I read the exact answer distribution from the logits of open models, base and instruction-tuned, and turned it into bits and guesses-to-crack.

The post How Many Guesses Does It Take to Crack an AI's "Random" Number? appeared first on Towards Data Science.