[论文] Language on Demand, Knowledge at Core: Composing LLMs with Encoder-Dec...

## 论文概要 **研究领域**: NLP **作者**: Mengyu Bu, Yang Feng **发布时间**: 2025-03-18 **arXiv**: [2503.13831](https://arxiv.org/abs/2503.13831) ## 中文摘要 大型语言模型（LLMs）表现出强大的通用智能，但其多语言性能仍然高度不平衡。尽管LLMs在统一语义空间中编码了大量跨语言知识，但它们往往难以可靠地将这些知识与低资源或未见过的语言对接。幸运的是，预训练的编码器-解码器翻译模型已经具备平衡的多语言能力，为LLMs提供了自然的补充。在这项工作中，我们提出了XBridge，一种组合式编码器-LLM-解码器架构，将多语言理解和生成卸载到外部预训练翻译模型，同时保留LLM作为以英语为中心的核心进行通用知识处理。为解决跨模型产生的表示不对齐问题，我们引入了轻量级跨模型映射层和基于最优传输的对齐目标，实现多语言生成的细粒度语义一致性。在四种LLM上的多语言理解、推理、摘要和生成实验表明，XBridge优于强基线，特别是在低资源和以前未见过的语言上，且无需重新训练LLM。 ## 原文摘要 Large language models (LLMs) exhibit strong general intelligence, yet their multilingual performance remains highly imbalanced. Although LLMs encode substantial cross-lingual knowledge in a unified semantic space, they often struggle to reliably interface this knowledge with low-resource or unseen languages. Fortunately, pretrained encoder-decoder translation models already possess balanced multilingual capability, suggesting a natural complement to LLMs. In this work, we propose XBridge, a compositional encoder-LLM-decoder architecture that offloads multilingual understanding and generation to external pretrained translation models, while preserving the LLM as an English-centric core for general knowledge processing. To address the resulting representation misalignment across models, we i... --- *自动采集于 2026-03-19* #论文 #arXiv #NLP #小凯