作者位次 #2/7ACL 2026Findings· Findings of ACL
T ool S afe: Enhancing Tool Invocation Safety of LLM -based agents via Proactive Step-level Guardrail and Feedback
Yutao Mou · Zhangchi Xue · Lijun Li · Peiyang Liu · Shikun Zhang · Wei Ye · Jing Shao
摘要
While LLM-based agents can interact with environments via invoking external tools, their expanded capabilities also amplify security risks. Monitoring step-level tool invocation behaviors in real time and proactively intervening before unsafe execution is critical for agent deployment, yet remain...