Evo-DexBench: Benchmarking Generalization across Related Tasks for Dexterous Manipulation
作者: Bing Cheng, Yiran Mao, Zewei Ye, Yifan Jin, Xinyi Chen, Ruifeng Tu, Shuhuai Wang, Tao Lin, Zhimu Zhou, Yuxin Du, Fangwei Zhong, Junchi Yan, Bo Zhao
分类: cs.RO
发布日期: 2026-10-15
💡 一句话要点
提出Evo-DexBench以解决灵巧操控任务间知识迁移问题
🎯 匹配领域: 支柱一：机器人控制 (Robot Control) 支柱九：具身大模型 (Embodied Foundation Models)
关键词: 灵巧操控 知识迁移 任务基准 多模态演示 机器人学习 功能交互 联合训练
📋 核心要点
- 现有基准无法有效区分知识迁移与任务记忆，且提供过多目标演示，导致评估不准确。
- 提出Core-Gen原则，定义核心任务与生成任务，设计Evo-DexBench以控制任务相关性与曝光。
- 实验结果显示最佳方法成功率仅为38.5%，联合训练在性能上分别提升12%和38%，验证了方法有效性。
📝 摘要（中文）
通用灵巧操控要求策略能够从相关任务中迁移知识到新任务，且仅需少量目标演示。现有基准往往提供过多目标演示，并且未能有效控制源任务与目标任务之间的差异，导致无法清晰区分知识迁移与任务记忆。本文提出了Core-Gen原则，定义了核心任务与生成任务，并基于此设计了Evo-DexBench基准，控制源任务相关性与目标任务曝光。实验表明，即使是最佳方法的平均成功率也仅为38.5%，显示出现有方法在知识迁移中的挑战。联合Core+Gen训练在性能上分别提升了12%和38%，表明联合训练更有效利用核心知识。
🔬 方法详解
问题定义：本文旨在解决灵巧操控任务中知识迁移的评估问题。现有方法往往提供过多目标演示，且未能控制源任务与目标任务之间的差异，导致无法清晰区分知识迁移与任务记忆。
核心思路：论文提出Core-Gen原则，核心任务是相关任务，提供源知识；生成任务是同一功能家族中的新目标，仅需少量演示。该设计旨在测试策略是否能够重用功能交互知识。
技术框架：Evo-DexBench包括30个任务，分为六个功能家族，并配备虚拟现实远程操作管道。数据集涵盖每个任务的三个日常场景，包含1,248个多模态演示，使用模拟的Shadow Hands收集，配备123个触觉传感器。
关键创新：最重要的创新在于Core-Gen原则的提出及其在基准设计中的应用，能够有效控制任务相关性与曝光，从而更准确地评估知识迁移能力。
关键设计：在实验中，联合Core+Gen训练的设计使得策略更好地利用核心知识，且功能匹配的核心数据在成功率上比不匹配的数据获得更大提升。
🖼️ 关键图片
📊 实验亮点
实验结果显示，最佳方法的平均成功率仅为38.5%，强调了现有方法在功能交互知识迁移中的挑战。联合Core+Gen训练在性能上分别提升了12%和38%，表明这种训练方式能够更有效地利用核心知识。
🎯 应用场景
该研究的潜在应用领域包括机器人抓取、自动化装配和人机协作等场景。通过提高灵巧操控策略的知识迁移能力，能够在实际应用中实现更高效的任务执行，降低对大量演示的依赖，提升机器人在复杂环境中的适应性和灵活性。
📄 摘要（原文）
General-purpose dexterous manipulation requires policies to transfer knowledge from related tasks to new tasks with only a few target demonstrations. Existing benchmarks often provide too many target demonstrations. They also leave the differences between source and target tasks uncontrolled. As a result, they cannot clearly distinguish knowledge transfer from task memorization. Our key insight is the Core-Gen principle. Core tasks are related tasks that share a functional interaction and provide source knowledge, while a Gen task is a novel target within the same functional family that has only a few demonstrations. This setup tests whether a policy can reuse functional interaction knowledge. Based on this principle, we propose Evo-DexBench, which controls both source-task relatedness and target-task exposure. It comprises 30 tasks in six functional families and a VR teleoperation pipeline. The resulting dataset covers three everyday scenes per task. It contains 1,248 multimodal demonstrations collected with simulated Shadow Hands equipped with 123 taxels. Experiments show that even the best-performing method achieves an average success rate of only 38.5%, highlighting the challenges existing methods face in transferring functional interaction knowledge. Further analysis shows that, across the two tasks, joint Core+Gen training improves performance by 12% and 38%, respectively, over training on Core followed by fine-tuning on Gen. This indicates that joint training makes better use of Core knowledge and that functionally matched Core data yields greater gains in success rate than mismatched data. We release the Evo-DexBench dataset and benchmark to support the evaluation of dexterous manipulation policies' generalization across related tasks.