What it does
Its example cutoff comes from a few checked video titles, not a rule for other data. Three unchecked titles above it may be mistakes; the checked no-side disagreement yields no usable cutoff.
Benchmarks & research
jev-trust-cut
Audit where Jev disagrees with existing yes-or-no rules, then use human checks to explore separate thresholds for each answer.
Bookmark: jev-trust-cut
Keep this in your collection.
Leave a noteWhat would you try with this?
: jev-trust-cut
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
How you can use it
Start with decisions your project already made and let Jev answer the same yes-or-no question. Send confident disagreements to a person instead of changing the live rules straight away. The report groups disagreements and prepares a sheet where a reviewer can enter the real answer.
Use your own checked examples to decide how sure Jev must be before acting on a yes or a no. Check more than the easy cases. EatList checked video titles, not the videos; do not copy its 0.73 starting point. If the strongest checked disagreement is wrong, keep the old decision for that answer.