Unverified50% confidenceFactExact time
Google在2018年发布的研究发现其ASR系统对非裔美国人英语(AAE)的错误率比对标准美国英语高出约一倍
1
Sources
50%
Confidence
Long-term
Relevance
8/11/2026
First Seen
Sources
Related Claims
UnverifiedResearch by the Columbia Journalism Review (CJR) shows that AI search tools gave incorrect answers to more than 60% of queries.53% similarUnverified当前主流 ASR 引擎的通用词错率（WER）已降至5%以下，但在医疗垂直领域的专有名词上错误率可能飙升至15%-30%，除非进行领域微调和自定义词典配置53% similarUnverifiedGoogle 内部统计数据表明，E2E 测试与单元测试的数量比例建议维持在 1:10 乃至更低52% similarUnverifiedTop large language models have generally surpassed 90% on MMLU, causing the test's discriminative power to decline.51% similarUnverifiedGoogle内部研究显示，当工具误报率超过10%时，开发者的采纳率会急剧下降50% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/731901API
curl https://kongchang.com/api/v1/knowledge/claims/731901MCP
get_claim(id=731901)