Cross-Lingual Ability of Multilingual BERT: An Empirical Study
Not on the map. Why?
Abstract (arXiv)
Recent work has exhibited the surprising cross-lingual abilities of multilingual BERT (M-BERT) -- surprising since it is trained without any cross-lingual objective and with no aligned data. In this work, we provide a comprehensive study of the contribution of different components in M-BERT to its cross-lingual ability. We study the impact of linguistic properties of the languages, the architecture of the model, and the learning objectives. The experimental study is done in the context of three typologically different languages -- Spanish, Hindi, and Russian -- and using two conceptually different NLP tasks, textual entailment and named entity recognition. Among our key conclusions is the fact that the lexical overlap between languages plays a negligible role in the cross-lingual success, while the depth of the network is an integral part of it. All our models and implementations can be found on our project page: http://cogcomp.org/page/publication_view/900 .
Key facts. Cited by 182 papers in the map (data through 7 Oct 2026). First posted on arXiv 17 Dec 2019.
Builds on
No reference list yet.
Built on by
Citing papers per year: 2021: 53; 2022: 39; 2023: 32; 2024: 25; 2025: 14; 2026: 19.
182 papers.
Sort:
By area:
- 2021CLIPScore: A Reference-free Evaluation Metric for Image Captioningcited by 2,388
- 2023ChatGPT Beyond English: Towards a Comprehensive Evaluation of Large Language Models in Multilingual Learningcited by 203
- 2023Okapi: Instruction-tuned Large Language Models in Multiple Languages with Reinforcement Learning from Human Feedbackcited by 189
- 2024How do Large Language Models Handle Multilingualism?cited by 148
- 2022Lifting the Curse of Multilinguality by Pre-training Modular Transformerscited by 145
- 2022MASSIVE: A 1M-Example Multilingual Natural Language Understanding Dataset with 51 Typologically-Diverse Languagescited by 142
- 2021Between words and characters: A Brief History of Open-Vocabulary Modeling and Tokenization in NLPcited by 130
- 2024Language-Specific Neurons: The Key to Multilingual Capabilities in Large Language Modelscited by 124
- 2021BanglaBERT: Language Model Pretraining and Benchmarks for Low-Resource Language Understanding Evaluation in Banglacited by 123
- 2024The Language Barrier: Dissecting Safety Challenges of LLMs in Multilingual Contextscited by 92
- 2024A Survey on Multilingual Large Language Models: Corpora, Alignment, and Biascited by 82
- 2021xGQA: Cross-Lingual Visual Question Answeringcited by 72
38 more
- 2024Multilingual Instruction Tuning With Just a Pinch of Multilingualitycited by 71
- 2022A Survey on Aspect-Based Sentiment Analysis: Tasks, Methods, and Challengescited by 61
- 2023Med-UniC: Unifying Cross-Lingual Medical Vision-Language Pre-Training by Diminishing Biascited by 57
- 2024ALLaM: Large Language Models for Arabic and Englishcited by 55
- 2023Towards a Common Understanding of Contributing Factors for Cross-Lingual Transfer in Multilingual Language Models: A Reviewcited by 50
- 2021A Primer on Pretrained Multilingual Language Modelscited by 49
- 2023EcomGPT: Instruction-tuning Large Language Models with Chain-of-Task Tasks for E-commercecited by 48
- 2021Revisiting the Primacy of English in Zero-shot Cross-lingual Transfercited by 47
- 2021MT6: Multilingual Pretrained Text-to-Text Transformer with Translation Pairscited by 41
- 2021Improving Pretrained Cross-Lingual Language Models via Self-Labeled Word Alignmentcited by 41
- 2021When is BERT Multilingual? Isolating Crucial Ingredients for Cross-lingual Transfercited by 41
- 2022MasakhaNER 2.0: Africa-centric Transfer Learning for Named Entity Recognitioncited by 38
