1,432 multiple-choice questions from Brazil's national secondary-school exam (ENEM), spanning 2009-2017 and 2022-2023, across humanities, languages, sciences and mathematics.
unassessed
| Category | knowledge |
|---|---|
| Subcategory | Brazilian national secondary-school exam multiple-choice QA (Portuguese) |
| Page status | active |
| Metric | exact_match (accuracy) |
| Direction | higher_is_better |
| Unit | % |
| Dataset size | 1432 |
| Dataset licence | Not established -- the Hugging Face dataset card sets no licence tag and its README body contains no licence statement beyond the dataset's structural (YAML) metadata. |
ENEM Challenge tests a model on real questions from the Exame Nacional do Ensino Medio (ENEM), the Brazilian government's high-school-level exam used nationwide as a university-admission gateway. Questions span four broad areas -- Humanities, Languages, Sciences (natural sciences and mathematics together, per the exam's own structure) -- and are written in Portuguese for Brazilian secondary-school graduates, so answering them draws on general academic knowledge and Portuguese reading comprehension together, not one narrow skill. The dataset compiled for this benchmark spans exam sittings from 2009 through 2023, though not every intervening year is present.
Multiple-choice question with five labelled options (A-E), single correct answer, in Portuguese; HELM's implementation drops any question the source data marks as officially annulled ("ANULADO") and scores the model's chosen option by exact match against the correct text.
No model card in ModelSpec reports this benchmark yet.