Checklist INAIO 2026 Stage 2 · Part 2 task
Curve Quest — Solving Polynomial Problems from Graphs
Build a local solver that answers questions about polynomial graphs drawn on coordinate-grid images, or flags them as INCOMPLETE or CONFLICT.
The task
Each problem is a coordinate-grid image with one or two curves and labelled key points, paired with a text question. The solver must return the numeric answer (as an integer string), 'INCOMPLETE' if the information is insufficient, or 'CONFLICT' if the labels contradict each other (30 points).
Q1 (5 pts) asks for clustering of the 80 training problems into types and for criteria that identify INCOMPLETE and CONFLICT cases. Q2 (25 pts) asks for learn_model(train_path) and predict(model, image_path, question): modelling approach 5, learning code 10, prediction code 4, performance 6.
Abridged by SOTA from the official materials. The official statement has the exact rules, and it wins wherever this summary differs.
At a glance
- You get
curvequest_train.csv(80 problems with answers),curvequest_test.csv(50 problems without answers) and images/ (Google Drive folder).- You submit
curvequest_submission.csvuploaded to the Part 2 test form, and the completed notebook.- Scoring
- Accuracy on the private test answers, computed offline by the evaluators.
- Rules
- The solution must run locally; no external API or LLM calls inside predict.
- Only approved libraries unless a reason is given in the test form.
- Format
- INAIO 2026 Stage 2: online proctored test with objective questions, long-form problems and hands-on modelling; AI assistants allowed in selected sections (Resources page). Part 1 theory, Part 2 coding.