- 2023OFA: A Framework of Initializing Unseen Subword Embeddings for Efficient Large-scale Multilingual Continued Pretrainingcited by 37
- 2022Analyzing the Mono- and Cross-Lingual Pretraining Dynamics of Multilingual Language Modelscited by 32
- 2023When Is Multilinguality a Curse? Language Modeling for 250 High- and Low-Resource Languagescited by 32
- 2021Multilingual Language Models Predict Human Reading Behaviorcited by 29
- 2022A Balanced Data Approach for Evaluating Cross-Lingual Transfer: Mapping the Linguistic Blood Bankcited by 29
- 2022Pretraining with Artificial Language: Studying Transferable Knowledge in Language Modelscited by 28
- 2022Discovering Language-neutral Sub-networks in Multilingual Language Modelscited by 28
- 2022Naamapadam: A Large-Scale Named Entity Annotated Data for Indic Languagescited by 28
- 2023Zero-shot cross-lingual transfer language selection using linguistic similaritycited by 28
- 2024Goldfish: Monolingual Language Models for 350 Languagescited by 27
- 2022Match the Script, Adapt if Multilingual: Analyzing the Effect of Multilingual Pretraining on Cross-lingual Transferabilitycited by 26
- 2022Towards Best Practices for Training Multilingual Dense Retrieval Modelscited by 26
- 2024Reuse Your Rewards: Reward Model Transfer for Zero-Shot Cross-Lingual Alignmentcited by 26
- 2021An Isotropy Analysis in the Multilingual BERT Embedding Spacecited by 25
- 2022Overlap-based Vocabulary Generation Improves Cross-lingual Transfer Among Related Languagescited by 25
- 2022Insights into Pre-training via Simpler Synthetic Taskscited by 25
- 2021Predicting the Performance of Multilingual NLP Modelscited by 24
- 2021Code-Mixing on Sesame Street: Dawn of the Adversarial Polyglotscited by 23
- 2021Multilingual Offensive Language Identification for Low-resource Languagescited by 22
- 2021On Cross-Lingual Retrieval with Multilingual Text Encoderscited by 22
- 2023Why Does Zero-Shot Cross-Lingual Generation Fail? An Explanation and a Solutioncited by 22
- 2021Improving Zero-Shot Cross-Lingual Transfer Learning via Robust Trainingcited by 21
- 2021Specializing Multilingual Language Models: An Empirical Studycited by 21
- 2021Multilingual Domain Adaptation for NMT: Decoupling Language and Domain Information with Adapterscited by 21
- 2022LVP-M3: Language-aware Visual Prompt for Multilingual Multimodal Machine Translationcited by 21
- 2022Multi Task Learning For Zero Shot Performance Prediction of Multilingual Modelscited by 20
50 most cited of 182 All 182 (JSON)
- 2026Why Pretraining Fails to Share Cross-Lingual Knowledge16 Sep 2026
- 2026Structural priors for data-efficient language learning10 Sep 2026
- 2026Cross-Lingual Alignment Without Joint Training: Do Monolingual Language Models Converge on Universal Representations?27 Aug 2026
- 2026One Form to Transfer Them All: Pretraining Multilingual Language Models Beyond Native Orthography26 Aug 2026
- 2026When English Isn't the Best Teacher: Source Language Effects in Cross-Lingual In-Context Learning16 Jun 2026
- 2026Multilingual Unlearning in LLMs: Transfer, Dynamics, and Reversibility2 Jun 2026
- 2026From Traditional Taggers to LLMs: A Comparative Study of POS Tagging for Medieval Romance Languages9 May 2026
- 2026The Impact of Vocabulary Overlaps on Knowledge Transfer in Multilingual Machine Translation5 May 2026
- 2026Is Cross-Lingual Transfer in Bilingual Models Human-Like? A Study with Overlapping Word Forms in Dutch and English8 Apr 2026
- 2026Rashid: A Cipher-Based Framework for Exploring In-Context Language Learning23 Mar 2026
- 2026Why Better Cross-Lingual Alignment Fails for Better Cross-Lingual Transfer: Case of Encoders19 Mar 2026
