News for hackers. Stuff that matters.
Re:Can LLMs write better code if you keep asking them to "wr... by minitech on Friday January 03, 2025 @04:47PM (#42589814)
They’re proposing choosing the output values randomly according to the distribution obtained by choosing input values uniformly at random for the original algorithm.
Re:Can LLMs write better code if you keep asking them to "wr... by ryao on Friday January 03, 2025 @07:52PM (#42591159)
That removes the random element to this. The way that random numbers work is that it is possible (although unlikely) that the minimum and maximal values in the range will not be selected when generating the million random numbers. If you assume that they will always be selected and thus always return the same output, then your output will be wrong at least some of the time.
Re:Can LLMs write better code if you keep asking them to "wr... by minitech on Saturday January 04, 2025 @06:41PM (#42598502)
I don’t know how you got “always return the same output” from “choose the output randomly according to [a non-uniform] distribution”.