发表机构
Korea Electronics Technology Institute (KETI)（韩国电子技术研究院（KETI））
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
SAGE提出符号动作门控与局部编辑机制，以零token安全监控和高效恢复提升LLM任务规划器的安全性与完成率，并在多个基准上验证了其优势。
AI 中文摘要
大型语言模型（LLM）现已成为具身家庭智能体的默认认知核心，然而它们生成的计划在执行前很少依据环境的基础模型进行检查，而且它们报告的任务成功率往往是在高度饱和的基准上测得的，以至于任何方法都无法与其他方法区分开来。我们提出SAGE（符号动作门控与编辑），这是一个由两种轻量机制构建的单LLM规划器：一个与领域无关的符号门（约250行Python代码，零token，复杂度为O(|π|)），作为运行时安全监控器，以类型化原因阻止违反前提条件的动作；以及一个局部编辑机制，仅重新生成失败子目标的后缀，保留已完成和未受影响的工作；混合的种子加实时记忆存储支持冷启动覆盖。我们在无泄漏协议（留一法检索）下，对五个开放权重模型和一个75任务AI2-THOR基准进行了评估。在标准基准上，目标完成率趋于饱和（52%的实例被平凡解决），SAGE与强分层基线持平。在更困难、与方法无关的多目标组合上，SAGE的完成率优势重新显著出现（四个模型上+0.06至+0.23）。在注入的中途执行失败下，SAGE以2.4至3.3倍更少的LLM调用，与全计划重规划器一样可靠地恢复。作为执行前验证门，符号监控器在执行前阻止不安全动作，并为每个测试的规划器提高模拟器报告的分步成功率（最高+0.11），这是验证器从未见过的信号（非循环）。由于该门不调用任何模型（0.008毫秒/计划），它是一个在边缘上几乎免费运行的安全层：SAGE规划在Jetson AGX Orin上复现了其质量，其中小模型验证帮助最大。我们发布了基准、无泄漏协议、恢复和安全门工具集，以及验证器可移植性研究（在ALFWorld上自动诱导，保留0.89）。
英文摘要
Large language models (LLMs) are now the default cognitive core of embodied household agents, yet the plans they emit are rarely checked against a grounded model of the environment before execution, and the task-success they report is often measured on benchmarks so saturated that no method can be separated from another. We present SAGE (Symbolic Action-Gating and Editing), a single-LLM planner built from two lightweight mechanisms: a domain-agnostic symbolic gate (~250 lines of Python, zero tokens, $O(|π|)$) that blocks precondition-violating actions with typed reasons as a runtime safety monitor, and a local edit that regenerates only the failed sub-goal's suffix, keeping completed and untouched work intact; a hybrid seed+live memory store supports cold-start coverage. We evaluate under a leak-free protocol (leave-one-out retrieval) over five open-weight models and a 75-task AI2-THOR benchmark. On the standard benchmark goal-completeness saturates (52% of instances trivially solved) and SAGE ties strong hierarchical baselines. On a harder, method-agnostic multi-goal composition, SAGE's completeness lead re-emerges large (+0.06 to +0.23 across four models). Under injected mid-execution failures, SAGE recovers as reliably as whole-plan replanners at 2.4-3.3x fewer LLM calls. As a verify-before-execute gate, the symbolic monitor blocks unsafe actions before actuation and raises simulator-reported step-success for every planner tested (up to +0.11), a signal the verifier never sees (non-circular). Because the gate calls no model (0.008 ms/plan), it is a safety layer that runs essentially free on the edge: SAGE planning reproduces its quality on a Jetson AGX Orin, where small-model verification helps most. We release the benchmark, the leak-free protocol, the recovery and safety-gate harnesses, and a verifier-portability study (auto-induced on ALFWorld, 0.89 held-out).
Comments8 pages, 2 figures, 1 table, 3 algorithms. Submitted to IEEE Robotics and Automation Letters (RA-L). Code and benchmark: https://github.com/mtbui2010/sage_release