- 2026From FusHa to Folk: Exploring Cross-Lingual Transfer in Arabic Language Models10 Feb 2026
38 more
- 2026Stop Jostling: Adaptive Negative Sampling Reduces the Marginalization of Low-Resource Language Tokens by Cross-Entropy Loss30 Jan 2026
- 2026A Study of Crosslinguistic Influence in Language Models29 Jan 2026
- 2026XProvence: Zero-Cost Multilingual Context Pruning for Retrieval-Augmented Generation26 Jan 2026
- 2026The Roots of Performance Disparity in Multilingual Language Models: Intrinsic Modeling Difficulty or Design Choices?12 Jan 2026
- 2026Evaluating Cross-Lingual Unlearning in Multilingual Language Models10 Jan 2026
- 2026The Grammar of Transformers: A Systematic Review of Interpretability Research on Syntactic Knowledge in Language Models9 Jan 2026
- 2026CRANE: Causal Relevance Analysis of Language-Specific Neurons in Multilingual Large Language Models8 Jan 2026
- 2025TransAlign: Machine Translation Encoders are Strong Word Aligners, Too31 Oct 2025
- 2025A-IPO: Adaptive Intent-driven Preference Optimization11 Oct 2025
- 2025Model-Based Ranking of Source Languages for Zero-Shot Cross-Lingual Transfer3 Oct 2025
- 2025Evaluating the Limits of Large Language Models in Multilingual Legal Reasoning26 Sep 2025
- 2025False Friends Are Not Foes: Investigating Vocabulary Overlap in Multilingual Language Models23 Sep 2025
- 2025Beyond the Rosetta Stone: Unification Forces in Generalization Dynamics14 Aug 2025
- 2025Cross-lingual Aspect-Based Sentiment Analysis: A Survey on Tasks, Approaches, and Challenges13 Aug 2025
- 2025Are Knowledge and Reference in Multilingual Language Models Cross-Lingually Consistent?17 Jul 2025
- 2025Multilinguality Does not Make Sense: Investigating Factors Behind Zero-Shot Transfer in Sense-Aware Tasks30 May 2025
- 2025The Devil Is in the Word Alignment Details: On Translation-Based Cross-Lingual Transfer for Token Classification Tasks15 May 2025
- 2025Untangling the Influence of Typology, Data and Model Architecture on Ranking Transfer Languages for Cross-Lingual POS Tagging25 Mar 2025
- 2025Multi-Granular Multimodal Clue Fusion for Meme Understanding16 Mar 2025
- 2025FewTopNER: Integrating Few-Shot Learning with Topic Modeling and Named Entity Recognition in a Multilingual Framework4 Feb 2025
- 2025Can MLLMs Generalize to Multi-Party dialog? Exploring Multilingual Response Generation in Complex Scenarios20 Jan 2025
- 2024Lens: Rethinking Multilingual Enhancement for Large Language Models6 Oct 2024
- 2024Goldfish: Monolingual Language Models for 350 Languages19 Aug 2024
- 2024PreAlign: Boosting Cross-Lingual Transfer by Early Establishment of Multilingual Alignment23 Jul 2024
- 2024ALLaM: Large Language Models for Arabic and English22 Jul 2024
- 2024Data, Data Everywhere: A Guide for Pretraining Dataset Construction8 Jul 2024
- 2024Teaching LLMs to Abstain across Languages via Multilingual Feedback22 Jun 2024
- 2024Learn and Unlearn: Addressing Misinformation in Multilingual LLMs19 Jun 2024
- 2024Decipherment-Aware Multilingual Learning in Jointly Trained Language Models11 Jun 2024
- 2024Unknown Script: Impact of Script on Cross-Lingual Transfer29 Apr 2024
- 2024mOthello: When Do Cross-Lingual Representation Alignment and Cross-Lingual Transfer Emerge in Multilingual Models?18 Apr 2024
- 2024Reuse Your Rewards: Reward Model Transfer for Zero-Shot Cross-Lingual Alignment18 Apr 2024
- 2024The Role of Language Imbalance in Cross-lingual Generalisation: Insights from Cloned Language Experiments11 Apr 2024
- 2024On the Effect of (Near) Duplicate Subwords in Language Modelling9 Apr 2024
- 2024A Survey on Multilingual Large Language Models: Corpora, Alignment, and Bias1 Apr 2024
