Claims

Multimodal XGBoost models achieved rubric-aligned scoring comparable to expert ratings, and no single modality approached multimodal performance

Claim · All claims
Evidence · 1 study (2 entries) · 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

Reading the evidence codes

q is the design tier, i the impact, n the sample, and the kind with its r the rigour judged within that kind: what each code means.

This claim: no standardized effect is recorded in observations/ for it yet, so no benchmark reading is possible.

q2 i? Seven speaker-independent XGBoost regression models trained on 10,360 MOOC video segments reached R² = 0.48–0.61, Spearman's ρ = 0.69–0.78, and MAE = 0.43–0.57 against expert BARS ratings. → Hung-Yue Suen and Kuo-En Hung 2026 q2 i? Unimodal baselines using only acoustic, facial, oculomotor, or textual features performed noticeably weaker (R² mostly 0.12–0.32; ρ around 0.34–0.56). → Hung-Yue Suen and Kuo-En Hung 2026 (2)

Evidence

Hung-Yue Suen and Kuo-En Hung 2026

Hung-Yue Suen and Kuo-En Hung. (2026). An Interpretable Closed-Loop Intelligent Tutoring System for Multimodal Affective Feedback in Asynchronous Presentation Training. IEEE Transactions on Learning Technologies. https://doi.org/10.1109/TLT.2026.3693864

q2 · i? · design · r2

Speaker-independent evaluation on a held-out test set (15% partition) of seven XGBoost regression models trained on 4,374-dimensional early-fused multimodal features from 10,360 MOOC segments annotated by about 30 experts.

"Across the dimensions, R² values ranged from 0.48 to 0.61, Spearman's ρ fell between 0.69 and 0.78, and MAE values were generally between 0.43 and 0.57."

Hung-Yue Suen and Kuo-En Hung 2026 (2)

Hung-Yue Suen and Kuo-En Hung. (2026). An Interpretable Closed-Loop Intelligent Tutoring System for Multimodal Affective Feedback in Asynchronous Presentation Training. IEEE Transactions on Learning Technologies. https://doi.org/10.1109/TLT.2026.3693864

q2 · i? · design · r2

Unimodal baseline models trained with only acoustic, facial, oculomotor, or textual features; "None of the single-modality models approached the consistency observed in the multimodal setting," supporting early feature-level fusion.

"unimodal R² values mostly fell between 0.12 and 0.32, with Spearman correlations around 0.34 to 0.56. None of the single-modality models approached the consistency observed in the multimodal setting."

Discussion

Page metadata
Field Value
type claim
id multimodal-xgboost-expert-comparable-scoring
title Multimodal XGBoost models achieved rubric-aligned scoring comparable to expert ratings, and no single modality approached multimodal performance
description Multimodal XGBoost models achieved rubric-aligned scoring comparable to expert ratings, and no single modality approached multimodal performance
status review
evidence_strength moderate
newest_evidence 2026
recheck_by 2029
generated by process:wiki-ingest on 2026-10-11
trust unverified — no one has confirmed this page's content
sources 2 — hung-yue-suen-and-kuo-en-hung-2026, hung-yue-suen-and-kuo-en-hung-2026-2

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