Learn

Glossary

Account: An optional Google sign-in. It keeps the results you save to it, and the checks you start while signed in. You can open them on any device where you sign in.
AI provider: The company behind your AI, such as OpenAI, Anthropic, Google, Microsoft or xAI.
Analytics: The public page of totals and medians from Quick and Deep checks. It never shows an individual result.
Answer: The part of the AI’s reply that belongs to one question.
API: A way for programs to send questions to an AI without a chat app. Developers use it.
App: How you use the AI: a web chat, a desktop or mobile app, a coding tool or an API.
Browser key: A random value in a cookie that lets a browser open the results of checks started in it while signed out. It is made the first time you start a check or save a pack while signed out. It lasts 90 days from then and isn’t renewed.
Change: The later score minus the earlier one, in points. Shown only for two comparable checks.
Check: One run of a fixed set of questions against your AI, from copying the questions to getting a result.
Comparable: Two checks that used the same questions, the same scoring code and the same setup, with nothing set to Unknown or Other. Both must have a score, and they must have started at different times.
Custom Pack: A set of questions written by a person, with the answer they expect for each. HIBN doesn’t confirm that those answers are right.
Expected answer: The answer a pack’s author says is right. It is shown to the author and to no other visitor.
Fully correct: A question with a question score of 1. The count is shown beside the score, because a partly correct question adds to the score but not to the count.
Label: The heading that marks each answer in the reply: Q01 to Q06 for Quick, D01 to D10 for Deep, T01 onwards for a pack.
Level: The thinking or effort setting, or the mode, that your app shows. Where you choose your setup, it is called “Thinking level”.
Location: The country and city recorded with a check. It is estimated from your connection. On the home page or a pack’s page you can change it or choose not to say, and then we store only your choice. Analytics uses only the country.
Median: The middle score when a group of scores is put in order.
Missing: An answer we couldn’t find in the reply. A missing answer means no overall score, unless you tell us the AI refused or didn’t answer.
Model: The named model your app shows, when it shows one.
No score: A check with no overall score, because an answer is missing or couldn’t be scored. We don’t count that answer as zero.
Outcome: What happened with one question, such as Correct, Partly correct or Missing.
Points: The unit for a change between two scores. From 50 to 67 is a change of +17 points.
Private: On a result and in My Results: no other visitor can open it. For a pack: not published. No other visitor can open or run it.
Question score: The score for one question, from 0 to 1.
Quick and Deep: HIBN’s 2 fixed question sets. Quick has 6 questions. Deep has 10 different ones, about twice as long on average. Their scores are never combined or compared.
Refused (you told us) and Not answered (you told us): An answer was missing, and you told us the AI refused or didn’t answer. The question scores 0 and shows this outcome.
Reply: Everything the AI wrote back to the questions. We store it exactly as you paste it.
Result: The saved record of one check: the score, the outcome of each question, the reply and the setup.
Score: The average of the question scores, multiplied by 100 and rounded. Shown only when every question has a question score.
Scoring code: The fixed program on HIBN’s server that scores each answer. No AI model takes part.
Setup: The provider, app, model and level you tell us you used. HIBN can’t confirm it.
Stopwatch: An optional clock that you start and stop yourself. It includes copying and switching apps, so it isn’t the AI’s speed.
Unknown: The option to choose when you’re not sure of part of your setup. A check with Unknown or Other in its setup can’t be compared with other checks.
Version: One saved state of a pack. Saving a change makes a new version, and results from different versions can’t be compared.
You told us: A mark on anything HIBN recorded because you said so: the setup, a refusal, an AI that didn’t answer.