CIMCQA

A HELM scenario for CS-education concept-inventory multiple-choice questions that cannot be run outside its authors because the item data is private.

unassessed

This page is a discovery lead. Nobody has yet assessed it against the catalogue contract, so it carries no disposition. Absence of evidence here is not evidence of staleness.
Categorydomain
SubcategoryCS education concept-inventory multiple-choice questions
Page statusunknown
Metricexact match (single-letter multiple-choice accuracy)
Directionhigher_is_better
Unit%

What it measures

ci_mcqa is the internal name HELM (Stanford's Holistic Evaluation of Language Models) uses for a scenario class called CIMCQAScenario, registered as the run spec "ci_mcqa". The scenario's own docstring describes it as "a multiple-choice question answering (MCQA) dataset designed to study concept inventories in CS Education." A concept inventory is a standardized instrument used in education research to detect specific, well-documented misconceptions a learner holds about a subject; the format originated in physics education and has since been adapted to other fields, including, here, computer science. Each item is a short question with several lettered answer options and exactly one option marked correct in the source data.

Task format

Multiple-choice question answering; HELM's adapter instructs the model to answer with a single letter chosen from the options given. The scenario code reads a separate test split and a (normally disabled) few-shot training split from local files.

Models reporting this benchmark

No model card in ModelSpec reports this benchmark yet.

Data

This page as JSON · Edit on GitHub