发表机构
Beijing University of Posts and Telecommunications; State Key Laboratory of General Artificial Intelligence, BIGAI（北京邮电大学; 通用人工智能国家重点实验室，字节跳动公司）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
研究VLM驱动具身智能体的过程级安全问题，引入SAFERELBENCH基准测试，含507个样本。评估七个智能体发现任务成功与安全合规有差距，该基准明确测试行动前安全条件，凸显空间关系在安全评估中的核心地位。
AI 中文摘要
视觉语言模型（VLMs）越来越多地被用作具身智能体的推理主干，使机器人能够解释视觉场景、遵循语言指令并规划多步行动。在家庭环境中，安全不仅取决于识别物体，还取决于行动如何随时间改变物理场景。现有具身安全评估主要集中在静态风险识别、不安全指令拒绝或最终状态任务完成上。因此，由支撑、包含和接近等空间关系引起的过程级安全故障研究不足。为填补这一空白，我们引入了SAFERELBENCH，这是一个具有507个可执行评估样本的空间关系感知安全基准测试，包括248个空间关系样本和259个非空间控制样本。使用SAFERELBENCH评估七个开源和闭源VLM驱动的具身智能体，我们发现任务成功与过程级安全合规之间存在很大差距：模型经常在违反过程级安全约束的情况下完成请求的任务。与先前的基准测试不同，SAFERELBENCH明确测试智能体在易发生风险的行动之前是否满足安全条件，使空间关系成为具身安全评估的核心维度。更广泛地说，我们的结果表明，安全的具身智能不仅需要更强的感知和规划能力，还需要可靠地推理物体关系在交互过程中如何塑造风险。
英文摘要
Vision-language models (VLMs) are increasingly used as the reasoning backbone of embodied agents, enabling robots to interpret visual scenes, follow language instructions, and plan multi-step actions. In household environments, however, safety depends not only on recognizing objects, but also on how actions change the physical scene over time. Existing embodied safety evaluations largely focus on static risk recognition, unsafe instruction refusal, or final-state task completion. As a result, process-level safety failures induced by spatial relations such as support, containment, and proximity remain insufficiently studied. To address this gap, we introduce SAFERELBENCH, a spatial-relation-aware safety benchmark with 507 executable evaluation samples, including 248 spatial-relation samples and 259 non-spatial control samples. Using SAFERELBENCH to evaluate seven open- and closed-source VLM-driven embodied agents, we find a substantial gap between task success and process-level safety compliance: models often complete the requested task while violating process-level safety constraints. Unlike prior benchmarks, SAFERELBENCH explicitly tests whether agents satisfy safety conditions before risk-prone actions, making spatial relations a core dimension in embodied safety assessment. More broadly, our results show that safe embodied intelligence requires not only stronger perception and planning, but also reliable reasoning about how object relations shape risk during interaction.
CommentsPreprint. 10 pages, 6 figures, 4 tables
