Unverified50% confidenceFactExact time
Several widely used multilingual models exhibited embedding space collapse, i.e., representations of different inputs became highly similar, making them unsuitable for similarity-based clustering tasks.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/27/2026
First Seen
Valid until: 1/25/2027
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.80% similarUnverifiedCross-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.77% similarUnverifiedThe language model component lacks the ability to explicitly model spatial relationships, causing the model to make errors easily in tasks requiring precise numerical reading and cross-element alignment.77% similarUnverifiedLarge language models show preliminary capabilities on certain Theory of Mind tasks, but still have significant limitations in complex multi-turn, multi-party scenarios.76% similarUnverifiedMultilingual large language models often give inconsistent answers to semantically identical questions in different languages, a phenomenon known as the Cross-Lingual Consistency (CLC) deficiency.75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1102142API
curl https://kongchang.com/api/v1/knowledge/claims/1102142MCP
get_claim(id=1102142)