"Congrats to the champion for holding us off this time," OpenAI wrote after its model came second at the AtCoder Heuristics ...

That study found that when asked to choose random numbers between one and five, the LLMS would choose three or four. For between one and 10, most would choose five and seven, and between one and 100, ...