Smart Question Generator and Evaluation System

Problem

Manual question design often suffers from weak alignment with curricula or exam standards, distractors that do not reflect real misconceptions, and difficulty or coverage estimates based mostly on intuition. This skill structures the workflow so Bloom levels, difficulty, knowledge points, item types, and point values are constrained during generation rather than reviewed after the fact.

How It Works

After receiving subject, knowledge scope, item types, counts, and difficulty ratios, it breaks the domain into weighted knowledge points and generates multiple-choice, true/false, fill-in-the-blank, short-answer, essay, case-analysis, or programming items. Each item should include a Bloom level, estimated difficulty, targeted knowledge point, and score. Multiple-choice distractors must come from common errors, and explanations should include reasoning steps and likely mistakes. Paper assembly checks difficulty distribution, item independence, single-answer uniqueness, and high-weight knowledge coverage, with an 80% minimum target for high-weight points.

Boundaries and Notes

It is not for grading, learning analytics, curriculum design, or simple knowledge lookup. The workflow does not allow automatic web search for real-time data; scenario-based data should preferably use sources from the last two years or be dated explicitly in the stem. When questions must be based on a specific textbook, the skill should first Read the provided source before generating items.