JevMade Sign in
← Back to experiments

Benchmarks & research

How sure is Jev?

Turns Jev probability distributions into one plain-language sureness verdict, from certain through torn to clueless.

Source screenshot of How sure is Jev?
SOURCE SCREENSHOTFull screenshot ↗

What it does

The zero-dependency library combines six uncertainty measures. In the maker's 72-question study, Jev's Choice confidence closely tracked the winning probability.

How you can use it

When an AI model gives percentage chances for several choices, your developer can use this code to turn those raw numbers into a plain verdict. It combines multiple math formulas to label an answer as certain, confident, leaning, torn, or clueless.

Your system can act on clear picks immediately. When choices are too close or the model is confused, the system can stop and ask a person instead. A high rating only means the numbers strongly favor one option, not that the choice is correct.

Maker-reported (not independently measured by JevMade): 72 real questions asked to compare Jev's confidence field against six metrics (author-reported) · On choice answers Jev's confidence equalled max_prob on every answer, largest gap 0.02 (author-reported) · A two-option 75/25 answer scores 0.50 under Jev's max_prob but 0.19 under entropy (author-reported)

Primitives
choice, score
Platform
Python
Added
Project created
GitHub stars
0 (snapshot, not live)

Keep this for later

Sign in to bookmark experiments, guides and videos, and keep notes only you can see.

Continue to sign in

We’ll bring you back to this listing.