“patch that is like the golden answer set. So even if the…” — The Pragmatic Engineer at 55:18
“patch that is like the golden answer set. So even if the model didn't write the exact code that the benchmark was expecting, was it functionally equivalent? And the next one is like a code quality review from another judge model. And like, that's better, but it's not sufficient. And this is why”
0:00
1:32:24