Jev and CodeGraph for security review
The typesafe-security-review project combines Jev’s probability judgments with a code graph and OWASP/CWE data for low-cost security scanning.
Jev Model Benchmarked on Phishing Email Subject Lines
Jev (jev-1.13.0), Clef, and Clef-flash were tested on phishing detection using only email subject lines (60 scam subjects vs. 79 official site titles). Direct questioning was ineffective; combining feature-based responses yielded AUC 0.90–0.93, but 15–25% were missed. Genuine emails were not tested.
詐欺メールの件名だけで AI は見分けられるか。Jev（jev-1.13.0）・Clef・Clef-flash に件名のみを入力し、詐欺の件名60件と公式サイトの題名79件で実測。直接「詐欺か」と聞くのは不向き。特徴を聞いて合算すると AUC 0.90〜0.93、ただし15〜25%は見逃す。本物のメールは未検証。
· 0 likesOpen on X
More like this
Jev Delivers 5x Cheaper and More Accurate Security Pipeline Results
A user shares experience using Jev in a security pipeline: over 5x cheaper, faster, and more accurate than smaller, dumber models.
Jev Use Cases and Demo in AI Security
Introduces Jev use cases and a demo in AI security, leveraging its choice-and-probability model with Unity AI Gateway and MLflow integration for safety checks.
