Multilingual Sentence Embeddings for Linguistic-Integrated Reliability Audit

A new study from researchers Ummugul Bezirhan, Ji Yoon Jung, and Matthias von Davier finds that multilingual sentence embeddings can replace translated English input for Linguistic-Integrated Reliability Auditing (LiRA) across 11 PIRLS constructed-response items and three embedding models, with native-language embeddings reproducing translation-based reliability estimates closely and recovering responses excluded after translation failure without meaningful change in reliability.

Multilingual Sentence Embeddings for Linguistic-Integrated Reliability Audit By Ummugul Bezirhan, Ji Yoon Jung, Matthias von DavierSource: arXiv cs.CL https://arxiv.org/list/cs.CL/recent arXiv:2607.17466v1 Announce Type: new Abstract: Multilingual assessment systems commonly rely on translation for scoring and quality-control processes. We evaluate whether multilingual sentence embeddings can replace translated English input for Linguistic-Integrated Reliability Auditing LiRA across 11 PIRLS constructed-response items and three embedding /glossary/embedding models. Native-language embeddings reproduced translation-based reliability estimates closely while recovering responses excluded after translation failure, with no meaningful change in reliability.Get AI news in your inbox Daily digest of what matters in AI.