Why ¾ and 0.75 are the same answer — and CO and Co are not

A student writes 3/4. Your answer key says 0.75. Every teacher knows those are the same answer — and every grading tool that compares text would mark it wrong. SuperGrader compares answers as what they mean, not how they were typed or written, and that one decision shapes everything about how it grades math, science and code.

Numbers are values

Before any comparison, both the key and the student answer are normalised. Fractions become decimals, so 3/4, 0.75 and 6/8 all land on the same value. Scientific notation is evaluated, so 3×10⁸, 3e8 and 3*10^8 are one number. Percent signs, currency symbols and thousands separators are treated as presentation. A tolerance on the key — 9.8 ± 0.2 — becomes a range, and anything inside it is correct.

Expressions are evaluated too: 2π and 6.283… match, √16 and 4 match. Comparison operators fold together, so x ≤ 5, x <= 5 and x =< 5 are the same inequality.

Units have to be right as well

In physics the value is only half the answer. 9.8 m/s² and 9.8 m/s^2 are the same; 9.8 m/s² and 9.8 m/s are not. Unit spellings are folded — µm and um, Ω and ohm, N·m and Nm — but the unit itself is checked alongside the number. One place where nothing is folded: mL and ml are different units, because in a science classroom the case of a unit is meaning, not style.

Where case is the whole answer

That principle runs through chemistry and code. H₂O and H2O are the same molecule, and 2H2+O2->2H2O is the same equation as 2 H₂ + O₂ → 2 H₂O. But CO is carbon monoxide and Co is cobalt, and the grader keeps them apart. In a program, whitespace is noise but identifier case is not: myVar and myvar are two different variables, and x <= 5 is never the same as x < 5. In biology, genotype case is the answer — AaBb and BbAa match because only the ordering is forgiven, while AA, Aa and aa are three different answers.

Rules, not opinions

None of this is an AI deciding what is "close enough". The AI reads the handwriting; a deterministic engine normalises and compares. The same paper gets the same score every time, and you can see exactly why. When a read is genuinely uncertain — an ambiguous digit, two answers written and neither crossed out — the question is flagged for you with the reason, instead of being silently scored.

Missing a symbol your subject uses? There is a request form inside the app, and requests go straight to the people who write the rules.