Jev Code

A diagnosis-guided repair agent whose CWE classification comes from Jev. Jev beats six open-weight models on every classification and ranking metric and runs at 6.27x lower median latency than GPT-5.6-Sol; with Jev diagnosing, repair lifts the detector-measured security pass rate from 63.5% to 70.7%.

AgentJev-0.6B scores coding completion with a Qwen3-0.6B backbone stripped of its LM head: 57.8% accuracy (606/1048) on the coding set and 87.2% agreement with the teacher on invoice questions.

A CLI that asks Jev which files and declarations answer a repository question, then prints verbatim excerpts for a coding agent to use.