Salvato in:
Dettagli Bibliografici
Autori principali: Xia, Xiao-li, Li, Hou-biao
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2512.06629
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866914185277865984
author Xia, Xiao-li
Li, Hou-biao
author_facet Xia, Xiao-li
Li, Hou-biao
contents Knowledge Tracing (KT) models face a critical ``Performance-Complexity Trap'': capturing complex cognitive dynamics like learning sessions and memory decay typically requires deep hierarchical architectures, which incur prohibitive computational costs for real-time deployment. To resolve this, we propose FlatFormer, a streamlined architecture based on the novel design paradigm of ``Information Injection over Structural Stacking.'' Unlike parameter-heavy hierarchical models, FlatFormer leverages a standard flat Transformer augmented with two lightweight injection mechanisms: (i) a hybrid input encoding strategy combining learnable session identifiers with fixed sinusoidal step embeddings; and (ii) a pre-computed power-law bias integrated directly into attention logits to explicitly model the forgetting curve. Extensive experiments on four large-scale datasets (e.g., EdNet, Junyi) show that FlatFormer achieves state-of-the-art performance. For example, on the EdNet dataset, compared to the strongest hierarchical baseline (HiTSKT), its absolute AUC increased by 8.3%, while using less than 15% of parameters, and inference speed was about three times faster. These results validate that high cognitive fidelity does not necessitate architectural complexity.
format Preprint
id arxiv_https___arxiv_org_abs_2512_06629
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle FlatFormer: A Flat Transformer Knowledge Tracing Model Based on Cognitive Bias Injection
Xia, Xiao-li
Li, Hou-biao
Artificial Intelligence
Information Theory
68T07, 68T37
I.2
Knowledge Tracing (KT) models face a critical ``Performance-Complexity Trap'': capturing complex cognitive dynamics like learning sessions and memory decay typically requires deep hierarchical architectures, which incur prohibitive computational costs for real-time deployment. To resolve this, we propose FlatFormer, a streamlined architecture based on the novel design paradigm of ``Information Injection over Structural Stacking.'' Unlike parameter-heavy hierarchical models, FlatFormer leverages a standard flat Transformer augmented with two lightweight injection mechanisms: (i) a hybrid input encoding strategy combining learnable session identifiers with fixed sinusoidal step embeddings; and (ii) a pre-computed power-law bias integrated directly into attention logits to explicitly model the forgetting curve. Extensive experiments on four large-scale datasets (e.g., EdNet, Junyi) show that FlatFormer achieves state-of-the-art performance. For example, on the EdNet dataset, compared to the strongest hierarchical baseline (HiTSKT), its absolute AUC increased by 8.3%, while using less than 15% of parameters, and inference speed was about three times faster. These results validate that high cognitive fidelity does not necessitate architectural complexity.
title FlatFormer: A Flat Transformer Knowledge Tracing Model Based on Cognitive Bias Injection
topic Artificial Intelligence
Information Theory
68T07, 68T37
I.2
url https://arxiv.org/abs/2512.06629