I have no idea what you're talking about because they still screw up in the exact same way as gpt3.
The hallucination quantity and severity is way less in new frontier models.
But not more predictable or regular.
The hallucination quantity and severity is way less in new frontier models.
But not more predictable or regular.