I'd generally be quite surprised to see LLMs spam unsafe blocks, both because that's behavior that I haven't observed while using them and because that contradicts my mental model of them where they imitate the styles of code that they were trained on (which in rust generally does not include spamming unsafe).
I'd generally be quite surprised to see LLMs spam unsafe blocks, both because that's behavior that I haven't observed while using them and because that contradicts my mental model of them where they imitate the styles of code that they were trained on (which in rust generally does not include spamming unsafe).