3 ms·
> key question is not just force balance: it’s whether the inextensible This is definitely not astra, taking a guess this is 5.6, perhaps not even Sol, which d
by RomanKornev 20d ago
> key question is not just force balance: it’s whether the inextensible
This is definitely not astra, taking a guess this is 5.6, perhaps not even Sol, which does not reflect the state of the frontier (what the research was about).
And yes, the paper is already outdated
- amluto 20d ago> And yes, the paper is already outdated The paper is about the fact that the benchmark’s evaluator is prone to egregious incorrect rejections of what answers that it should accept. A new model will not invalidate that issue.
- emil-lp 19d agoThis is getting ridiculous. Is 5.6 not considered good enough to conduct experiments with anymore? Two months ago, if you weren't using 5.6, you were doing it wrong. Now, you can't conduct experiments using the default model?
- RomanKornev 19d agoWhy is it so surprising that frontier can move that fast?