Built to the skills the test measures — and checked before your child sees it.
Korva prepares Year 6 students for the ACER ASET, the exam used for selective entry into Western Australian public schools. Below is what we align to, how questions are made, and — just as important — what we check before one is allowed to reach a child.
What we align to
The exam tests four things, and Korva practises the same four. We align to the skills — we do not reproduce real test questions, and we could not: every item your child sees is generated for them.
Reading comprehension
Full-length passages at genuine exam word counts. Questions turn on inference, tone, authorial choice and the function of a paragraph — not on finding a fact and copying it back out.
Quantitative reasoning
Multi-step reasoning over ratio, rate, pattern and deduction, often built on a shared stimulus — a table, a diagram, a price list — the way the real paper does it. Not single-operation arithmetic.
Abstract reasoning
Entirely visual, as on the real paper. Every figure is drawn by our own rendering engine and shown as a picture. A shape described in words would be a different task, so we never do it.
Written expression
One extended piece from an abstract statement, in whatever form your child chooses, marked against the three criteria the exam scores: ideas, organisation and expression.
How a question is made
Generated to a specification
Each section has a written specification describing the passage types, question shapes and option vocabulary the exam uses. A question is generated against that specification, at one of four difficulty tiers, and at Year 6 level — there is no easier year band hiding underneath.
Deliberately varied
Before generating, the format is pre-picked from a weighted set, so the same easy shape cannot keep winning. A child who has only met one question shape has practised recognition, not reasoning.
Checked, not trusted
Generated content is checked before it is stored. Cheap structural checks run first — length, coherence, whether the options are actually distinct — and only the survivors go on to be judged on substance.
Explained, then re-asked
Every answer comes back with the reasoning, including why the tempting wrong option was tempting. Anything your child got wrong returns the next day automatically, and their skill map updates after every single attempt.
What we check before it ships
Abstract reasoning is solved independently before it is served
The most damaging thing a generator can do is mark the wrong option correct and write a confident justification for it. So an abstract puzzle is not accepted on the generator’s word. A second, stronger model is handed a plain-English description of the figure and asked to solve it. If its answer disagrees with the one marked correct, the question is thrown away and regenerated. It gets three attempts, and then it is dropped.
The hard tier has to earn the label
“Hard” is a claim about reasoning depth — three or more independent moves, with the route to the answer hidden rather than signposted — not about how long a question looks. Every item claiming that tier is scored against a written difficulty rubric by an independent judge, and the judge is fail-closed: an error, a missing key or a reply we cannot read all mean the same thing, which is that the question does not ship.
We do not grade our own reading content against a paper we do not have
Our difficulty calibration is not equally strong in every section, and we would rather say so than round it up. Abstract and quantitative difficulty are calibrated against past-paper material we hold. Reading is not. It is judged against a written criterion only — we hold no reading paper, so any claim that reading was measured against one would be a claim we cannot support. That is also why a question labelled hard by structure alone is re-checked on substance before it is put in front of a child.
Written expression is marked, not scored
A piece of writing has no single correct answer, so Korva never gives writing a percentage score — not on your child’s home page, not on your dashboard, and not inside the test-ready score. It reports what was written and the feedback against each of the three criteria. A percentage there would be a number we invented, and an invented number is worse than no number.
Practising like the real day
Four sections, on the clock
A full paper each week at the real per-question pacing, with results and a section-by-section review afterwards.
Reviewed before release
A harder set each week, released only after a human has approved that week’s content. Some weeks it opens later than others, and the app says so rather than showing an empty page.
See a question before you decide.
14 days free, no card. $7 a week after that.