Luganda Linguistic Pedagogical Knowledge (LLPK) Benchmark
Multiple-choice benchmark measuring whether a model understands how to teach foundational literacy in the Ugandan context. Built via a hybrid LLM-generation + human-review pipeline grounded in the global reading science (incl. the GEEAP report), the Ministry of Education P1 Luganda Teacher's Guide, and Pilkington's 1915 A Handbook of Luganda. Generated with Gemini 2.5 Flash and validated by an education expert and native Luganda experts. 100 MCQs per language.
Configurations
This dataset has two language configurations:
english— questions written in English (about Luganda)luganda— same questions translated into Luganda
Columns
| Column | Description |
|---|---|
id |
Stable question identifier (e.g. APP_LUG_001) |
section |
Pedagogical theme (e.g. instructional_sequencing, oral_language) |
skill |
Specific skill targeted (e.g. phonics_order, morphological_awareness) |
prompt |
Full multiple-choice question text including options (A)–(D) |
correct_answer |
Single-letter correct answer (A–D) |
source_ref |
References to linguistic / pedagogical sources |
rationale |
Why the correct answer is correct (and distractors are not) |
Source & methodology
Built by AI for Education as part of the Small Language Model finetuning project with Crane AI Labs. See the source repository for the full evaluation harness, scoring code, and reproducibility notes.
- Downloads last month
- 14