DiM\textsuperscript{3}: Bridging Multilingual and Multimodal Models via Direction- and Magnitude-Aware Merging

• URL: http://arxiv.org/abs/2605.12960v2
• Date: Wed, 20 May 2026 10:12:11 GMT
• Title: DiM\textsuperscript{3}: Bridging Multilingual and Multimodal Models via Direction- and Magnitude-Aware Merging
• Abstract summary: We propose Direction- and Magnitude-aware Multimodal merging (DiM3)<n>Experiments on multilingual benchmarks in both text-only and vision-language settings, covering 57 languages across LLaVA- and Qwen-based backbones.
• Score: 60.709970092170074
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Towards more general and human-like intelligence, large language models should seamlessly integrate both multilingual and multimodal capabilities; however, extending an existing multimodal model to many languages typically requires expensive multilingual multimodal data construction and repeated end-to-end retraining. We study a training-free alternative: injecting multilingual capability into an existing multimodal model by composing residual updates in the shared language model backbone. The key challenge is that multilingual and multimodal updates are heterogeneous, reflecting different functional roles in the shared model. To address this, we propose Direction- and Magnitude-aware Multilingual Multimodal merging (DiM3), which selectively composes the two updates at each parameter dimension while preserving the original vision encoder and multimodal projector. Experiments on multilingual benchmarks in both text-only and vision-language settings, covering 57 languages across LLaVA- and Qwen-based backbones, show that DiM3 consistently outperforms existing merging baselines, substantially improves multilingual performance over the original multimodal model, and remains competitive with dedicated multilingual multimodal fine-tuning while largely retaining general multimodal ability. We further show that DiM3 can be directly applied to already trained multilingual multimodal models and still yield additional gains. Further interpretability analysis shows that DiM3 primarily reshapes intermediate-layer semantic representations, strengthening cross-lingual alignment under both text-only and multimodal inputs while preserving higher-layer task-sensitive structure. Our repository is on https://github.com/wzj1718/DiM3.

Related papers

• M3DR: Towards Universal Multilingual Multimodal Document Retrieval [0.0]
  We present M3DR (Multilingual Multimodal Document Retrieval), a framework designed to bridge this gap across languages.<n>We generalize across different vision-language architectures and model sizes, enabling robust cross-lingual and cross-modal alignment.<n>Our models, NetraEmbed and ColNetraEmbed achieve state-of-the-art performance with 150% relative improvements on cross-lingual retrieval.
  arXiv  Detail & Related papers  (2025-12-03T07:17:59Z)
• CACARA: Cross-Modal Alignment Leveraging a Text-Centric Approach for Cost-Effective Multimodal and Multilingual Learning [6.162206820356373]
  We propose a multimodal and multilingual architecture, CACARA, trained through emergent alignment learning.<n>By fine-tuning the newly incorporated modality only on data aligned with the English language, our model develops support for over 100 languages.<n>Our strategy achieves up to a 14.24 percentage points improvement in R@1 audio-to-text retrieval, outperforming state-of-the-art multimodal models.
  arXiv  Detail & Related papers  (2025-11-29T14:04:27Z)
• LUSIFER: Language Universal Space Integration for Enhanced Multilingual Embeddings with Large Language Models [89.13128402847943]
  We present LUSIFER, a novel zero-shot approach that adapts LLM-based embedding models for multilingual tasks without requiring multilingual supervision.<n>LUSIFER's architecture combines a multilingual encoder, serving as a language-universal learner, with an LLM-based embedding model optimized for embedding-specific tasks.<n>We introduce a new benchmark encompassing 5 primary embedding tasks, 123 diverse datasets, and coverage across 14 languages.
  arXiv  Detail & Related papers  (2025-01-01T15:43:07Z)
• U3M: Unbiased Multiscale Modal Fusion Model for Multimodal Semantic Segmentation [63.31007867379312]
  We introduce U3M: An Unbiased Multiscale Modal Fusion Model for Multimodal Semantics. We employ feature fusion at multiple scales to ensure the effective extraction and integration of both global and local features. Experimental results demonstrate that our approach achieves superior performance across multiple datasets.
  arXiv  Detail & Related papers  (2024-05-24T08:58:48Z)
• AnyGPT: Unified Multimodal LLM with Discrete Sequence Modeling [115.56746545958522]
  We introduce AnyGPT, an any-to-any multimodal language model that utilizes discrete representations for the unified processing of various modalities.<n>We build a multimodal text-centric dataset for multimodal alignment pre-training.<n>We show that AnyGPT is capable of facilitating any-to-any multimodal conversation while achieving performance comparable to specialized models across all modalities.
  arXiv  Detail & Related papers  (2024-02-19T15:33:10Z)
• Large Scale Multi-Lingual Multi-Modal Summarization Dataset [26.92121230628835]
  We present the current largest multi-lingual multi-modal summarization dataset (M3LS) It consists of over a million instances of document-image pairs along with a professionally annotated multi-modal summary for each pair. It is also the largest summarization dataset for 13 languages and consists of cross-lingual summarization data for 2 languages.
  arXiv  Detail & Related papers  (2023-02-13T18:00:23Z)
• xGQA: Cross-Lingual Visual Question Answering [100.35229218735938]
  xGQA is a new multilingual evaluation benchmark for the visual question answering task. We extend the established English GQA dataset to 7 typologically diverse languages. We propose new adapter-based approaches to adapt multimodal transformer-based models to become multilingual.
  arXiv  Detail & Related papers  (2021-09-13T15:58:21Z)
• M3P: Learning Universal Representations via Multitask Multilingual Multimodal Pre-training [119.16007395162431]
  M3P is a Multilingual Multimodal Pre-trained model that combines multilingual pre-training and multimodal pre-training. We show that M3P can achieve comparable results for English and new state-of-the-art results for non-English languages.
  arXiv  Detail & Related papers  (2020-06-04T03:54:29Z)

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.