3 ms·

I wouldn’t look for questions with yes/no answers, but for questions where the answers can have correct/incorrect reasoning. Of course, you can’t turn those int

by layer8 2y ago

I wouldn’t look for questions with yes/no answers, but for questions where the answers can have correct/incorrect reasoning. Of course, you can’t turn those into automated benchmarks, but that’s maybe kinda the point.

varjag 2y ago
I think that's the point: correctness can be a sliding scale. There is Newton correct, and there's Einstein correct.
- rightbyte 2y ago
  It wouldn't surprise me if Newton correct is closer to correct for 'small numbers' then Einstein correct will be in the end in the general case.