- 2024The Impact of Syntactic and Semantic Proximity on Machine Translation with Back-Translation26 Mar 2024
- 2024Tracing the Roots of Facts in Multilingual Language Models: Independent, Shared, and Transferred Knowledge8 Mar 2024
- 2024How do Large Language Models Handle Multilingualism?29 Feb 2024
50 newest of 182 All 182 (JSON)
- 2023Okapi: Instruction-tuned Large Language Models in Multiple Languages with Reinforcement Learning from Human Feedbackcited by 189
- 2024The Language Barrier: Dissecting Safety Challenges of LLMs in Multilingual Contextscited by 92
- 2024Reuse Your Rewards: Reward Model Transfer for Zero-Shot Cross-Lingual Alignmentcited by 26
- 2025A-IPO: Adaptive Intent-driven Preference Optimizationcited by 1
- 2025A-IPO: Adaptive Intent-driven Preference Optimization11 Oct 2025
- 2024Reuse Your Rewards: Reward Model Transfer for Zero-Shot Cross-Lingual Alignment18 Apr 2024
- 2024The Language Barrier: Dissecting Safety Challenges of LLMs in Multilingual Contexts23 Jan 2024
- 2023Okapi: Instruction-tuned Large Language Models in Multiple Languages with Reinforcement Learning from Human Feedback29 Jul 2023
- 2021CLIPScore: A Reference-free Evaluation Metric for Image Captioningcited by 2,388
- 2023ChatGPT Beyond English: Towards a Comprehensive Evaluation of Large Language Models in Multilingual Learningcited by 203
- 2024How do Large Language Models Handle Multilingualism?cited by 148
- 2022Lifting the Curse of Multilinguality by Pre-training Modular Transformerscited by 145
- 2022MASSIVE: A 1M-Example Multilingual Natural Language Understanding Dataset with 51 Typologically-Diverse Languagescited by 142
- 2021Between words and characters: A Brief History of Open-Vocabulary Modeling and Tokenization in NLPcited by 130
- 2024Language-Specific Neurons: The Key to Multilingual Capabilities in Large Language Modelscited by 124
- 2021BanglaBERT: Language Model Pretraining and Benchmarks for Low-Resource Language Understanding Evaluation in Banglacited by 123
- 2024A Survey on Multilingual Large Language Models: Corpora, Alignment, and Biascited by 82
- 2021xGQA: Cross-Lingual Visual Question Answeringcited by 72
- 2024Multilingual Instruction Tuning With Just a Pinch of Multilingualitycited by 71
- 2022A Survey on Aspect-Based Sentiment Analysis: Tasks, Methods, and Challengescited by 61
12 most cited of 171 All 171 (JSON)
- 2026Why Pretraining Fails to Share Cross-Lingual Knowledge16 Sep 2026
- 2026Structural priors for data-efficient language learning10 Sep 2026
- 2026Cross-Lingual Alignment Without Joint Training: Do Monolingual Language Models Converge on Universal Representations?27 Aug 2026
- 2026One Form to Transfer Them All: Pretraining Multilingual Language Models Beyond Native Orthography26 Aug 2026
- 2026When English Isn't the Best Teacher: Source Language Effects in Cross-Lingual In-Context Learning16 Jun 2026
- 2026From Traditional Taggers to LLMs: A Comparative Study of POS Tagging for Medieval Romance Languages9 May 2026
- 2026The Impact of Vocabulary Overlaps on Knowledge Transfer in Multilingual Machine Translation5 May 2026
- 2026Is Cross-Lingual Transfer in Bilingual Models Human-Like? A Study with Overlapping Word Forms in Dutch and English8 Apr 2026
- 2026Rashid: A Cipher-Based Framework for Exploring In-Context Language Learning23 Mar 2026
- 2026Why Better Cross-Lingual Alignment Fails for Better Cross-Lingual Transfer: Case of Encoders19 Mar 2026
- 2026From FusHa to Folk: Exploring Cross-Lingual Transfer in Arabic Language Models10 Feb 2026
- 2026Stop Jostling: Adaptive Negative Sampling Reduces the Marginalization of Low-Resource Language Tokens by Cross-Entropy Loss30 Jan 2026
12 newest of 171 All 171 (JSON)