works on
code generation 2performance optimization 2benchmarking 1boundary taxonomy 1execution feedback 1gpu programming 1isolation 1LLM agents 1parallel computing 1program synthesis 1prompt injection 1reinforcement learning 1

From the 3 of 3 linked papers with an AI index.

collaborators
Showing cs.AIShow all

1 paper · 1 filter

cs.AI2026

Isolation as a First-Class Principle for LLM-Agent System Safety: Concepts, Taxonomy, Challenges and Future Directions

Huihao Jing, Wenbin Hu, Shaojin Chen +10

The paper surveys how isolating components such as user inputs, tools, execution, inter‑agent communication, and environment can improve safety of LLM‑agent systems, presenting a b…