jev-spam-eval
用 Noul 问题做零样本垃圾邮件过滤，对照 TF-IDF 基线。
本页由英文清单自动生成。
来自 readme
Zero-shot email classification with TypeSafe’s Jev TypeSafe’s Jev reached 98.64% accuracy on a 5,733-email ham/spam/phishing test using written category definitions and email context, without task-specific fine-tuning or labeled examples in its requests. A TF-IDF logistic regression classifier trained on roughly 4,600 labeled messages per fold reached …
这是什么
- 形态
- 数据集或基准
- 宿主智能体
- 独立运行
- 面向人群
- 研究人员
成熟度
文档
报告了实测数据