Unverified50% confidenceFactExact time
General-purpose models trained primarily on English corpora often exhibit grammatical confusion or semantic drift in German's case inflection and gender system
1
Sources
50%
Confidence
Long-term
Relevance
10/4/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMost mainstream large models are trained with English as the center, and support for morphologically complex languages such as German is often inadequate79% 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 tokenization70% similarUnverified系统提示词的语言选择会影响输出风格，当系统提示词语言与用户输入语言不一致时会产生行为叠加效应66% 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 accent66% similarUnverifiedWriting scoring rules into a prompt in natural language has drawbacks: wording variations across different calls may cause the model's understanding to drift, and the scoring logic cannot be directly parsed or reused by programs.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/972206API
curl https://kongchang.com/api/v1/knowledge/claims/972206MCP
get_claim(id=972206)