3 ms·
That "blind ranking" is limited to about 2,000 tokens of context. So it's certainly not evaluating how good the models are at complex assignments.
by staticman2 3y ago
That "blind ranking" is limited to about 2,000 tokens of context. So it's certainly not evaluating how good the models are at complex assignments.