Extended Response Question generator · IB DP Physics 2025 syllabus

Authentic IB-grade Paper 2 questions, generated to spec.

An Extended Response Question (ERQ) is the multi-part 10–16 mark problem at the heart of IB Physics Paper 2 — a novel real-world stem, parts (a)/(b)/(c)/(d) walking a strict cognitive ladder, and a discriminator that asks the student to critique the model. This site generates ERQs that hit the IBO bar, anchored on 26 tagged past-paper exemplars and a critic ensemble that enforces every rule of the contract.

What's an ERQ?

IB Physics Paper 2 contains a handful of ERQs worth 10–16 marks each. Unlike a Paper 1 multiple-choice, an ERQ has a multi-sentence stem (a real-world scenario with data), followed by parts labelled (a), (b), (c), (d). Each part has a strict mark allocation and a specific command term that signals the assessment objective:

Authentic ERQs always blend two sub-topics from different themes (e.g., kinematics + radioactive decay) — that cross-pollination is what separates a real IB question from a textbook drill. Generators that miss this constraint feel "off" to teachers. We enforce it as a hard rule.

Cross-theme integration

Every question blends two sub-topics from different themes (A/B/C/D/E), because real IB Paper 2 ERQs always do.

Cognitive ladder + Rule of One

Part (a) eases the student in. Middle parts demand multi-step calculation. The final part critiques the model. Every mark equals one distinct operation.

ECF-aware mark schemes

Mark schemes tag each method-step independently. A wrong upstream value never dooms a downstream method mark.

Anchored on the corpus

Every generation is grounded on the nearest past-paper exemplars from the IBO archive — same vocabulary, same register, same examiner expectations.

How it works

  1. Planner picks the stem archetype + cognitive ladder + cross-pollination point.
  2. Author writes the question + mark scheme, grounded on 2 nearest exemplars from the tagged corpus.
  3. Critic ensemble (5 specialists) scores topic accuracy, AO ladder, Rule-of-One, cross-theme, discriminator. Each returns pass/fail + reasoning.
  4. Reviser re-runs the author with critic feedback. Up to 2 retries before bubbling up to human review.
  5. You grade the output. Verdicts update the rubric, which sharpens the next generation.

A typical run takes 60–90 seconds. Watch live progress when you generate one.