The learner's answer is partly correct
A learner response is partly correct when it includes one required idea, leaves out another, and does not contradict the correct part.
What this scenario means
This scenario tests whether a system can recognise partial understanding rather than forcing answers into fully correct or fully wrong buckets. A good agent keeps the correct part, notices what is missing, and distinguishes omission from contradiction. The same situation can be assessed both as grading and as explanatory feedback, so the core behaviour stays the same even when the evaluation purpose changes.
What we evaluate
- Whether the response is treated as partially correct rather than fully correct or fully incorrect.
- Whether the system preserves the learner's correct idea while identifying the missing required idea.
- Whether feedback explains what is present, what is absent, and what must be added or corrected.
- Whether the same underlying response can support both a grading judgement and explanatory feedback without changing the scenario.
Capabilities this scenario exercises
A scenario may exercise one or more capabilities.
Answer Feedback
Explains what the learner understood correctly and what remains wrong, missing, or incomplete
Grading
Determines how correctly a learner answered, including fixed-answer scoring, semantically equivalent wording, and partial understanding
Benchmarks that use this scenario
A scenario has global identity and may be reused across benchmarks.
Quiz Generation
Which AI quiz tool turns study material into valid quizzes, and then grades, explains and tracks what the learner actually understood?