty safe approach. You never know which parts are crude averages of the training data and which parts got interpolated.
If you happen have subject matter expertise, you can tell where it all went wrong. On the other hand, at that point you already know what the right answer is so why bother asking an LLM in the first place. It’s a very paradoxical situation.
Yeah, that’s a pretty safe approach. You never know which parts are crude averages of the training data and which parts got interpolated.
If you happen have subject matter expertise, you can tell where it all went wrong. On the other hand, at that point you already know what the right answer is so why bother asking an LLM in the first place. It’s a very paradoxical situation.