Transformer-encoder knowledge tracing framework with IEU/IIU input vectors

Research Method · All research methods
Evidence · 6 claims (6 for) · 1 study (1 design), q2 · 1 of 1 report an effect size · 6 claims rest on one study

Description

A deep knowledge tracing framework for the ASSISTments dataset that predicts students' responses to end-of-unit test problems from action logs of in-unit assignments. It constructs IEU vectors (problem, sequence, student, and class information) as queries and IIU vectors (adding action-type embeddings and action features with positional encoding) as keys and values for a Transformer Encoder. The article states it is "a novel Transformer-based framework for the ASSISTments dataset which integrates several data-preprocessing techniques with a Transformer-based predictive model". Code, saved models, and predictions are released on OSF.

Accounts

• Transformer-encoder knowledge tracing framework with IEU/IIU input vectors and released code, models, and predictions: A deep knowledge tracing framework for the ASSISTments dataset that predicts students' responses to end-of-unit test problems from action logs of in-unit assignments. It constructs IEU vectors (problem, sequence, student, and class information) as queries and IIU vectors (adding action-type embeddings and action features with positional encoding) as keys and values for a Transformer Encoder. The article states it is "a novel Transformer-based framework for the ASSISTments dataset which integrates several data-preprocessing techniques with a Transformer-based predictive model". Code, saved models, and predictions are released on OSF. (Lu et al. (2024))

Claims

• The single-head base model (Model 1) outperformed the more complex Model 2, suggesting overfitting in the larger architecture [+M]
• Ablation study: removing any component lowers evaluation AUC, and removing all additional features yields the lowest public and private AUCs [+M]
• Single instances of help-seeking actions (answer requested, explanation requested) carry more predictive information than single correct or open responses (ISA) [+M]
• The proposed Transformer-based framework achieved first place in the EDM Cup 2023 with an AUC of 78.969% on the private evaluation dataset [+M]
• NCA scores of IIU input vectors order action types identically to the main-effects ordering (Spearman's ρ = 1) [+M]
• In artificial action logs, \"answer requested\" yields the lowest average predicted probability of a correct end-of-unit response, below even \"wrong response\" [+M]

Related Research Methods

• Coherence analysis framework and its six activity-trace measures

Key Sources

• Lu, Y., Tong, L., & Cheng, Y. (2024). Advanced Knowledge Tracing: Incorporating Process Data and Curricula Information via an Attention-Based Framework for Accuracy and Interpretability. Journal of Educational Data Mining, 16(2). https://osf.io/mdpzc/. doi:10.31219/osf.io/3k7z2

Page metadata

Field	Value
type	research-method
id	transformer-encoder-knowledge-tracing-framework-with-ieuiiu-input-vectors
title	Transformer-encoder knowledge tracing framework with IEU/IIU input vectors
description	A deep knowledge tracing framework for the ASSISTments dataset that predicts students' responses to end-of-unit test problems from action logs of in-unit assignments.
status	review
moved_from	element/transformer-ieu-iiu-kt-framework
generated	by process:sweep-kinds on 2026-10-10
trust	unverified — no one has confirmed this page's content
sources	1 — lu-2024

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