Can Embedding Similarity Predict Cross-Lingual Transfer? A Systematic Study on African Languages
Not on the map. Why?
Abstract (arXiv)
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 ≈ 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.
Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 31 of its 40 references that are in the map. First posted on arXiv 6 Jan 2026.
Builds on
31 of 40 references.
References · 31
- 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
- 2019Unsupervised Cross-lingual Representation Learning at Scalecited by 3,935
- 2021SimCSE: Simple Contrastive Learning of Sentence Embeddingscited by 2,173
- 2019Similarity of Neural Network Representations Revisitedcited by 1,755
- 2018XNLI: Evaluating Cross-lingual Sentence Representationscited by 915
- 2019How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddingscited by 784
- 2019How multilingual is Multilingual BERT?cited by 645
- 2017Word Translation Without Parallel Datacited by 487
- 2018Massively Multilingual Sentence Embeddings for Zero-Shot Cross-Lingual Transfer and Beyondcited by 434
- 2020On the Stability of Fine-tuning BERT: Misconceptions, Explanations, and Strong Baselinescited by 255
- 2019Emerging Cross-lingual Structure in Pretrained Language Modelscited by 181
- 2021MasakhaNER: Named Entity Recognition for African Languagescited by 155
19 more
- 2022Adapting Pre-trained Language Models to African Languages via Multilingual Adaptive Fine-Tuningcited by 149
- 2019Choosing Transfer Languages for Cross-Lingual Learningcited by 144
- 2022A Few Thousand Translations Go a Long Way! Leveraging Pre-trained Models for African News Translationcited by 98
- 2023AfriSenti: A Twitter Sentiment Analysis Benchmark for African Languagescited by 68
- 2021First Align, then Predict: Understanding the Cross-Lingual Ability of Multilingual BERTcited by 65
- 2020Transfer Learning and Distant Supervision for Multilingual Transformer Models: A Study on African Languagescited by 52
- 2021Revisiting the Primacy of English in Zero-shot Cross-lingual Transfercited by 47
- 2022SERENGETI: Massively Multilingual Language Models for Africacited by 40
- 2024Probing the Emergence of Cross-lingual Alignment during LLM Trainingcited by 39
- 2020From Zero to Hero: On the Limitations of Zero-Shot Cross-Lingual Transfer with Multilingual Transformerscited by 35
- 2021A Cluster-based Approach for Improving Isotropy in Contextual Embedding Spacecited by 27
- 2024MEXA: Multilingual Evaluation of English-Centric LLMs via Cross-Lingual Alignmentcited by 25
- 2023MasakhaPOS: Part-of-Speech Tagging for Typologically Diverse African Languagescited by 23
- 2021Evaluating Multilingual Text Encoders for Unsupervised Cross-Lingual Retrievalcited by 16
- 2025Can you map it to English? The Role of Cross-Lingual Alignment in Multilingual Performance of LLMscited by 12
- 2022Languages You Know Influence Those You Learn: Impact of Language Characteristics on Multi-Lingual Text-to-Text Transfercited by 11
- 2025Untangling the Influence of Typology, Data and Model Architecture on Ranking Transfer Languages for Cross-Lingual POS Taggingcited by 9
- 2022Cross-lingual Similarity of Multilingual Representations Revisitedcited by 8
- 2023A Joint Matrix Factorization Analysis of Multilingual Representationscited by 5
Built on by
Citing papers per year: 2026: 3.
3 papers.
- 2026A Cost-Efficient Routing Pipeline for Multilingual Short-Text Classification Using Small Language Modelscited by 1
- 2026Sequential Adapter Stacking for Cross-Lingual Low-Resource ASRcited by 1
- 2026Budget-Xfer: Budget-Constrained Source Language Selection for Cross-Lingual Transfer to African Languagescited by 0
- 2026Sequential Adapter Stacking for Cross-Lingual Low-Resource ASR14 Sep 2026
- 2026A Cost-Efficient Routing Pipeline for Multilingual Short-Text Classification Using Small Language Models11 Aug 2026
- 2026Budget-Xfer: Budget-Constrained Source Language Selection for Cross-Lingual Transfer to African Languages29 Mar 2026