BLUEX (Brazilian Leading Universities Entrance eXams)

Portuguese multiple-choice questions from Unicamp (Convest) and USP (Fuvest) entrance exams, including image-linked items that most text-only harnesses drop.

Also known as: Brazilian Leading Universities Entrance eXams

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.
Categoryknowledge
SubcategoryBrazilian university-entrance multiple-choice exams (Portuguese)
Page statusactive
Metricexact_match (HELM); accuracy in the original paper
Directionhigher_is_better
Unit%
Dataset size1422
PublisherUniversity of Campinas (UNICAMP) / Portuguese-Benchmark-Datasets

What it measures

BLUEX tests whether a model can answer official Brazilian vestibular questions written in Portuguese. Items come from Unicamp’s Convest and USP’s Fuvest papers. Subjects span high-school fields such as mathematics, history, and languages. Many questions need an accompanying figure. Metadata flags image use, mathematical reasoning, and Brazil-specific knowledge. HELM’s scenario is text-only: it skips any item marked has_associated_images.

Task format

Multiple-choice. HELM uses joint letter answering with a Portuguese instruction and one worked example, scored by exact match. The 2023 paper also used few-shot prompts drawn from a different year of the same university.

Models reporting this benchmark

No model card in ModelSpec reports this benchmark yet.

Data

This page as JSON · Edit on GitHub