# Ji Pei’s arXiv survey of dexterous robotic hands in the foundation-model era [Canonical reading](https://escapemeridian.com/readings/ji-pei-s-arxiv-survey-of-dexterous-robotic-hands-in-the-aaa3ffb3) Ji Pei’s survey argues that progress in dexterous robotic hands is hard to assess when hardware, data, learning and evaluation are considered separately. It organizes research into five interdependent layers, with human hand-object understanding and tactile sensing as cross-cutting perspectives. Its central point is that more human data is not enough: value depends on kinematic, visual and contact alignment. Shared action representations and foundation policies also leave embodiment-specific constraints on force, latency and contact. And task completion alone does not establish physical correctness or deployment readiness. These distinctions matter because a completed task may leave questions about how the hand acted and whether the result is ready for deployment. Pei identifies limited evidence for zero-shot transfer of contact-rich skills between morphologically different hands, long-duration reliability and cross-sensor tactile generalization. The survey connects design choices through comparative tables and case analyses, while distinguishing within-study comparisons from heterogeneous corpus statistics, vendor specifications and company-reported results. It is a structured narrative synthesis, not a pooled performance ranking or an exhaustive systematic review. Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims. ## Original source arXiv cs.AI · Article Original source date: Nov 4, 2026 (2026-11-04). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-09T12:47:23.490Z. First observed: 2026-11-05T05:06:42.722Z. Retrieved: 2026-11-09T12:46:42.780Z. - [Dexterous Robotic Hands in the Foundation-Model Era: A Full-Stack Survey of Hardware, Data, Cross-Embodiment Transfer, Learning, and Evaluation](https://arxiv.org/abs/2611.06724) ## Limits The paper is presented through its abstract and publisher-page information; the detailed case analyses and comparative tables are not available in this evidence. The review is a structured narrative synthesis, not a pooled performance ranking or an exhaustive systematic review. The abstract says evidence remains limited for zero-shot transfer of contact-rich skills between morphologically different hands, long-duration reliability and cross-sensor tactile generalization. ## Open question What controlled comparisons could establish whether contact-rich skills transfer between morphologically different hands? ## Source-stated Ji Pei’s survey organizes research on dexterous robotic hands across five interdependent layers and two cross-cutting perspectives. > We organize the literature into five interdependent layers: hardware; data collection and curation; cross-embodiment transfer; learning and foundation models; and evaluation and applications. Human hand-object understanding and tactile sensing provide two cross-cutting perspectives. - [Supporting source](https://arxiv.org/abs/2611.06724) ## Source-stated The survey argues that human data’s value depends on alignment, not simply the amount collected, and that shared representations do not remove embodiment-specific constraints. > the value of human data depends on kinematic, visual and contact alignment, not recording volume alone. Second, shared action representations and foundation policies do not remove embodiment-specific constraints on force, latency and contact. - [Supporting source](https://arxiv.org/abs/2611.06724) ## Source-stated The survey says task completion alone does not establish physical correctness or deployment readiness, and identifies evidence gaps in transfer, reliability and tactile generalization. > task completion is insufficient to establish physical correctness or deployment readiness. Evidence remains limited for zero-shot transfer of contact-rich skills between morphologically different hands, long-duration reliability, and cross-sensor tactile generalization. - [Supporting source](https://arxiv.org/abs/2611.06724) ## Source-stated The review is a structured narrative synthesis, not a pooled performance ranking or exhaustive systematic review. > The review is a structured narrative synthesis rather than a pooled performance ranking or an exhaustive systematic review. - [Supporting source](https://arxiv.org/abs/2611.06724) ## Source and contributor context Pei, Ji · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Foundation model A large model trained on broad data that can be adapted to many different tasks. - [Definition reference](https://developers.google.com/machine-learning/glossary#foundation-model) ## Latency The delay between sending a request and receiving the result. - [Definition reference](https://developers.google.com/machine-learning/glossary#latency) ## Zero-shot Doing a task without worked examples (zero-shot) or with only a few examples given in the prompt (few-shot). - [Definition reference](https://developers.google.com/machine-learning/glossary#zero-shot-prompting) ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## Editorial provenance Version 1. Publication ID: 52c535de-05a2-45e6-81dd-7bdfdb9efe4f. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-09T12:47:23.490Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: useful\_perspective. Observation: c9e2dc40f4ff8c78f40984cf07928857528684fb262e8ba5581d9cd28a006c96. Source text hash: c77f70fb1255fc57972146a9d0fb487ae9647f4c6c71cc8dbe64386a30e5f658. ## Revision and source-check history ### Version 1 — 2026-11-09T12:47:23.490Z Ji Pei’s survey argues that progress in dexterous robotic hands is hard to assess when hardware, data, learning and evaluation are considered separately. It organizes research into five interdependent layers, with human hand-object understanding and tactile sensing as cross-cutting perspectives. Its central point is that more human data is not enough: value depends on kinematic, visual and contact alignment. Shared action representations and foundation policies also leave embodiment-specific constraints on force, latency and contact. And task completion alone does not establish physical correctness or deployment readiness. These distinctions matter because a completed task may leave questions about how the hand acted and whether the result is ready for deployment. Pei identifies limited evidence for zero-shot transfer of contact-rich skills between morphologically different hands, long-duration reliability and cross-sensor tactile generalization. The survey connects design choices through comparative tables and case analyses, while distinguishing within-study comparisons from heterogeneous corpus statistics, vendor specifications and company-reported results. It is a structured narrative synthesis, not a pooled performance ranking or an exhaustive systematic review. ## Original source arXiv cs.AI · Article Original source date: Nov 4, 2026 (2026-11-04). