HomeGuard: 基于视觉-语言模型的具身安全防护系统用于识别家庭任务中的上下文风险
HomeGuard: VLM-based Embodied Safeguard for Identifying Contextual Risk in Household Task
AI总结：
本文提出HomeGuard系统，通过Context-Guided Chain-of-Thought机制，结合定制数据集和强化微调策略，提升家庭任务中风险识别精度，减少误判，同时为后续规划提供空间约束。
摘要：
视觉-语言模型（VLMs）使具身代理能够执行复杂指令，但仍然容易受到上下文安全风险的影响，即良性命令因微妙的环境状态而变得危险。现有安全措施往往不足。基于规则的方法在物体密集场景中缺乏可扩展性，而基于模型的方法依赖于提示工程，导致感知不集中，从而遗漏风险或产生幻觉。为此，我们提出了一种架构无关的安全防护机制，即上下文引导的推理链（CG-CoT）。该机制将风险评估分解为主动感知，即按顺序将注意力锚定在交互目标和相关空间邻域，随后基于此视觉证据进行语义判断。我们通过定制的 grounding 数据集和两阶段训练策略支持该方法，利用强化微调（RFT）和过程奖励来强制精确的中间锚定。实验表明，我们的模型HomeGuard显著提高了安全性，相比基础模型，风险匹配率提高了超过30%，同时减少了过度安全。除了危险检测，生成的视觉锚点还为后续规划者提供可操作的空间约束，促进显式碰撞避免和安全轨迹生成。代码和数据已发布在https://github.com/AI45Lab/HomeGuard
英文摘要：
Vision-Language Models (VLMs) empower embodied agents to execute complex instructions, yet they remain vulnerable to contextual safety risks where benign commands become hazardous due to subtle environmental states. Existing safeguards often prove inadequate. Rule-based methods lack scalability in object-dense scenes, whereas model-based approaches relying on prompt engineering suffer from unfocused perception, resulting in missed risks or hallucinations. To address this, we propose an architecture-agnostic safeguard featuring Context-Guided Chain-of-Thought (CG-CoT). This mechanism decomposes risk assessment into active perception that sequentially anchors attention to interaction targets and relevant spatial neighborhoods, followed by semantic judgment based on this visual evidence. We support this approach with a curated grounding dataset and a two-stage training strategy utilizing Reinforcement Fine-Tuning (RFT) with process rewards to enforce precise intermediate grounding. Experiments demonstrate that our model HomeGuard significantly enhances safety, improving risk match rates by over 30% compared to base models while reducing oversafety. Beyond hazard detection, the generated visual anchors serve as actionable spatial constraints for downstream planners, facilitating explicit collision avoidance and safety trajectory generation. Code and data are released under https://github.com/AI45Lab/HomeGuard
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
