Datasets:

maritaca-ai
/

poscomp

dataset_info:
  features:
    - name: question
      dtype: string
    - name: number
      dtype: int64
    - name: id
      dtype: string
    - name: alternatives
      sequence: string
    - name: associated_images
      sequence: binary
    - name: answer
      dtype: string
    - name: has_associated_images
      dtype: bool
    - name: alternatives_type
      dtype: string
    - name: subject
      sequence: string
    - name: IU
      dtype: bool
    - name: MR
      dtype: bool
    - name: CR
      dtype: bool
  splits:
    - name: train
      num_bytes: 783159
      num_examples: 140
  download_size: 505636
  dataset_size: 783159
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: cc-by-4.0
task_categories:
  - question-answering
  - multiple-choice
language:
  - pt
tags:
  - poscomp
  - portuguese
  - evaluation
  - benchmark
  - computer-science
size_categories:
  - n<1K

POSCOMP

Questions from the POSCOMP exams of 2022 and 2023. POSCOMP (Exame Nacional para Ingresso na Pós-Graduação em Computação) is the Brazilian national exam for admission to graduate programs in Computing, administered by the Sociedade Brasileira de Computação (SBC). The dataset contains 140 multiple-choice questions in Portuguese covering mathematics, computer fundamentals, and computer technology.

This dataset was released as part of PoETa v2.

Dataset Structure

A single train split with 140 rows. Each example has the following fields:

Field	Type	Description
`question`	string	The question text.
`number`	int64	Question number within its exam.
`id`	string	Unique identifier (e.g. `POSCOMP_2023_56`).
`alternatives`	list[string]	The multiple-choice options (A–E).
`associated_images`	list	Images associated with the question, if any.
`answer`	string	The correct alternative (letter).
`has_associated_images`	bool	Whether the question has associated images.
`alternatives_type`	string	Type of the alternatives.
`subject`	list[string]	Subject area(s): `mathematics`, `computer_fundamentals`, `computer_technology`.
`IU`	bool	Requires image understanding.
`MR`	bool	Requires mathematical reasoning.
`CR`	bool	Requires complex reasoning.

Usage

from datasets import load_dataset

ds = load_dataset("maritaca-ai/poscomp", split="train")
print(ds[0])

Citation

If you use this dataset, please cite:

@article{almeida2025poeta,
  title={PoETa v2: Toward More Robust Evaluation of Large Language Models in Portuguese},
  author={Almeida, Thales Rog{\'e}rio Sales and Pires, Ramon and Abonizio, Hugo and Nogueira, Rodrigo and Pedrini, Helio},
  journal={IEEE Access},
  volume={13},
  pages={214180--214200},
  year={2025},
  publisher={IEEE}
}