研究人员开发了一种名为漂移增强评分（Drift-Augmented Scoring, DAS）的新技术，以提高零样本音频-语言分类模型在声学噪声下的鲁棒性。该方法在余弦评分上增加了一个小额奖励，当噪声音频嵌入朝着文本提示预测的方向漂移时，会奖励相应的类别。DAS在UrbanSound8K和FSD50K等基准数据集上展示了显著的改进，在各种噪声条件下提高了准确率和mAP得分。 AI
影响 增强了音频AI系统在嘈杂环境中的可靠性，可能改进语音助手和内容审核等应用。
排序理由 该集群包含一篇详细介绍音频-语言分类新方法的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2606.04844v1 Announce Type: cross Abstract: Contrastive audio-language models such as CLAP enable zero-shot audio classification: a sound is labelled by matching its embedding to text prompt embeddings, with no labelled audio. This matching breaks down under acoustic noise,…
Contrastive audio-language models such as CLAP enable zero-shot audio classification: a sound is labelled by matching its embedding to text prompt embeddings, with no labelled audio. This matching breaks down under acoustic noise, where accuracy and mAP fall by 12-30 percentage p…