Unverified50% confidenceFactExact time
Cross-lingual transfer in multilingual models mainly relies on a shared multilingual representation space, which exhibits a significant alignment gap between language pairs with large differences in grammatical structure.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedCross-lingual jailbreaking exploits the asymmetry between a model's concept space and its language space, as the learning of safety-refusal behavior is highly language-specific.79% similarUnverified根据语言枢纽假说，英语等高资源语言在多语言模型内部表示中往往充当隐式中间语言，非英语低资源语言之间的迁移倾向于经由英语语义空间中转78% similarUnverifiedThe widespread use of large language models may lead to a sustained reduction in linguistic diversity, giving rise to what linguists call 'technology-mediated language convergence.'73% similarVerifiedModern multilingual large models establish language-agnostic semantic representations during the pre-training phase, building cross-lingual semantic alignment in high-dimensional vector space through joint training on massive multilingual corpora.70% similarUnverifiedCross-lingual transfer learning, which enables acoustic knowledge accumulated in high-resource languages to be transferred to low-resource languages, is the core mechanism by which a single model supports 83 languages while maintaining consistent audio quality.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/894734API
curl https://kongchang.com/api/v1/knowledge/claims/894734MCP
get_claim(id=894734)