Unverified50% confidenceOpinionExact time
Most mainstream large models are trained with English as the center, and support for morphologically complex languages such as German is often inadequate
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/4/2026
First Seen
Valid until: 1/2/2027
Sources
Related Entities
Related Claims
UnverifiedGeneral-purpose models trained primarily on English corpora often exhibit grammatical confusion or semantic drift in German's case inflection and gender system79% similarUnverifiedEnglish content in mainstream large language model training corpora typically exceeds 50%, leading to significant performance gaps for non-English languages.73% similarUnverifiedThe training data of current mainstream large models contains a relatively high proportion of English corpus, which may cause the model to favor English prosodic patterns in syllable boundary processing and pause rhythm, producing a foreign accent72% similarUnverifiedGerman is a highly inflected-synthetic language whose compound word formation mechanism causes the vocabulary to expand sharply, posing challenges for large models based on subword tokenization72% similarUnverified通用大模型在特定语言的方言处理、形态变化丰富的语言以及文化特定表达上仍存在明显不足72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/972212API
curl https://kongchang.com/api/v1/knowledge/claims/972212MCP
get_claim(id=972212)