When English Isn't the Best Teacher: Source Language Effects in Cross-Lingual In-Context Learning
arXiv:2606.18033v1 Announce Type: new Abstract: Cross-lingual transfer in multilingual NLP has been widely explored in supervised fine-tuning contexts, where factors like data availability and linguistic similarity largely determine transfer quality. As the field shifts toward few-shot In-Context Learning (ICL), it is often presumed that insights from fine-tuning carry over unchanged. Yet this assumption has not been rigorously evaluated, leaving open the question of how to choose source languages for cross-lingual ICL. We conduct a broad empirical study of cross-lingual transfer in ICL spanni
The rapid development and deployment of multilingual large language models make understanding optimal cross-lingual transfer critical for model performance and efficiency.
This research highlights that presumed best practices in cross-lingual AI transfer, derived from fine-tuning, may not apply to In-Context Learning, impacting multilingual AI development strategies.
The understanding of how to effectively select source languages for cross-lingual In-Context Learning will shift, leading to more optimized model training and deployment.
- · AI researchers in multilingual NLP
- · Developers of multilingual AI applications
- · Users of diverse language AI tools
- · AI models relying solely on English-centric transfer assumptions
- · Companies with suboptimal multilingual AI strategies
Improved performance and efficiency of multilingual AI systems through better source language selection.
Reduced computational costs and increased accessibility of advanced AI capabilities to a wider range of languages.
Enhanced global adoption of AI as linguistic barriers are more efficiently addressed, potentially fostering more equitable technological development.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL