How Edvana Generates and Checks IB Practice

How we generate papers

Edvana does not ask a general chatbot to invent an exam from a blank prompt. A paper starts with structured choices such as subject, level, paper type, mark total, time limit, calculator setting, and syllabus outcomes.

The system then builds questions inside those boundaries. It also checks the surrounding paper context so questions do not keep repeating the same setup, and it retries generation when a question fails review.

How we mark

Multiple-choice answers are marked directly against the answer key. Longer written or calculation answers are marked with the original question, mark scheme, and student response in view, then cleaned before the feedback is shown.

How we differ from ChatGPT

ChatGPT is a flexible conversation tool. Edvana is a constrained exam-practice workflow. It uses known paper settings, selected syllabus outcomes, structured responses, review checks, retry loops, and formatting cleanup before content is used in the product.

Examiner criteria alignment

Edvana is not endorsed by the IB. The alignment comes from practical checks: outcomes must match the selected syllabus data, question parts must add up to the declared marks, calculator rules must match the paper, and command words are checked against the number of marks available.

Our review process

Before a generated paper is accepted, Edvana checks for common failure modes: wrong mark totals, invalid topic IDs, calculator mismatches, missing parts, placeholder text, data references without data, repeated math expressions, command-word mismatches, broken math notation, and unsafe diagram content.