Gender Disambiguation in Machine Translation: Diagnostic Evaluation in Decoder-Only Architectures

2026

Read Gender Disambiguation in Machine Translation: Diagnostic Evaluation in Decoder-Only Architectures (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-17952.txt.

Description

Manna and co-authors test gender handling in decoder-only translation models. Languages mark gender in unlike ways, so translation must often guess what the source left vague. Their framework adds a Prior Bias score for a model's default gender lean. Decoder-only systems did not clearly beat encoder-decoder ones on gender metrics, they find. Extra post-training sharpened context use and cut the masculine lean. Scale alone did not fix the bias.

Catalog record

YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE37238
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.17952

Browse the full library · About the Tower of Babel