1 paper · 1 filter
Sajjad Khan
The paper shows that several popular LLM‑agent frameworks do not correctly enforce their advertised control primitives (approval, cancellation, timeout), identifies multiple enforc…