Unverified50% confidenceFactExact time
通用ASR模型在呼叫中心场景中WER可能从通用场景的5%飙升至20-30%
1
Sources
50%
Confidence
Long-term
Relevance
8/5/2026
First Seen
Sources
Related Claims
Unverified在ASR场景中，即使单一维度的分布偏移就可能导致WER增加50%以上，多维度偏移会使模型性能断崖式下降70% similarUnverifiedThe solve rate of top models on SWE-bench jumped from under 5% in 2023 to over 30% in 2025.62% similarUnverifiedMid-tier models typically achieve more than 80% of flagship model scores on benchmarks such as MMLU and HumanEval, while requiring less than 30% of the inference cost.62% similarUnverifiedIn early 2023, mainstream models' SWE-bench scores were generally below 5%; by mid-2025, top models had surpassed 70%.60% 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60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/692313API
curl https://kongchang.com/api/v1/knowledge/claims/692313MCP
get_claim(id=692313)