A multitask matrix factorization approach that jointly learns grades across multiple assessment labels outperforms a single-task (binary relevance) baseline that learns each label separately.¶
Claim · All claims
Evidence · 1 study · designr2·q2
Currency · newest study 2018 · checked against its sources 2026-10-30 (automated, 1 of 1 entry) · not reviewed by a person
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? A multitask matrix factorization approach that jointly learns grades across multiple assessment labels outperforms a single-task baseline that learns each label separately. → Luaces 2018
Evidence¶
Luaces 2018¶
Luaces, O., Díez, J., & Bahamonde, A. (2018). A peer assessment method to provide feedback, consistent grading and reduce students’ burden in massive teaching settings. Computers & Education. https://doi.org/10.1016/j.compedu.2018.07.016
q2 · i? · design · r2
In experiments across three courses, the multitask approach, which learns grading patterns for all labels simultaneously, achieved better accuracy than a binary relevance baseline that learns a separate model per label.
"We have presented a method that uses a multitask approach to search for grading patterns in all labels at the same time. Multitask leverages the accuracy of a baseline that successively focuses on each label separately."
Discussion¶
Related Claims¶
- A matrix factorization model trained on sparse peer assessment data can predict grades for answers from students who did not participate in grading, reducing the grading burden on students. — related
- Jointly predicting courses and grades reduces grade-prediction MAE by 46.4% versus an identical Transformer that predicts grades only — a broader claim this one bears on
- Multi-task learning on a large multi-topic dataset outperforms separate topic-specific essay scoring models — a broader claim this one bears on
- KL-divergence and binary cross-entropy losses for course prediction yield grade prediction within 1-2% of each other — related
Page metadata
| Field | Value |
|---|---|
type |
claim |
id |
multitask-matrix-factorization-outperforms-single-task-baseline |
title |
A multitask matrix factorization approach that jointly learns grades across multiple assessment labels outperforms a single-task (binary relevance) baseline that learns each label separately. |
description |
A multitask matrix factorization approach that jointly learns grades across multiple assessment labels outperforms a single-task (binary relevance) baseline that learns each label separately. |
status |
review |
evidence_strength |
moderate |
newest_evidence |
2018 |
source_checked |
2026-10-30 |
generated |
by process:wiki-ingest on 2026-10-30 |
trust |
unverified — no one has confirmed this page's content |
sources |
1 — luaces-2018 |
Frontmatter is the page's machine-readable record — see the Schema & Guide.