Unverified70% confidenceFactExact time
English content in mainstream large language model training corpora typically exceeds 50%, leading to significant performance gaps for non-English languages.
2
Sources
70%
Confidence
Long-term
Relevance
5/25/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedEnglish data accounts for an overwhelming proportion of mainstream large language model training corpora; taking public datasets such as Common Crawl as an example, English content typically constitutes 40%–60% of the total, while languages such as Vietnamese and Chinese often account for less than 5%.81% similarUnverifiedThe English pivot hypothesis stems from the extremely high proportion of English corpus in multilingual model pretraining, so translation between non-English language pairs may implicitly pass through English as an intermediary.74% similarUnverifiedMost mainstream large models are trained with English as the center, and support for morphologically complex languages such as German is often inadequate73% similarUnverifiedIn general-purpose corpora such as Common Crawl, English typically accounts for more than 45%, while Portuguese accounts for less than 2%.73% similarUnverifiedIndian languages account for a far smaller share of the training corpora of large language models than English, which leads to relatively weak safety guardrails for the model on languages such as Hindi.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/2166API
curl https://kongchang.com/api/v1/knowledge/claims/2166MCP
get_claim(id=2166)