Cross-Lingual Steering for Figurative Language Generation

跨语言引导：比喻性语言生成研究

Abstract: Multilingual large language models can generate figurative language, but whether the internal signals driving this behavior are language-specific or reusable across languages is unclear. 摘要： 多语言大语言模型能够生成比喻性语言，但驱动这种行为的内部信号究竟是特定于某种语言的，还是可以在不同语言间复用的，目前尚不明确。

Using activation steering as a probe, we estimate a direction for a figurative category from figurative—literal activation differences in one language and apply it during generation. 我们利用激活引导（activation steering）作为探测手段，通过分析一种语言中比喻性与字面性表达之间的激活差异，估算出该比喻类别的引导方向，并将其应用于生成过程中。

Across five figurative categories, six languages, and four multilingual LLMs, these directions steer reliably within their own language, most robustly for metaphor and simile. 在涵盖五个比喻类别、六种语言以及四个多语言大语言模型的实验中，这些引导方向在各自语言内均能实现可靠的控制，其中对比喻（metaphor）和明喻（simile）的引导效果最为稳健。

More importantly, they transfer across languages: a direction learned in one increases the target behavior when applied to another, with German among the most receptive targets. 更重要的是，这些引导方向具有跨语言迁移能力：在一个语言中学习到的方向，应用于另一种语言时能增强相应的目标行为，其中德语表现出极高的接受度。

Going further, directions assembled from other languages can match or even surpass a target language’s own native direction, while removing this shared component weakens native steering. 进一步研究发现，由其他语言组合而成的引导方向可以达到甚至超过目标语言自身的原生引导效果，而移除这些共享成分则会削弱原生引导的效力。

Together, these results provide direct evidence of a reusable but target-dependent cross-lingual signal for figurative generation. 综上所述，这些结果为比喻生成中存在一种“可复用但依赖于目标语言”的跨语言信号提供了直接证据。