Cross-Lingual Steering for Figurative Language Generation
This research provides direct evidence of a reusable, yet target-dependent, cross-lingual signal for figurative language generation, which is significant for researchers working on multilingual large language models.
This paper investigates whether internal signals for figurative language generation in multilingual LLMs are language-specific or reusable. By estimating figurative category directions from activation differences in one language and applying them during generation, the authors found that these directions reliably steer within their own language and, more importantly, transfer across languages, with German being a particularly receptive target. Directions assembled from other languages can even surpass a target language's native direction.
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. 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. Across five figurative categories, six languages, and four multilingual LLMs, these directions steer reliably within their own language, most robustly for metaphor and simile. 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. 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. Together, these results provide direct evidence of a reusable but target-dependent cross-lingual signal for figurative generation.