← All experiments

No. 030 · Week 41 · Michael Taylor · Exploring

Tier-list grades for checks

The same 200 outputs judged twice, once with yes/no checks and once with S-to-F tier grades.

Does a tier-list grade tell you more about a model than pass or fail?

See the rest of the write-up and 1 more image.A free Every account opens this. Nothing to buy.