SafeToolBench: pioneering a prospective benchmark to evaluating tool utilization safety in LLMs

Xia, Hongfei, Wang, Hongru, Liu, Zeming · Findings of the Association for Computational Linguistics: EMNLP 2025 · 2025