# EvoMem-VLA：用于长程机器人操作的状态演化记忆

> 原文链接：https://robotics.hellobike.com/papers/2026-10-04-2610-05418
> 频道：papers | 更新日期：2026-10-04 | 标签：arxiv, cs.RO
> 核心摘要：EvoMem-VLA通过显式编码历史状态变化构建状态演化记忆，并在仿真基准和真实长程操作任务上验证性能。

# EvoMem-VLA：用于长程机器人操作的状态演化记忆

::: v-pre
> 原文链接：[EvoMem-VLA: State-Evolution Memory for Long-Horizon Robot Manipulation](https://arxiv.org/abs/2610.05418)（arXiv 2610.05418 · 发布日期：2026-10-04）

**EvoMem-VLA: State-Evolution Memory for Long-Horizon Robot Manipulation**

EvoMem-VLA通过显式编码历史状态变化构建状态演化记忆，并在仿真基准和真实长程操作任务上验证性能。

## 核心要点

- 现有VLA依赖当前观测，证据离开视野后丢失，限制长程记忆任务表现。
- 方法显式保留历史状态间变化，用条件delta tokenization编码有序帧对。
- 共享VLM骨干支持任务自适应路由，长程任务直接动作，多阶段任务先生成子任务。
- 在RMBench、RoboMME与真实任务上成功率80.7%、82.0%、83.8%，均超既有SOTA。

**验证范围与局限**：验证范围包括RMBench、RoboMME仿真基准，以及涵盖两种机器人形态的四个真实任务。

## 英文摘要 (Abstract)

Most vision-language-action (VLA) models rely on current observations and lose task-relevant evidence once it leaves view, limiting performance on long-horizon, memory-dependent tasks. Existing efforts incorporate compressed historical features or sparse visual keyframes. However, isolated snapshots can leave the policy uncertain about what changed during past interactions and which action should follow. To overcome this limitation, we propose EvoMem-VLA, which constructs state-evolution memory by explicitly encoding and retaining observed changes between historical states. These change representations preserve evidence of interaction outcomes, allowing the policy to track task progress beyond isolated snapshots. Specifically, we introduce conditional delta tokenization to encode ordered frame pairs into directional, source-conditioned delta tokens, each associated with its corresponding state evidence. A shared VLM backbone supports task-adaptive routing: normal long-horizon tasks follow a direct action route, whereas multi-stage tasks use a subtask route that generates an executable subtask as an additional input for action generation. With a single jointly trained policy for each simulation benchmark, EvoMem-VLA achieves success rates of 80.7\% on RMBench, 82.0\% on RoboMME and 83.8\% across four real-world tasks spanning two robot embodiments. These results represent substantial improvements over the previous state of the art in all three evaluation settings.

作者：Yuheng Na, Zhide Zhong, Junjie He, Junfeng Li, Haodong Yan, Jiaan Wang, Jiaguan Zhu, Yangyang Zheng, Tianyu Huang, Haoang Li
:::

## 原始参考来源 (Sources)

- [arXiv 2610.05418：EvoMem-VLA: State-Evolution Memory for Long-Horizon Robot Manipulation](https://arxiv.org/abs/2610.05418)