---
configs:
- config_name: alba_mcq
  data_files:
  - split: validation
    path: data/alba_mcq/validation-*.parquet
- config_name: cultura_viva_pt_mcq
  data_files:
  - split: validation
    path: data/cultura_viva_pt_mcq/validation-*.parquet
- config_name: openbookqa_mt_pt
  data_files:
  - split: validation
    path: data/openbookqa_mt_pt/validation-*.parquet
- config_name: portugal_basic_qa
  data_files:
  - split: validation
    path: data/portugal_basic_qa/validation-*.parquet
- config_name: pt_exams_history_geography
  data_files:
  - split: validation
    path: data/pt_exams_history_geography/validation-*.parquet
- config_name: saudade_pt
  data_files:
  - split: validation
    path: data/saudade_pt/validation-*.parquet
- config_name: sst2_pt_mini
  data_files:
  - split: validation
    path: data/sst2_pt_mini/validation-*.parquet
---

# PTCORE

Standardized European Portuguese CORE-style multiple-choice evaluation dataset.

Each config is one PTCORE task and exposes a `validation` split.

## Schema

- `id`: stable row id
- `task`: task/subtask name
- `source`: original Hugging Face dataset
- `source_config`: original config/subset
- `source_split`: original split
- `question`: prompt/question text
- `choices`: answer options
- `correct_choice`: zero-based index into `choices`
- `metadata`: source-specific JSON string

## Counts

| Config / task | Rows |
|---|---:|
| `alba_mcq` | 240 |
| `cultura_viva_pt_mcq` | 1000 |
| `openbookqa_mt_pt` | 500 |
| `portugal_basic_qa` | 50 |
| `pt_exams_history_geography` | 544 |
| `saudade_pt` | 8573 |
| `sst2_pt_mini` | 2048 |