1 paper · 1 filter
Yash Datta, Sharath Rajasekar
We present JavelinGuard, a suite of low-cost, high-performance model architectures designed for detecting malicious intent in Large Language Model (LLM) interactions, optimized spe…