Verified against ChatGPT · 2026-08-13
Get your wrong exam answer explained the way a real grader would mark it, including where you'd have earned partial credit
Reviews your actual attempted answer against the correct one, pinpointing exactly where your reasoning diverged and how many marks that specific error would likely cost under real exam marking, rather than just restating the right answer.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Review my actual attempted answer to this question and explain precisely where it went wrong, the way an exam grader marking against a scheme would, not just by restating the correct solution. QUESTION A car accelerates uniformly from 5 m/s to 20 m/s over 6 seconds. Calculate the distance travelled during this time. MY ATTEMPTED ANSWER a = (20-5)/6 = 2.5 m/s^2. s = ut + 0.5at^2 = 5(6) + 0.5(2.5)(6) = 30 + 7.5 = 37.5 m CORRECT ANSWER OR MARK SCHEME (IF I HAVE IT) Official answer: 75 m, method mark for correct SUVAT equation, mark for correct substitution, mark for final answer EXAM MARKING STYLE GCSE Physics, method marks awarded for correct equation and substitution even if final arithmetic is wrong RULES First identify exactly where my reasoning departs from a correct path — quote or paraphrase the specific line or step in my answer where the error starts, not just "the final answer is wrong." Distinguish between an error in method (I used the wrong approach entirely) and an error in execution (I used the right approach but made an arithmetic, sign, or transcription slip partway through), since these are graded very differently on most real mark schemes. Estimate, based on the marking style described, roughly how many of the available marks my attempt would likely have earned given where the error occurred — method marks for correct steps before the error, versus losing marks only from the error onward. If I did not provide an official mark scheme, say explicitly that your mark estimate is an approximation based on typical marking conventions for this exam type, not an authoritative score. Do not simply present the fully correct solution as your primary response — the correction should be secondary to, and clearly derived from, the diagnosis of my specific error. OUTPUT FORMAT 1. Where the error starts (quote the specific step). 2. Method error or execution error, and why. 3. Estimated partial credit under the marking style given, with the caveat if no official scheme was provided. 4. The corrected path forward from the error point only, not a full restated solution from scratch.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
When simply asked to check an answer, GPT-5.1 tends to solve the problem fresh and present the correct solution as the primary output, treating the student's actual attempt as an afterthought to compare against at the end — this is the opposite of what's diagnostically useful, since a student reviewing a wrong answer already knows what's ultimately correct in outline and specifically needs to know where their own reasoning diverged. Requiring the model to quote the specific step where the error starts forces it to actually trace through the given attempt line by line rather than pattern-matching "this final number doesn't match" and working backward from the correct answer, which is a meaningfully different and more careful diagnostic pass. The method-versus-execution distinction is mechanically important because it maps directly onto how real mark schemes allocate partial credit — in the worked example above, the student used the exact right SUVAT equation but substituted t=6 twice instead of squaring it (0.5(2.5)(6) instead of 0.5(2.5)(6^2)), an execution slip that a real grader would likely still award method marks for, whereas GPT-5.1's default "the answer is wrong, here's the right one" response would obscure that the underlying approach was actually sound. The explicit caveat requirement when no official mark scheme is supplied matters because a partial-credit estimate stated with unwarranted confidence could mislead a student about how lenient or strict their real exam's grading actually is, when the model is in fact guessing at typical conventions rather than citing a scheme it was actually given.
What you get back
1. Error starts at: '0.5(2.5)(6)' — the time value was substituted as t instead of t-squared. 2. This is an execution error, not a method error: the correct equation (s = ut + 0.5at^2) and the correct acceleration value were both used correctly up to that point. 3. Under GCSE-style method marking, this would likely earn 2 of 3 marks — one for the correct equation choice, one for correct substitution of u and a, with the final mark lost for the squaring slip. 4. Corrected from that point: 0.5(2.5)(6^2) = 0.5(2.5)(36) = 45, giving s = 30 + 45 = 75 m.
Verified against
ChatGPT GPT-5.1 · 2026-08-13
Changelog
- 2026-08-13 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
