发表机构
Carnegie Mellon University; Tulane University; New York University; Princeton University; Columbia University（卡内基梅隆大学; 杜兰大学; 纽约大学; 普林斯顿大学; 哥伦比亚大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
提出SimpleARM，一种免训练记忆层，通过监控、感知工具和结构化检索为冻结通用策略提供任务相关状态，在RoboMME基准上以67.17%成功率超越基线44.51%。
AI 中文摘要
视觉记忆系统通常保留或压缩过去的观察。机器人控制额外需要由交互衍生的状态，这些状态可能无法由任何单个帧显式表示，例如持久的身份关系、累积的进展或有序的程序。我们引入了简单智能体机器人记忆（SimpleARM），一种用于冻结的通用机器人策略的免训练记忆层。根据任务指令，SimpleARM指定要监控的内容；冻结的感知工具在线维护紧凑的带类型状态；结构化访问仅在提议的子目标依赖于历史时才检索该状态；当前视图接地在执行前解析回忆的实体。我们在RoboMME上评估SimpleARM，这是一个记忆依赖的机器人操作任务基准，需要当前观察中不再可用的历史信息。在所有16个任务和三个策略种子中，SimpleARM实现了67.17%的平均成功率，而最强的非预言机基线为44.51%。匹配的消融研究显示了机制的特异性：移除关系、参考、进展或路线状态会在受影响状态被检索用于控制的任务中产生较大损失，而在其他任务中则基本不受影响。这些结果支持基于状态的机器人记忆观点：用于控制的有效记忆不仅仅是保留的视觉历史，而是从交互历史中衍生的紧凑任务相关状态。
英文摘要
Visual-memory systems commonly retain or compress past observations. Robot control additionally requires interaction-derived state that no individual frame may explicitly represent, such as persistent identity relations, accumulated progress, or ordered procedures. We introduce Simple Agentic Robot Memory (SimpleARM), a training-free memory layer for frozen generalist robot policies. From the task instruction, SimpleARM specifies what to monitor; frozen perceptual tools maintain compact typed state online; structured access retrieves that state only when a proposed subgoal depends on history; and current-view grounding resolves recalled entities before execution. We evaluate SimpleARM on RoboMME, a benchmark of memory-dependent robot manipulation tasks that require history information no longer available in the current observation. Across all 16 tasks and three policy seeds, SimpleARM achieves 67.17% mean success, compared with 44.51% for the strongest non-oracle baseline. Matched ablations show mechanism specificity: removing relation, reference, progress, or route state produces large losses where the affected state is retrieved for control, while largely sparing other tasks. These results support a state-based view of robot memory: effective memory for control is not simply retained visual history, but compact task-relevant state derived from the interaction history.
Comments34 pages, 13 figures. Project page: https://simplearm.github.io/
