---
license: apache-2.0
task_categories:
- question-answering
- multiple-choice
language:
- lg  # Luganda
- nyn # Nyankore
pretty_name: Pedagogy Benchmark - Multilingual
size_categories:
- 1K<n<10K
tags:
- education
- pedagogy
- multilingual
- african-languages
- teacher-training
configs:
- config_name: luganda
  data_files:
  - split: cdpk_main
    path: luganda/cdpk_main-*.parquet
  - split: cdpk_send
    path: luganda/cdpk_send-*.parquet
- config_name: nyankore
  data_files:
  - split: cdpk_main
    path: nyankore/cdpk_main-*.parquet
  - split: cdpk_send
    path: nyankore/cdpk_send-*.parquet
---

# Pedagogy Benchmark - Multilingual

A multilingual translation of the [AI-for-Education/pedagogy-benchmark](https://huggingface.co/datasets/AI-for-Education/pedagogy-benchmark) dataset into African languages.

## Dataset Description

This dataset provides translations of Chilean teacher training exam questions into African languages. The original dataset contains multiple-choice questions covering various pedagogical domains, education levels, and subject areas.

### Languages

- **Luganda (Uganda)**
- **Nyankore (Uganda)**

**Coming soon:** Swahili

### Translation Methodology

All translations were performed using **Google's Gemini 2.0 Flash** model with:
- Specialized system prompts for educational content
- Temperature 0.3 for consistent translations
- Human validation on sample sets
- Preservation of all metadata and structure

## Dataset Structure

### Configs and Splits

Each language has two splits mirroring the original dataset:

- **`cdpk_main`**: Main pedagogy questions (920 questions)
- **`cdpk_send`**: Special Educational Needs and Disabilities (SEND) questions (223 questions)

### Data Fields

- `question_id` (int): Unique identifier
- `question` (string): Multiple-choice question text
- `answer_a`, `answer_b`, `answer_c`, `answer_d` (string): Answer options
- `answer_e`, `answer_f`, `answer_g` (string, nullable): Additional answer options
- `correct_answer` (string): Correct answer letter (A-G)
- `category` (string): Subject category (e.g., Science, SEND, General)
- `pedagogical_subdomain` (string): Pedagogical area (e.g., Assessment, Teaching strategies)
- `age_group` (string): Educational level (Pre-primary, Primary, Secondary)
- `year` (int): Exam year
- `secondary_category` (string, nullable): Additional categorization for SEND questions

### Data Splits

| Config | Language | cdpk_main | cdpk_send | Total |
|--------|----------|-----------|-----------|-------|
| luganda | Luganda | 920 | 223 | 1,143 |
| nyankore | Nyankore | 920 | 223 | 1,143 |

## Usage

### Load a specific language and split

```python
from datasets import load_dataset

# Load Luganda main questions
dataset = load_dataset("CraneAILabs/pedagogy-benchmark-multilingual", "luganda", split="cdpk_main")

# Load Nyankore SEND questions
dataset = load_dataset("CraneAILabs/pedagogy-benchmark-multilingual", "nyankore", split="cdpk_send")

# Access a question
print(dataset[0]['question'])
print(f"A) {dataset[0]['answer_a']}")
print(f"B) {dataset[0]['answer_b']}")
print(f"C) {dataset[0]['answer_c']}")
print(f"D) {dataset[0]['answer_d']}")
print(f"Correct: {dataset[0]['correct_answer']}")
```

### Load all splits for a language

```python
from datasets import load_dataset

# Load both splits
dataset_dict = load_dataset("CraneAILabs/pedagogy-benchmark-multilingual", "luganda")

print(f"Main questions: {len(dataset_dict['cdpk_main'])}")
print(f"SEND questions: {len(dataset_dict['cdpk_send'])}")
```

## Dataset Statistics

### Question Categories (cdpk_main)

The main split covers various subject areas:
- Science
- Language and Literature
- Mathematics
- Social Studies
- Arts and Physical Education
- General Pedagogy

### Pedagogical Subdomains

Questions assess knowledge across:
- Assessment and evaluation
- Teaching strategies
- Student understanding
- Curriculum and planning
- Education theories
- Professional development

### Education Levels

- **Pre-primary**: Early childhood education
- **Primary**: Elementary education
- **Secondary**: High school education

### SEND Questions (cdpk_send)

The SEND split focuses on:
- Special educational needs
- Inclusive education strategies
- Differentiated instruction
- Individual education plans
- Assistive technologies

## Original Dataset

This is a translated version of:

**Pedagogy Benchmark**
Original Dataset: [AI-for-Education/pedagogy-benchmark](https://huggingface.co/datasets/AI-for-Education/pedagogy-benchmark)

**Citation:**
```bibtex
@misc{pedagogy-benchmark-2024,
  title={Pedagogy Benchmark: Chilean Teacher Training Exams},
  author={AI for Education},
  year={2024},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/AI-for-Education/pedagogy-benchmark}
}
```

## Translation Details

**Translated by:** CraneAI Labs
**Translation Date:** October 2025
**Model:** Google Gemini 2.0 Flash
**Languages:** Luganda, Nyankore (Swahili in progress)

### Quality Assurance

- Educational terminology verified by native speakers
- Sample validation across all question types
- Metadata and structure preserved exactly
- Checkpoint-based translation with error recovery

## Limitations

- Automated translations may not capture all cultural nuances
- Some pedagogical terms may have multiple valid translations
- Context-specific educational references may need adaptation
- Recommended for use with human review for high-stakes applications

## License

This dataset maintains the **Apache 2.0** license from the original pedagogy-benchmark dataset.

## Contact

For questions or issues with translations:
- **Organization:** [CraneAI Labs](https://huggingface.co/CraneAILabs)
- **Original Dataset:** [AI-for-Education](https://huggingface.co/AI-for-Education)

## Acknowledgments

- Original dataset creators at AI-for-Education
- Chilean Ministry of Education (MINEDUC) for source materials
- Google for Gemini API access
- African language speakers who validated translations

## Authors

Bakunga, Bronson and Mubiru, Kato Steven and Tukamushaba, Catherine — Crane AI Labs

## Citation

```bibtex
@misc{craneailabs2026benchmark,
  title={Multilingual Pedagogy Benchmark for Ugandan Languages},
  author={Bakunga, Bronson and Mubiru, Kato Steven and Tukamushaba, Catherine},
  year={2026},
  publisher={Crane AI Labs}
}
```