Benchmarking report · 01

AI Evaluation vs. Human Examiner

Initial benchmark from the AS1020 Fluid Mechanics undergraduate quiz.

Paper-by-paper comparison of AI and human examiner scores for nine Fluid Mechanics quiz papers

Summary

9 anonymised handwritten answer sheets from the AS1020 Fluid Mechanics quiz. Each paper was evaluated on a 30-mark scale and compared against the human examiner score. The AI and examiner scores had a Pearson correlation of 0.92, with an average deviation of 1.44 marks. Broader validation · Coming soon: Multi-subject benchmarking will be published as future evaluation reports are completed.

Case study · 02

Faint Handwriting Recognition Comparison

EduSathi's answer-sheet preprocessing improves faint, low-contrast handwriting before recognition. In this sample, EduSathi produced a more complete and accurate transcription than the frontier vision models shown in the comparison.