SafeArena: Evaluating the Safety of Autonomous Web Agents
원문 초록
LLM-based agents are becoming increasingly proficient at solving web-based tasks. With this capability comes a greater risk of misuse for malicious purposes, such as posting misinformation in an online forum or selling illicit substances on a website. To evaluate these risks, we propose SafeArena, the first benchmark to focus on the deliberate misuse of web agents. SafeArena comprises 250 safe and 250 harmful tasks across four websites. We classify the harmful tasks into five harm categories -- misinformation, illegal activity, harassment, cybercrime, and social bias, designed to assess realistic misuses of web agents. We evaluate leading LLM-based web agents, including GPT-4o, Claude-3.5 Sonnet, Qwen-2-VL 72B, and Llama-3.2 90B, on our benchmark. To systematically assess their susceptibility to harmful tasks, we introduce the Agent Risk Assessment framework that categorizes agent behavior across four risk levels. We find agents are surprisingly compliant with malicious requests, with GPT-4o and Qwen-2 completing 34.7% and 27.3% of harmful requests, respectively. Our findings highlight the urgent need for safety alignment procedures for web agents. Our benchmark is available here: https://safearena.github.io
함께 살펴볼 논문
공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.
Safe RLHF-V: Safe Reinforcement Learning from Human Feedback in Multimodal Large Language Models
The Devil behind the mask: An emergent safety vulnerability of Diffusion LLMs
Remember Past, Anticipate Future: Learning Continual Multimodal Misinformation Detectors
Robustness of Misinformation Classification Systems to Adversarial Examples Through BeamAttack
Persona-Assigned Large Language Models Exhibit Human-Like Motivated Reasoning
Safe Pruning LoRA: Robust Distance-Guided Pruning for Safety Alignment in Adaptation of LLMs
토론
이 논문과 연결된 포럼 글이 아직 없습니다.