Evaluation · bring your own key

Generation and question quality

Eight fixed topics, the same retrieved passages for everyone, and your chosen model. Every generated question is checked for citations, grounded numbers, answerability, distractors and duplicates, and you rate it on a short rubric. Nothing here is published; runs and ratings stay in your browser. Retrieval results are on Evaluation.

Run setup

No key set. Add one to run live, or try the practice run.

Topics (8 of 8)Sampling methods: simple random, stratified, cluster and systematic samples 1.2Levels of measurement: nominal, ordinal, interval and ratio data 1.3Experimental design: random assignment, control groups, placebos and blinding 1.4Box plots, quartiles and the five-number summary 2.4Mean, median and mode, and how skewness affects them 2.5Standard deviation, variance and z-scores 2.7Least-squares regression, residuals and the correlation coefficient 12.2Outliers and influential points in linear regression 12.5

Task: Likely student questions
Passages per topic: 5, dense, precomputed
Temperature: 0 where the model allows
Bootstrap seed: 20241009
Estimated cost: –

Calls go from your browser to api.anthropic.com and are listed in the AI log. Runs and ratings stay in this browser.

Generation evaluation

Same topics, same passages, your model.

Each selected topic retrieves the same five passages for everyone, and your model writes four multiple-choice and three short-answer questions from them. Every item is checked automatically, and you rate it on a short rubric. The summary reports both with intervals, and how often the checks agree with you.