Can Embedding Similarity Predict Cross-Lingual Transfer? A Systematic Study on African Languages

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Idris, Tewodros Kederalah, Mitra, Prasenjit, Eiselen, Roald
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912805202952192
author Idris, Tewodros Kederalah
Mitra, Prasenjit
Eiselen, Roald
author_facet Idris, Tewodros Kederalah
Mitra, Prasenjit
Eiselen, Roald
contents Cross-lingual transfer is essential for building NLP systems for low-resource African languages, but practitioners lack reliable methods for selecting source languages. We systematically evaluate five embedding similarity metrics across 816 transfer experiments spanning three NLP tasks, three African-centric multilingual models, and 12 languages from four language families. We find that cosine gap and retrieval-based metrics (P@1, CSLS) reliably predict transfer success ($ρ= 0.4-0.6$), while CKA shows negligible predictive power ($ρ\approx 0.1$). Critically, correlation signs reverse when pooling across models (Simpson's Paradox), so practitioners must validate per-model. Embedding metrics achieve comparable predictive power to URIEL linguistic typology. Our results provide concrete guidance for source language selection and highlight the importance of model-specific analysis.
format Preprint
id arxiv_https___arxiv_org_abs_2601_03168
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Can Embedding Similarity Predict Cross-Lingual Transfer? A Systematic Study on African Languages
Idris, Tewodros Kederalah
Mitra, Prasenjit
Eiselen, Roald
Computation and Language
Machine Learning
Cross-lingual transfer is essential for building NLP systems for low-resource African languages, but practitioners lack reliable methods for selecting source languages. We systematically evaluate five embedding similarity metrics across 816 transfer experiments spanning three NLP tasks, three African-centric multilingual models, and 12 languages from four language families. We find that cosine gap and retrieval-based metrics (P@1, CSLS) reliably predict transfer success ($ρ= 0.4-0.6$), while CKA shows negligible predictive power ($ρ\approx 0.1$). Critically, correlation signs reverse when pooling across models (Simpson's Paradox), so practitioners must validate per-model. Embedding metrics achieve comparable predictive power to URIEL linguistic typology. Our results provide concrete guidance for source language selection and highlight the importance of model-specific analysis.
title Can Embedding Similarity Predict Cross-Lingual Transfer? A Systematic Study on African Languages
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2601.03168