Asyncform (AF) dataset

Product or Programme · All products and programmes
Evidence · 6 claims (5 for, 1 mixed) · 1 study (1 design), q2 · 0 of 1 report an effect size · 6 claims rest on one study

Description

A corpus of 863 AI-generated educational questions in English, Mathematics, and Science, labeled by K–12 teachers with Bloom Taxonomy levels and related subject, grade-level, objective, and question metadata; it was created and released by Michael Lawrence Castanares et al.

Components

• Asyncform (AF) dataset of 863 teacher-labeled AI-generated educational questions: The Asyncform (AF) dataset is an out-of-distribution corpus of AI-assisted educational questions created by the authors to test Bloom-level classifier transfer. The article states: "we generated 863 AEQ covering subjects in English, Mathematics, and Science. AF dataset was labeled with corresponding Bloom Taxonomy level by three K-12 teachers who taught the subjects." Each question received a single Bloom label due to limited staff, and annotation included subject, grade level, learning objectives, and question text. (Michael Lawrence Castanares et al. (2026))

Claims

• Retrained BERT relies on high-impact nouns, verbs, and adjectives for correct predictions, but this reliance is muted on the AF dataset [+W]
• Appending learning objectives to questions improves Bloom classification on the AF dataset [+M]
• Feature-based ML Bloom classifiers trained on IID data degrade sharply on out-of-distribution AI-assisted questions [+W]
• AI-generated OOD question datasets are longer, syntactically deeper, and overlap less with Bloom trigger verbs than human-curated IID data [+W]
• LLMs show robust zero- and few-shot Bloom classification of OOD AI-assisted questions compared with ML and BERT models [+W]
• Model retraining on labeled OOD data provides the largest improvement in Bloom classification across models and datasets [~W]

Related Products and Programmes

• Dataset of 434 expert-labelled authentic learner questions augmented to balance class distributions

Key Sources

• Michael Lawrence Castanares, Princess Ventures, and Allan Tan. (2026). Evaluation of Pre-trained Models for Pedagogical Assessment of NovelAI-Assisted Educational Questions. Preprint. https://arxiv.org/abs/2609.27749

Page metadata

Field	Value
type	product
id	asyncform-af-dataset
title	Asyncform (AF) dataset
description	A corpus of 863 AI-generated educational questions in English, Mathematics, and Science, labeled by K–12 teachers with Bloom Taxonomy levels and related subject, grade-level, objective, and question metadata; it was created and released by Michael Lawrence Castanares et al.
status	review
product_kind	dataset
generated	by process:settle-candidates on 2026-10-10
trust	unverified — no one has confirmed this page's content

Frontmatter is the page's machine-readable record — see the Schema & Guide.