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 evaluation on BANKING77 and adversarial injection analysis
The referenced blog reports Jev zero-shot accuracy of 81.2% on BANKING77, below MiniLM+linear head's 89.9%; injected text moved prediction probability from 0.76 to 0.48, suggesting a potential security risk.
⚠️ Caveats: • Vendor evals score agreement with GPT-6 Astra + Fable 5.1, not ground truth • BANKING77 (@manojlds): MiniLM + linear head 89.9%, Jev zero-shot 81.2% • Injected text moved a block prob 0.76 → 0.48 https://stacktoheap.com/blog/2026/09/22/jev-attack-of-the-classifiers/…
· 1 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.
