How we check tasks
The diagnostic result is a set of numbers that parents and tutors use to decide what to practise. That is why the answer key must be correct. Here is how we make sure it is.
The result comes from a key, not from AI
We compare the child's answer with the key stored with the task. Artificial intelligence does not grade answers and does not set the result. If we ever add it as a helper, it will be clearly labelled.
Where our keys come from
- Official exam papers (CKE). Tasks from official exam papers use the key from the official marking scheme. We checked part of them independently as well: we solved the tasks without seeing the key and only then compared the results.
- Generated tasks. We calculate the answer in code (using fractions and whole numbers, with no rounding), and the wrong options are typical student mistakes. Each generator has a separate test that computes the same answer by a different method.
- Tasks that need a drawing or earlier context are not used in the diagnostic, because they cannot be settled from the text alone.
What you see next to the result
Under the result we show how many tasks have a key from an official exam paper and how many have a key calculated in code. The topic map is indicative: when there are few tasks in one topic, we mark it as less certain.
Corrections register
When we find an error in a task, we correct it and list it here. A correction does not change past results, only later diagnostics.
Polish eighth-grade exam, mathematics, May 2025, task 11.2
In our copy of the task the key said "true", while the official CKE marking scheme says "false". We caught it during an independent key check.
Corrected in the source data. Applying the correction to the task bank is in progress.
Spotted an error in a task?
Write to support@sprelli.com and include the task text. We review every report. You can read more about how Sprelli works in the platform overview.