Unverified50% confidenceBenchmarkExact time
On the SADA corpus, after replay-mix fine-tuning with roughly 90% Saudi dialect plus 7% English, WER on Najdi and Hijazi dialects dropped from 55.05% to 29.96%.
1
Sources
50%
Confidence
Long-term
Relevance
10/17/2026
First Seen
Sources
Related Entities
Related Claims
Unverified据估计互联网上超过97%的阿拉伯语文本不包含完整的变音符号标注68% similarUnverifiedFor Hindi, Korean and Swahili, the score drops brought about by cross-language transfer clearly exceed the fluctuations produced by English rewording.68% similarUnverifiedMainstream BPE tokenizers have far lower token efficiency for CJK languages than for English, and the same content may require 2 to 3 times as many tokens as in English.60% similarUnverifiedModels exhibit significant performance degradation on East Asian languages (such as Chinese, Japanese, and Korean) when faced with distorted statements, with cross-lingual performance gaps of up to 28 percentage points and relative drops as high as 49%60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1053534API
curl https://kongchang.com/api/v1/knowledge/claims/1053534MCP
get_claim(id=1053534)