GPT-4 surpasses other LLMs in the AGE query selection with an accuracy rate 15% better¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2024 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? GPT-4 surpassed other LLMs with the accuracy rate 15% better in the AGE model on the UNT dataset. → Mohammad Iqbal 2024

Evidence¶

Mohammad Iqbal 2024¶

Mohammad Iqbal, Tsamara Rana Nugraha and Hsing-Kuo Pao. (2024). Active Grade Estimator on Short Answer Assessment. https://doi.org/10.21203/rs.3.rs-5651820/v1

q2 · i? · design · r2

Numerical comparison of different LLMs (BERT, SGPT, GPT-4) within AGE on the UNT dataset shown in Fig. 7(b). The article reports "GPT-4 surpassed other LLMs w ith the accuracy rate is 15% better".

"It turns out that GPT-4 surpassed other LLMs w ith the accuracy rate is 15% better. That is, the proposed model is designed with GPT-4 as the LLMs on the query selections."

Discussion¶

Related Claims¶

• AGE reduces teacher grading burden by requiring only 1% of student answers in the UNT dataset to be graded — related
• The two-stage query with SCS elevates grading process up to 40% on the LL dataset and up to 30% at the beginning on the JRL dataset compared to conventional AL — related
• Hyperparameter setting with maximum token 53 and temperature 1 (Set C) yields the most accurate AGE grading performance
• On the SciC classification task, AGE achieves more than 80% accuracy after roughly the first 30 graded student answers
• On the mismatch-grade dataset, all models without the voting mechanism diverge, while AGE with VM estimates student answers increasingly precisely