Does Continued Pretraining on a Learner Corpus Improve Automated Essay Scoring on English Proficiency Tests? Evidence from EFCAMDAT
arXiv:2605.25924v1 Announce Type: new Abstract: Recent automated essay scoring (AES) studies increasingly use pretrained transformer models, but these models are usually pretrained on general-domain English and may under-represent second-language learner writing. This study investigates whether domain-adaptive continued pretraining (DAPT) on the EFCAMDAT learner corpus improves transformer-based AES for English proficiency tests. We apply DAPT to three transformer encoders and evaluate them on FCE and IELTS in both in-domain scoring and few-shot cross-dataset transfer. Full-corpus DAPT produce
The proliferation of large language models and their increasing deployment in critical applications like automated assessment necessitates research into their performance and biases, particularly for non-native English speakers.
Improving automated essay scoring for English proficiency tests can standardize and enhance assessment reliability, impacting educational pathways and immigration for millions globally.
The potential for more accurate and equitable automated assessment, moving beyond general-domain models to specialized ones, could reduce biases against second-language learners.
- · Test-takers from non-English speaking backgrounds
- · Educational technology companies
- · English language testing organizations
- · AI developers specializing in domain adaptation
- · Providers of less accurate general-purpose AES
- · Manual essay graders in some contexts
Automated essay scoring systems become more reliable and fair for second-language learners, allowing for broader adoption.
Increased trust in AI-powered assessment could lead to changes in educational curricula and resource allocation for language learning.
Reduced barriers for international students and professionals relying on English proficiency tests, potentially influencing global talent mobility.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL