Multi-stage Training of Bilingual Islamic LLM for Neural Passage Retrieval

Vera Pavlova · 2025

Read Multi-stage Training of Bilingual Islamic LLM for Neural Passage Retrieval by Vera Pavlova (2025) 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-2501-10175.txt.

Description

Vera Pavlova trains a bilingual search model for Islamic texts, where rich source material exists in Arabic but little in English. The 2025 study builds on a multilingual model and trims it with a language reduction trick. The goal is passage retrieval that works across both tongues.

Training runs in stages, starting from a huge general search dataset and narrowing to small in-domain sets. The author also built a fresh English test set for the domain, grown by augmentation and checked against a trusted Islamic source. The combined recipe lets the bilingual model beat one-language rivals on retrieval tasks. The findings argue that staged training plus domain tuning is the winning mix for niche search.

Catalog record

CREATORVera Pavlova
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE30 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-10175

Browse the full library · About the Tower of Babel