ArmBench-ASR：亚美尼亚语语音识别基准发布

AI 摘要

Hugging Face 发布了 ArmBench-ASR v0.1，一个用于评估亚美尼亚语自动语音识别（ASR）模型的基准，包含近30个模型和约20.7小时的音频。Gemini 2.5 Pro 以14.31%的严格WER排名第一，而最强的开源模型是 NVIDIA 的 Armenian FastConformer，排名第9。基准显示模型性能高度依赖领域，电影音频最难，且归一化显著影响错误率。

正文节选

This initial v0.1 release evaluates almost 30 models, both open-weight and closed. The evaluation suite consists of 10,113 clips totaling approximately 20.7 hours of audio. Top result: Gemini 2.5 Pro ranks first by strict combined WER, with 14.31% WER and 6.42% normalized WER. Armenian remains underrepresented in speech technology. A model that performs well on a standard read-speech test set may still struggle with the variability of real-world speech. A single aggregate score cannot expose the

发布时间：—

抓取时间：2026-08-25 15:03

来源机构：Hugging Face