# ManiUnit基准发布：长程操作技能解耦评测与VLA策略验证

> 原文链接：https://robotics.hellobike.com/papers/2026-10-08-2610-12089
> 频道：papers | 更新日期：2026-10-08 | 标签：auto, arxiv
> 核心摘要：针对长程移动操作痛点，ManiUnit提供13.8万段技能数据与1260个测试实例。实测显示关节扰动致成功率降56%，技能策略局部成功率达78.7%，结合规划器使全任务成功率从4.0%升至18.0%。

# ManiUnit基准发布：长程操作技能解耦评测与VLA策略验证

::: v-pre
> 本文由 AI 根据公开内容整理，可能有误，请以原文为准：[ManiUnit: A Manipulation Skill Dataset and Benchmark for Long-Horizon Tasks](https://arxiv.org/abs/2610.12089)（arXiv 2610.12089）

## 核心内容

针对长程移动操作中观测歧义、状态偏离及诊断困难三大挑战，研究团队基于BEHAVIOR-1K构建ManiUnit数据集与基准。该资产包含源自50项活动的137,899个片段，覆盖21种技能类型与417个子任务，并配套1,260个测试实例。每条数据均绑定显式子任务指令，支持中间状态恢复与局部成功判定，解决了传统任务级指标无法精准定位技能缺陷的问题，为具身智能长程操作提供了细粒度评测标准。

评测揭示了视觉语言动作（VLA）模型在聚合分数掩盖下的技能差异及状态敏感性：起始关节配置扰动导致全基准成功率下降约56%。在两项长程活动中，基于ManiUnit训练的技能策略局部操作成功率达78.7%，显著优于完整演示训练的49.3%。进一步通过规划器协同任务与技能策略，将全任务成功率从4.0%提升至18.0%，验证了“技能解耦+规划编排”范式在提升系统鲁棒性方面的工程价值。

## 关键看点

- **数据规模**：包含137,899个操作片段、21种技能类型及417个子任务，提供1,260个测试实例支撑细粒度评测
- **状态敏感性**：起始关节配置扰动导致全基准成功率下降约56%，揭示VLA策略对初始状态依赖度高
- **技能策略优势**：技能策略局部操作成功率达78.7%，较完整演示训练的任务策略提升29.4个百分点
- **规划协同增益**：通过规划器协调任务与技能策略，长程活动全任务成功率从4.0%提升至18.0%

## 工程与产业观察

ManiUnit证实长程操作瓶颈在于技能级状态鲁棒性而非单纯语义理解。关节扰动致成功率腰斩的数据警示，具身算法需强化对非理想初始状态的适应能力。技能解耦训练配合规划器编排可大幅提升系统上限，这为VLA模型从端到端黑盒向模块化、可诊断的工程化落地提供了关键实证路径与数据基座。

## 英文摘要 (Abstract)

Long-horizon mobile manipulation requires a robot to navigate multi-room environments and execute a sequence of manipulation skills under a single natural language instruction. Learning and evaluating these skills present three challenges: similar observations under a fixed task instruction may make skill selection ambiguous; even when a preceding skill succeeds, the robot state inherited by the next skill may deviate from its demonstrated starting states and affect execution; and task-level metrics hinder skill-specific diagnosis, while early failures leave later skills untested. We therefore introduce ManiUnit, a manipulation skill dataset and benchmark built from 50 BEHAVIOR-1K activities. Its dataset contains 137,899 segments across 21 skill types and 417 subtasks, and its benchmark contains 1,260 test instances. Correspondingly, ManiUnit pairs each segment with an explicit subtask instruction; measures sensitivity to perturbations of the robot's starting base position or joint configuration; and restores intermediate simulator states and defines local success conditions so that each skill can be evaluated without executing preceding stages. Evaluations of representative vision-language-action (VLA) policies show that similar aggregate scores can hide substantial per-skill differences. The tested starting-state perturbations also degrade execution: on the full benchmark, joint perturbations reduce success rates by approximately 56% relative to those from demonstrated starting states. On two long-horizon activities, a skill policy trained on ManiUnit segments achieves 78.7% local manipulation success, compared with 49.3% for a task policy trained on complete demonstrations. The trained skills further support complete-task execution on these activities, as coordinating the task and skill policies through a planner raises full-task success from 4.0% to 18.0%.

作者：Guoting Wei, Dawei Yan, Xia Yuan, Gengming Zhang, Yelin He, Guodong Du, Jiaquan Ye, Heng Zhang, Xinming Wei, Xianbiao Qi, Chunxia Zhao, Haokui Zhang, Rong Xiao
:::

## 原始参考来源 (Sources)

- [arXiv 2610.12089：ManiUnit: A Manipulation Skill Dataset and Benchmark for Long-Horizon Tasks](https://arxiv.org/abs/2610.12089)