mmBERT: A Modern Multilingual Encoder with Annealed Language Learning

• URL: http://arxiv.org/abs/2509.06888v1
• Date: Mon, 08 Sep 2025 17:08:42 GMT
• Title: mmBERT: A Modern Multilingual Encoder with Annealed Language Learning
• Authors: Marc Marone, Orion Weller, William Fleshman, Eugene Yang, Dawn Lawrie, Benjamin Van Durme,
• Abstract summary: mmBERT is an encoder-only language model pretrained on 3T tokens of multilingual text.<n>We add over 1700 low-resource languages to the data mix only during the decay phase.<n>We show that mmBERT significantly outperforms the previous generation of models on classification and retrieval tasks.
• Score: 57.58071656545661
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Encoder-only languages models are frequently used for a variety of standard machine learning tasks, including classification and retrieval. However, there has been a lack of recent research for encoder models, especially with respect to multilingual models. We introduce mmBERT, an encoder-only language model pretrained on 3T tokens of multilingual text in over 1800 languages. To build mmBERT we introduce several novel elements, including an inverse mask ratio schedule and an inverse temperature sampling ratio. We add over 1700 low-resource languages to the data mix only during the decay phase, showing that it boosts performance dramatically and maximizes the gains from the relatively small amount of training data. Despite only including these low-resource languages in the short decay phase we achieve similar classification performance to models like OpenAI's o3 and Google's Gemini 2.5 Pro. Overall, we show that mmBERT significantly outperforms the previous generation of models on classification and retrieval tasks -- on both high and low-resource languages.

Related papers

• mOSCAR: A Large-scale Multilingual and Multimodal Document-level Corpus [52.83121058429025]
  Multimodal Large Language Models (mLLMs) are trained on a large amount of text-image data.<n>mOSCAR is the first large-scale multilingual and multimodal document corpus crawled from the web.<n>It covers 163 languages, 303M documents, 200B tokens and 1.15B images.
  arXiv  Detail & Related papers  (2024-06-13T00:13:32Z)
• Distilling Efficient Language-Specific Models for Cross-Lingual Transfer [75.32131584449786]
  Massively multilingual Transformers (MMTs) are widely used for cross-lingual transfer learning. MMTs' language coverage makes them unnecessarily expensive to deploy in terms of model size, inference time, energy, and hardware cost. We propose to extract compressed, language-specific models from MMTs which retain the capacity of the original MMTs for cross-lingual transfer.
  arXiv  Detail & Related papers  (2023-06-02T17:31:52Z)
• Distilling a Pretrained Language Model to a Multilingual ASR Model [3.4012007729454816]
  We distill the rich knowledge embedded inside a well-trained teacher text model to the student speech model. We show the superiority of our method on 20 low-resource languages of the CommonVoice dataset with less than 100 hours of speech data.
  arXiv  Detail & Related papers  (2022-06-25T12:36:11Z)
• Pre-training Data Quality and Quantity for a Low-Resource Language: New Corpus and BERT Models for Maltese [4.4681678689625715]
  We analyse the effect of pre-training with monolingual data for a low-resource language. We present a newly created corpus for Maltese, and determine the effect that the pre-training data size and domain have on the downstream performance. We compare two models on the new corpus: a monolingual BERT model trained from scratch (BERTu), and a further pre-trained multilingual BERT (mBERTu)
  arXiv  Detail & Related papers  (2022-05-21T06:44:59Z)
• mGPT: Few-Shot Learners Go Multilingual [1.4354798873010843]
  This paper introduces two autoregressive GPT-like models with 1.3 billion and 13 billion parameters trained on 60 languages. We reproduce the GPT-3 architecture using GPT-2 sources and the sparse attention mechanism. The resulting models show performance on par with the recently released XGLM models by Facebook.
  arXiv  Detail & Related papers  (2022-04-15T13:02:33Z)
• Breaking Down Multilingual Machine Translation [74.24795388967907]
  We show that multilingual training is beneficial to encoders in general, while it only benefits decoders for low-resource languages (LRLs) Our many-to-one models for high-resource languages and one-to-many models for LRLs outperform the best results reported by Aharoni et al.
  arXiv  Detail & Related papers  (2021-10-15T14:57:12Z)
• UNKs Everywhere: Adapting Multilingual Language Models to New Scripts [103.79021395138423]
  Massively multilingual language models such as multilingual BERT (mBERT) and XLM-R offer state-of-the-art cross-lingual transfer performance on a range of NLP tasks. Due to their limited capacity and large differences in pretraining data, there is a profound performance gap between resource-rich and resource-poor target languages. We propose novel data-efficient methods that enable quick and effective adaptation of pretrained multilingual models to such low-resource languages and unseen scripts.
  arXiv  Detail & Related papers  (2020-12-31T11:37:28Z)
• Cross-lingual Machine Reading Comprehension with Language Branch Knowledge Distillation [105.41167108465085]
  Cross-lingual Machine Reading (CLMRC) remains a challenging problem due to the lack of large-scale datasets in low-source languages. We propose a novel augmentation approach named Language Branch Machine Reading (LBMRC) LBMRC trains multiple machine reading comprehension (MRC) models proficient in individual language. We devise a multilingual distillation approach to amalgamate knowledge from multiple language branch models to a single model for all target languages.
  arXiv  Detail & Related papers  (2020-10-27T13:12:17Z)
• Explicit Alignment Objectives for Multilingual Bidirectional Encoders [111.65322283420805]
  We present a new method for learning multilingual encoders, AMBER (Aligned Multilingual Bi-directional EncodeR) AMBER is trained on additional parallel data using two explicit alignment objectives that align the multilingual representations at different granularities. Experimental results show that AMBER obtains gains of up to 1.1 average F1 score on sequence tagging and up to 27.3 average accuracy on retrieval over the XLMR-large model.
  arXiv  Detail & Related papers  (2020-10-15T18:34:13Z)
• WikiBERT models: deep transfer learning for many languages [1.3455090151301572]
  We introduce a simple, fully automated pipeline for creating languagespecific BERT models from Wikipedia data. We assess the merits of these models using the state-of-the-art UDify on Universal Dependencies data.
  arXiv  Detail & Related papers  (2020-06-02T11:57:53Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.