One fine-tuned model per tool
Each of our 400+ tools has its own dedicated model, trained by an expert examiner for that qualification and question.
Accuracy claims are only worth the study behind them. Ours are measured against board standardisation materials, published per question, and corroborated by the trusts that use us.
Experienced human examiners: ~0.65 (Fowles, 2009)
Human baseline: research on experienced GCSE English examiners marking against chief-examiner scores (Fowles, 2009) found a Pearson correlation of around 0.65, with only ~45% of marks inside the board's tolerance.
Each of our 400+ tools has its own dedicated model, trained by an expert examiner for that qualification and question.
Every tool is algorithmically benchmarked against official standardisation materials before it reaches a school.
Correlation, Mean Absolute Error and % in tolerance, published per question, so you can check the tool you'll use.
Every tool is tested against board-moderated scripts before it reaches a classroom, and we publish the results. Correlation, Mean Absolute Error and % in tolerance, question by question, so you can check the exact tool you'll be using.
Read our published accuracy studies20 minutes with our Schools Partnership team: a live demo, accuracy data for your tools, and pilot options for your school or trust.
We use cookies for analytics and marketing to improve your experience. These are only set if you accept. Decline and we'll only use cookies that are strictly necessary. (Live chat is always available either way.) Learn more in our Cookie Policy.