promptdojo_

The ML-system option — six boring stages that get a repo taken seriously — step 3 of 7

A pipeline run just finished. Training accuracy is a glowing 0.98. Held-out accuracy is 0.78. The bar is 0.80.

Predict all three output lines — the row count after validation, the accuracy line, and the gate's verdict. One of the four numbers in run never gets read at all. Which one?