Group then Scale: Dynamic Mixture-of-Experts Multilingual Language Model

• URL: http://arxiv.org/abs/2506.12388v1
• Date: Sat, 14 Jun 2025 07:56:18 GMT
• Title: Group then Scale: Dynamic Mixture-of-Experts Multilingual Language Model
• Authors: Chong Li, Yingzhuo Deng, Jiajun Zhang, Chengqing Zong,
• Abstract summary: Experimental results on 18 to 128 languages show that our method reduces the negative transfer between languages and significantly boosts multilingual performance with fewer parameters.<n>Such language group specialization on experts benefits the new language adaptation and reduces the inference on the previous multilingual knowledge learned.
• Score: 38.0723521889505
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The curse of multilinguality phenomenon is a fundamental problem of multilingual Large Language Models (LLMs), where the competition between massive languages results in inferior performance. It mainly comes from limited capacity and negative transfer between dissimilar languages. To address this issue, we propose a method to dynamically group and scale up the parameters of multilingual LLM while boosting positive transfer among similar languages. Specifically, the model is first tuned on monolingual corpus to determine the parameter deviation in each layer and quantify the similarity between languages. Layers with more deviations are extended to mixture-of-experts layers to reduce competition between languages, where one expert module serves one group of similar languages. Experimental results on 18 to 128 languages show that our method reduces the negative transfer between languages and significantly boosts multilingual performance with fewer parameters. Such language group specialization on experts benefits the new language adaptation and reduces the inference on the previous multilingual knowledge learned.

Related papers

• Modular Sentence Encoders: Separating Language Specialization from Cross-Lingual Alignment [50.80949663719335]
  multilingual sentence encoders (MSEs) are commonly obtained by training multilingual language models to map sentences from different languages into a shared semantic space.<n>MSEs are subject to curse of multilinguality, a loss of monolingual representational accuracy due to parameter sharing.<n>We train the cross-lingual adapters with two different types of data to resolve the conflicting requirements of different cross-lingual tasks.
  arXiv  Detail & Related papers  (2024-07-20T13:56:39Z)
• Understanding and Mitigating Language Confusion in LLMs [76.96033035093204]
  We evaluate 15 typologically diverse languages with existing and newly-created English and multilingual prompts.<n>We find that Llama Instruct and Mistral models exhibit high degrees of language confusion.<n>We find that language confusion can be partially mitigated via few-shot prompting, multilingual SFT and preference tuning.
  arXiv  Detail & Related papers  (2024-06-28T17:03:51Z)
• Multilingual Instruction Tuning With Just a Pinch of Multilinguality [31.360147312195068]
  We show that many languages transfer some instruction-following capabilities to other languages from even monolingual tuning. We observe that models tuned on multilingual mixtures exhibit comparable or superior performance in multiple languages. diversifying the instruction tuning set with even just 2-4 languages significantly improves cross-lingual generalization.
  arXiv  Detail & Related papers  (2024-01-03T17:48:10Z)
• GradSim: Gradient-Based Language Grouping for Effective Multilingual Training [13.730907708289331]
  We propose GradSim, a language grouping method based on gradient similarity. Our experiments on three diverse multilingual benchmark datasets show that it leads to the largest performance gains. Besides linguistic features, the topics of the datasets play an important role for language grouping.
  arXiv  Detail & Related papers  (2023-10-23T18:13:37Z)
• Multilingual Entity and Relation Extraction from Unified to Language-specific Training [29.778332361215636]
  Existing approaches for entity and relation extraction tasks mainly focus on the English corpora and ignore other languages. We propose a two-stage multilingual training method and a joint model called Multilingual Entity and Relation Extraction framework (mERE) to mitigate language interference. Our method outperforms both the monolingual and multilingual baseline methods.
  arXiv  Detail & Related papers  (2023-01-11T12:26:53Z)
• Lifting the Curse of Multilinguality by Pre-training Modular Transformers [72.46919537293068]
  multilingual pre-trained models suffer from the curse of multilinguality, which causes per-language performance to drop as they cover more languages. We introduce language-specific modules, which allows us to grow the total capacity of the model, while keeping the total number of trainable parameters per language constant. Our approach enables adding languages post-hoc with no measurable drop in performance, no longer limiting the model usage to the set of pre-trained languages.
  arXiv  Detail & Related papers  (2022-05-12T17:59:56Z)
• When is BERT Multilingual? Isolating Crucial Ingredients for Cross-lingual Transfer [15.578267998149743]
  We show that the absence of sub-word overlap significantly affects zero-shot transfer when languages differ in their word order. There is a strong correlation between transfer performance and word embedding alignment between languages. Our results call for focus in multilingual models on explicitly improving word embedding alignment between languages.
  arXiv  Detail & Related papers  (2021-10-27T21:25:39Z)
• Discovering Representation Sprachbund For Multilingual Pre-Training [139.05668687865688]
  We generate language representation from multilingual pre-trained models and conduct linguistic analysis. We cluster all the target languages into multiple groups and name each group as a representation sprachbund. Experiments are conducted on cross-lingual benchmarks and significant improvements are achieved compared to strong baselines.
  arXiv  Detail & Related papers  (2021-09-01T09:32:06Z)
• How Good is Your Tokenizer? On the Monolingual Performance of Multilingual Language Models [96.32118305166412]
  We study a set of nine typologically diverse languages with readily available pretrained monolingual models on a set of five diverse monolingual downstream tasks. We find that languages which are adequately represented in the multilingual model's vocabulary exhibit negligible performance decreases over their monolingual counterparts.
  arXiv  Detail & Related papers  (2020-12-31T14:11:00Z)
• On Negative Interference in Multilingual Models: Findings and A Meta-Learning Treatment [59.995385574274785]
  We show that, contrary to previous belief, negative interference also impacts low-resource languages. We present a meta-learning algorithm that obtains better cross-lingual transferability and alleviates negative interference.
  arXiv  Detail & Related papers  (2020-10-06T20:48:58Z)

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.