What it does
Organizes independent Jev studies by calibration, consistency, prompt injection, abstention, and other failure modes.
Benchmarks & research
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
Organizes independent Jev studies by calibration, consistency, prompt injection, abstention, and other failure modes.
Before building an app with Jev, read through this curated list to see how the model behaves. Check how small changes in wording, option order, or tricky inputs make its answers shift. This helps you identify common failure points before writing your own software.
If your developer tests Jev on your project, you can contribute your findings back to the collection. Write a short summary of what you measured, how many examples you tested, and the main result, then have your developer submit that update to the public list.