Large language models exhibit strong multilingual capabilities, yet significant performance gaps persist between dominant and non-dominant languages. Prior work attributes this gap to imbalances between shared and language-specific neurons in multilingual representations. We propose Cross-Lingual Activation Steering (CLAS), a training-free inference-time intervention that selectively modulates neuron activations. We evaluate CLAS on classification and generation benchmarks, achieving average improvements of 2.3% (Acc.) and 3.4% (F1) respectively, while maintaining high-resource language performance. We discover that effective transfer operates through functional divergence rather than strict alignment; performance gains correlate with increased language cluster separation. Our results demonstrate that targeted activation steering can unlock latent multilingual capacity in existing models without modification to model weights.

翻译：大型语言模型展现出强大的多语言能力，但在主导语言与非主导语言之间仍存在显著的性能差距。先前研究将此差距归因于多语言表征中共享神经元与语言特定神经元之间的不平衡。我们提出跨语言激活导向，一种无需训练、在推理时选择性调节神经元激活的干预方法。我们在分类与生成基准测试上评估了该方法，分别实现了平均2.3%（准确率）与3.4%（F1分数）的性能提升，同时保持了高资源语言的性能。我们发现，有效的跨语言迁移通过功能分化而非严格对齐实现；性能提升与语言簇分离度的增加相关。我们的结果表明，定向的激活导向能够在无需修改模型权重的情况下，释放现有模型中潜在的多语言能力。