Model Card for mt5-small-indo-bloom-5e

Model Details

This model is a fine-tuned version of google/mt5-small specifically designed for Pedagogical Automatic Question Generation (AQG) in the Indonesian language. Unlike standard AQG models that are 'pedagogically blind,' this model has been trained with Cognitive Control based on Bloom's Taxonomy (Level C1: Remembering, and Level C2: Understanding).

• Developed by: Firmansyah Ibrahim (Universitas Negeri Malang / UIN Alauddin Makassar)
• Model type: Text-to-Text Transformer (Seq2Seq)
• Language: Indonesian (id)
• Finetuned from model: google/mt5-small

Uses

Prompt Format: generate indonesian question level C[1_or_2]: [Your Reading Context]

Example: generate indonesian question level C2: Budi pergi ke pasar untuk membeli apel dan jeruk.

Citation

BibTeX:

@misc{ibrahim2026mt5bloom,
  author = {Ibrahim, Firmansyah},
  title = {mt5-small-indo-bloom-5e: A Pedagogically Controlled AQG Model for Indonesian},
  year = {2026},
  publisher = {Hugging Face},
  howpublished = {\url{[https://huggingface.co/](https://huggingface.co/)Firmansyah-Ibrahim/mt5-small-indo-bloom-5e}}
}

📊 Quantitative Benchmarking (Evaluation Results)

The model was evaluated using a test set of 100 samples (50 for C1 and 50 for C2) from the Indo-Bloom Corpus.

Metric	Score	Baseline (Awalurahman et al., 2024)	Status
BLEU-1	32.65	41.02	-
BLEU-4	17.38	14.25	SOTA 🚀
ROUGE-L	43.14	54.39	Competitive

Notes: - The score of 17.38 in BLEU-4 outperforms the previous national SOTA record of 14.25.

• This result confirms that pedagogical control (Bloom Taxonomy) improves the structural coherence of generated questions.