面向混合专家模型中多语言下游任务的路由对齐微调
RA-MoE: Routing-Aligned Fine-Tuning for Multilingual Adaptation of Mixture-of-Experts Models
AI总结：
针对混合专家模型在多语言下游任务中的路由结构异构问题，提出RA-MoE三阶段框架，通过中间层语言通用对齐区识别任务相关专家，并引入路由对齐损失增强目标语言路由，实验表明该方法优于标准微调和强基线。
摘要：
混合专家（MoE）模型已成为高效扩展LLM的主流范式，但将其适配到非英语下游任务仍然具有挑战性。现有的微调方法将MoE模型视为整体学习器，忽略了预训练期间形成的异构路由结构。我们在多个MoE模型和下游任务上验证，中间层形成了语言通用对齐区，其中路由发散性强烈预测了每种语言的任务性能差距。基于这一观察，我们提出了RA-MoE（路由对齐MoE微调），一个三阶段框架，该框架根据英语和目标语言的正确性将并行任务示例分类为四路分类法（cc/ci/ic/ii），识别中间层中与任务相关的专家，并用路由对齐损失增强标准SFT，该损失鼓励ci类型示例上的目标语言路由遵循英语任务专家激活模式。在三个MoE模型、三个任务和六种目标语言上的实验表明，RA-MoE始终优于标准SFT和强基线（包括Routing Steering和RISE），其中任务-语言对的ci比例可作为对齐收益的可靠预测指标。
英文摘要：
Mixture-of-Experts (MoE) models enable efficient LLM scaling, yet adapting them to non-English downstream tasks remains challenging. Standard multilingual fine-tuning largely ignores their heterogeneous routing structure. Across multiple MoE models and tasks, we find strong cross-lingual routing alignment in middle layers, with routing divergence associated with target-language performance gaps. Motivated by this observation, we propose RA-MoE (Routing-Aligned MoE Fine-Tuning), a three-stage framework for multilingual MoE adaptation. RA-MoE categorizes parallel examples into four correctness groups (cc/ci/ic/ii) and identifies task-relevant experts in middle layers. It then selectively aligns target-language routing on ci examples toward successful English routing patterns, jointly matching the total routing mass assigned to task experts and its relative allocation among them. Experiments across three MoE models, three downstream tasks, and six target languages show that RA-MoE consistently outperforms standard SFT and strong routing-aware baselines. Further analyses confirm the intended routing changes and reveal that middle-layer task routing is largely shared and transferable across languages, providing mechanistic evidence for the cross-language transferability of task-specific routing.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
