arXiv cs.AIPaper
Reva: Real-Time Vulnerability Detection through Contextual Probing of Large Language Models
As coding agents ship code faster than security teams can review it, you need vulnerability detection that runs at agent speed, not human speed. REVA cracks this by capturing what makes agent reasoning reliable (context gathering, evidence weighing) and baking it into a fast detector. If you're deploying agent-generated code in production, this is the kind of system that sits between agent output and merge.