HarnessRisk: A Lifecycle-Oriented Benchmark for Agent Harness Safety
HarnessRisk evaluates agent harness safety across six operational phases, revealing that configuration vulnerabilities and detection gaps allow high attack success despite preserve…
Hugging Face · Daily Papers
·Yajing Bai, Jinhao Duan
·
·▲ 7 upvotes
Este artigo está em destaque na seleção diária de papers do Hugging Face, curada pela comunidade de pesquisa em IA.
Autores: Yajing Bai, Jinhao Duan, Jie Peng, Xianfeng Wu, Sijia Liu, Song Wang
- 7 upvotes da comunidade
- Temas: agent harnesses, HarnessRisk, lifecycle benchmark, attack success rate, state persistence, action control
Resumo
Resumo original (em inglês), extraído do paper:
HarnessRisk evaluates agent harness safety across six operational phases, revealing that configuration vulnerabilities and detection gaps allow high attack success despite preserved utility.