A DDQN-based attention quantification model trained on multimodal EEG-fNIRS features produces predicted attention levels at one-second temporal resolution¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? After training, the optimized DDQN model quantifies student attention on test data, producing a predicted attention level (high, medium, or low) at one-second resolution. → Shixian Zhou 2026

Evidence¶

Shixian Zhou 2026¶

Shixian Zhou, Minghuan Shen, Xiaolin Wen, Zijun Qiu, Yongliang Jiang, Xiangyang Wu, Fei Wu, Yong Wang, and Zhiguang Zhou. (2026). SAVVY: Student Attention Visualization for Video-based Learning Analysis. arXiv. https://arxiv.org/abs/2607.29413 doi:10.48550/arxiv.2607.29413

q2 · i? · design · r2

Attention quantification model evaluation described in the Attention Modeling subsection: a DDQN trained on 80% of the labeled dataset and tested on 20%, outputting "a predicted attention level (high, medium, or low)" per second. No effect size or test statistic is printed in the available text.

"After training, the optimized model is used to quantify student attention on the test data, producing a predicted attention level (high, medium, or low) at a one- second temporal resolution."

Discussion¶

Related Claims¶

• A multimodal EEG-fNIRS dataset was collected from 31 participants watching three instructional videos, yielding over 4.5 million data points with expert-verified attention labels — related
• A modality-aware reference model outperforms sensor-free, statistical, deep temporal, foundation-model, and LLM-based baselines for estimating minute-level attention difficulty from wearable sensing — related
• The end-to-end attention-estimation pipeline meets the one-second prediction stride on an NVIDIA A100 GPU, with warm latency of 50.65 ms per window and 1.02 ms for the temporal model