Newshot

Evo-DexBench: Benchmarking Generalization across Related Tasks for Dexterous Manipulation

Summary

arXiv:2610.21180v3 Announce Type: replace Abstract: 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 and 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, three everyday layouts per task, and 1,248 multimodal demonstrations collected via…

Original text(via arXiv cs.RO)

Computer Science > Robotics

Title:Evo-DexBench: Benchmarking Generalization across Related Tasks for Dexterous Manipulation

Abstract: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 and 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, three everyday layouts per task, and 1,248 multimodal demonstrations collected via VR teleoperation with simulated Shadow Hands equipped with 123 taxels. We first compare few-shot adaptation strategies and find that joint Core + Gen training makes the most effective use of Core knowledge. Under this setting, we evaluate six representative policies. The best-performing policy, $\pi_{0.5}$, achieves a mean success rate of only 24.4%, revealing the limitations of existing methods in transferring functional interaction knowledge. Further analysis shows that on the tested Gen targets, functionally matched Core data yields larger success-rate gains than mismatched data, and that with matched Core data, 8 Gen demonstrations reach the performance of training on 16--32 Gen demonstrations alone. These results indicate that Evo-DexBench separates knowledge transfer from task memorization under controlled conditions and effectively evaluates the generalization of dexterous manipulation policies across related tasks.
Subjects:Robotics (cs.RO)
Cite as:arXiv:2610.21180 [cs.RO]
(or arXiv:2610.21180v3 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2610.21180 arXiv-issued DOI via DataCite

Submission history

Access Paper:

  • HTML (experimental)

Additional Features

  • Audio Summary

Current browse context:

References & Citations

  • Google Scholar
  • Semantic Scholar

BibTeX formatted citation

Bookmark

Bibliographic and Citation Tools

Code, Data and Media Associated with this Article

Demos

Recommenders and Search Tools

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

The text above was extracted automatically from the source page and may miss images, tables or code blocks. The original is authoritative.

Published by arXiv cs.RO
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