Test Jev on news relevance.

Run Jev beside the current model for news relevance. Keep its answer out of the product until a separate review.

The same 174 reviewed cases were run three times. Precision and recall show the lowest and highest result.

The main test checked all six current hiring rules together on 121 reviewed cases. Every model missed at least one required accuracy check.

The test asked whether 23 article pairs belonged together. It did not build complete groups.

The test asked which article should come first in 28 pairs. It did not rank a full report.

The test turned two yes-or-no answers into a tier for 50 cases. It did not create complete article assessments.

Only four reviewed cases were available. That is enough to describe these runs, not enough to choose a model.

The test checked 27 pairs of possible themes. It did not write or review a complete daily report.

No one had reviewed and fixed the correct sentiment labels before this research. There is no fair model comparison to show.

Jev allows 64,000 tokens per request and 32,000 tokens for the saved state plus the longest question. The character check below is only a rough warning.

People reviewed the correct answers before testing. A model's own answer was never treated as correct just because the model produced it.