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-09T12:47:23.490Z. First observed: 2026-11-05T05:06:42.722Z. Retrieved: 2026-11-09T12:46:42.780Z. - [Dexterous Robotic Hands in the Foundation-Model Era: A Full-Stack Survey of Hardware, Data, Cross-Embodiment Transfer, Learning, and Evaluation](https://arxiv.org/abs/2611.06724) ## Limits The paper is presented through its abstract and publisher-page information; the detailed case analyses and comparative tables are not available in this evidence. The review is a structured narrative synthesis, not a pooled performance ranking or an exhaustive systematic review. The abstract says evidence remains limited for zero-shot transfer of contact-rich skills between morphologically different hands, long-duration reliability and cross-sensor tactile generalization. ## Open question What controlled comparisons could establish whether contact-rich skills transfer between morphologically different hands? ## Source-stated Ji Pei’s survey organizes research on dexterous robotic hands across five interdependent layers and two cross-cutting perspectives. > We organize the literature into five interdependent layers: hardware; data collection and curation; cross-embodiment transfer; learning and foundation models; and evaluation and applications. Human hand-object understanding and tactile sensing provide two cross-cutting perspectives. - [Supporting source](https://arxiv.org/abs/2611.06724) ## Source-stated The survey argues that human data’s value depends on alignment, not simply the amount collected, and that shared representations do not remove embodiment-specific constraints. > the value of human data depends on kinematic, visual and contact alignment, not recording volume alone. Second, shared action representations and foundation policies do not remove embodiment-specific constraints on force, latency and contact. - [Supporting source](https://arxiv.org/abs/2611.06724) ## Source-stated The survey says task completion alone does not establish physical correctness or deployment readiness, and identifies evidence gaps in transfer, reliability and tactile generalization. > task completion is insufficient to establish physical correctness or deployment readiness. Evidence remains limited for zero-shot transfer of contact-rich skills between morphologically different hands, long-duration reliability, and cross-sensor tactile generalization. - [Supporting source](https://arxiv.org/abs/2611.06724) ## Source-stated The review is a structured narrative synthesis, not a pooled performance ranking or exhaustive systematic review. > The review is a structured narrative synthesis rather than a pooled performance ranking or an exhaustive systematic review. - [Supporting source](https://arxiv.org/abs/2611.06724) ## Source and contributor context Pei, Ji · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Foundation model A large model trained on broad data that can be adapted to many different tasks. - [Definition reference](https://developers.google.com/machine-learning/glossary#foundation-model) ## Latency The delay between sending a request and receiving the result. - [Definition reference](https://developers.google.com/machine-learning/glossary#latency) ## Zero-shot Doing a task without worked examples (zero-shot) or with only a few examples given in the prompt (few-shot). - [Definition reference](https://developers.google.com/machine-learning/glossary#zero-shot-prompting) ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## Editorial provenance Version 1. Publication ID: 52c535de-05a2-45e6-81dd-7bdfdb9efe4f. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-09T12:47:23.490Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: useful\_perspective. Observation: c9e2dc40f4ff8c78f40984cf07928857528684fb262e8ba5581d9cd28a006c96. Source text hash: c77f70fb1255fc57972146a9d0fb487ae9647f4c6c71cc8dbe64386a30e5f658. ### uncertain — 2026-11-09T12:47:23.490Z — anthropic/claude-opus-5.5 Publication: 52c535de-05a2-45e6-81dd-7bdfdb9efe4f attribution — pass: The page lists Ji Pei as the sole author of this arXiv survey. The reading attributes the survey's conclusions to Pei and keeps them as source arguments rather than established facts. > Authors:Ji Pei summary — uncertain: The summary follows the abstract faithfully, but a non-specialist would struggle with it. Central terms are never explained: foundation policies, kinematic, visual and contact alignment, embodiment-specific constraints, zero-shot transfer, morphologically different hands and cross-sensor tactile generalization. These terms carry the argument, so tooltips or brief plain glosses are needed. The reading also drops the survey's constructive contribution: eleven open problems and proposed reporting interfaces and controlled comparisons. Smallest repair: add one sentence on these proposals and gloss the key terms. The title and question are faithful. > We identify eleven open problems and propose reporting interfaces and controlled comparisons that can make full-stack progress more measurable. date — pass: The originalPublished date of 2026-11-04 matches the v1 submission date. The feed date of Nov 5 is later, so it is correctly not used. > \[Submitted on 4 Nov 2026\] support — pass: Each claim quotes the abstract verbatim. The sentence saying a completed task may leave questions about how the hand acted is a modest restatement of the source's claim about physical correctness and deployment readiness. It does not add new benefits or forecasts. The opening framing about separate assessment reasonably reflects the abstract. > progress in hardware, representation learning and evaluation is often assessed separately media — pass: The selected media is the verified arXiv abstract page, which is the direct original source. No better source media was found. ## Recorded references Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter. No public references recorded here yet. [Reference data (JSON)](/readings/aaa3ffb32785c3df32b4561a88f8c13456842aae4ce48b7a668331127f59daee/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/ji-pei-s-arxiv-survey-of-dexterous-robotic-hands-in-the-aaa3ffb3.json)