# arXiv cs.AI: Tactus authors present a glove for dexterity data collection [Canonical reading](https://escapemeridian.com/readings/arxiv-cs-ai-tactus-authors-present-a-glove-for-dexterity-data-9da9bfa3) The authors of Tactus present a textile-integrated glove designed to record fingertip force and hand pose during natural manipulation. They argue that collecting both signals matters for robot dexterous manipulation policies and world models, but say existing wearable platforms do not combine calibrated fingertip force sensing with hand tracking that resists occlusion while preserving natural movement. Tactus combines a thin, conformal knit design, tactile sensors placed at manipulation-relevant regions, and pose tracking that fuses data from inertial measurement units (IMUs). The authors report a 1.5 mm fingertip stackup and say the tracking maintains accuracy when the hand is obscured by an object. They also describe calibration procedures and benchmarking protocols, and compare Tactus with two commercially available tactile gloves. In their tests, the authors report a minimum detectable force of 0.59 N at 95% confidence, 1.6–2.7 times lower than the commercial baselines. They also report 0.87 cross-participant consistency and 11.0 mm fingertip keypoint relative error during manipulation. The authors say the combined signals support object property estimation, slip detection, and recording hand-object interactions for contact-aware policies and world models. The abstract does not establish how well systems trained on the resulting data perform beyond the reported tests. 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: Oct 29, 2026 (2026-10-29). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-13T03:31:32.712Z. First observed: 2026-10-30T06:42:56.048Z. Retrieved: 2026-10-31T08:47:45.393Z. - [Tactus: Tactile & Pose Sensing Glove for Natural Dexterity Data Collection](https://arxiv.org/abs/2610.39076) ## Limits The evidence is the paper’s abstract; it does not provide the detailed methods, test conditions, or results needed to assess the reported performance. The abstract says the signals support contact-aware policies and world models, but does not establish how well systems trained on the resulting data perform beyond the reported tests. ## Open question How well do Tactus-recorded demonstrations support contact-aware manipulation policies and world models beyond the reported tests? ## Source-stated The authors present Tactus as a glove designed to collect tactile force and hand-pose data during natural manipulation. > We present Tactus, a textile-integrated sensing glove - [Supporting source](https://arxiv.org/abs/2610.39076) ## Source-stated The authors say the glove combines calibrated fingertip force sensing and occlusion-resilient hand tracking in a form intended to preserve natural manipulation behavior. > no existing wearable platform provides both calibrated fingertip force sensing and occlusion-resilient hand tracking in a form factor that preserves natural manipulation behavior - [Supporting source](https://arxiv.org/abs/2610.39076) ## Source-stated The authors report a minimum detectable force of 0.59 N at 95% confidence, 1.6–2.7 times lower than the commercial baselines they tested. > Tactus resolves a minimum detectable force of 0.59 N at 95% confidence, 1.6 - 2.7x lower than the commercial baselines - [Supporting source](https://arxiv.org/abs/2610.39076) ## Source-stated The authors report 0.87 cross-participant consistency and 11.0 mm fingertip keypoint relative error during manipulation. > with 0.87 cross-participant consistency, while our pose tracking sustains 11.0 mm fingertip keypoint relative error during manipulation - [Supporting source](https://arxiv.org/abs/2610.39076) ## Source-stated The authors say the combined signals support object property estimation, slip detection, and recording hand-object interaction for contact-aware policies and world models. > We show that these combined signals support object property estimation, slip detection, and representative recording of hand-object interaction - [Supporting source](https://arxiv.org/abs/2610.39076) ## Source and contributor context Guan, Li, Granberry, Rachael, Aseeri, Sahar, Rostami, Sahar, Niehues, Taylor, Marri, Sai Dhanooshi, Huang, Mia, Zhang, Dingtian, Kou, Ziyi, Pursel, Jason, Whitmire, Eric, Mehta, Vatsal, Deshpande, Ashish, Trutna, Tristan, Belcher, Reed, Black, Emelia, Blake, Julian, Boldu, Roger, Bonaventure, Sean, Boyovich, Einar, Brennan, Jordan, Casillas, Elizabeth, Christensen, David, Christensen, Tyler, Colonnese, Nicholas, Dealey, Stuart, DeHaven, Gaige, Doran, Katie, Doughty, Chris, Drury, Robert, Espino, Silvia, Geyne, Gabriel, Goodrich, Beth, Gould, Kyra, Grisso, Joe, Guerrero, Susana, Gwozdziowski, Andrew, Ha, Monica, Hajiesmaili, Ehsan, Healy, Kate, Hope, Rose, Hurst, Michael, Ikemiya, Claire, Isibor, Daisy, James, Rick, Joginapally, Yashasvi, Johnson, Samuel, Jorgensen, Tyler, Jost, Kristy, Kamath, Ben, Karulin, Slava, Keener, Alexander, Kim, Woo, Krauskopf, Alison, Kumar, Ankit, Lang, Joshua, Li, Lizzy, Lynch, Cameron, Ma, Daniel, Martinon, Hector, Matos, Christina, McCurdy, Keagan, Mincey, Lee, Nezlek, Stuart, Nguyen, Tiffany, O'Hara, James, Osborne, Jason, Pai, Jacky, Pardakhtim, Roham, Perek, David, Poulin, Alexandre, Raines, Nick, Rasheed, Salwa, Resneck, Leah, Rhodes, Kevin, Ristani, Ergys, Roberts, Douglas, Sarker, Anik, Scarborough, Harrison, Schmidtler, Jonas, Shieh, Brian, Shuman, Benjamin, Simard, Garrett, Smith, Forrest, Somarathna, Udara, Sosin, Eliot, Stewart, Jonathan, Strom, Kristine, Swett, Adam, Talbot-Mantooth, Hal, Tanner, Julia, Thirunagari, Sai Sharan, Torrey, Curtis, Uzun, Simge, Wachs, Eileen, Wasson, Rylan, Wei, Henry, White, Jesse, Woodbyrne, Beau, Woods, Antonio · Article authors (partial credit list) 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. ## World model An AI system's internal model of how an environment behaves, used to predict what happens next or to plan. - [Definition reference](https://arxiv.org/abs/1803.10122) ## Benchmark A defined set of tests used to compare systems. Results depend on the tasks and measures selected. - [Definition reference](https://arxiv.org/abs/2211.09110) ## 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: 38f3a5fb-eb03-46c0-8c1c-4d1897062ded. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: passed by anthropic/claude-opus-5.5 on 2026-11-13T03:31:32.712Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: 13ea62c2acab6bb73ac448222662ea333446fd6b5e2dd346650a6de0cdcdc8c0. Source text hash: 2cc608a9633af0ae80b7902409ad23050b7a8d61121441eaa2a53122fd2687e2. ## Revision and source-check history ### Version 1 — 2026-11-13T03:31:32.712Z The authors of Tactus present a textile-integrated glove designed to record fingertip force and hand pose during natural manipulation. They argue that collecting both signals matters for robot dexterous manipulation policies and world models, but say existing wearable platforms do not combine calibrated fingertip force sensing with hand tracking that resists occlusion while preserving natural movement. Tactus combines a thin, conformal knit design, tactile sensors placed at manipulation-relevant regions, and pose tracking that fuses data from inertial measurement units (IMUs). The authors report a 1.5 mm fingertip stackup and say the tracking maintains accuracy when the hand is obscured by an object. They also describe calibration procedures and benchmarking protocols, and compare Tactus with two commercially available tactile gloves. In their tests, the authors report a minimum detectable force of 0.59 N at 95% confidence, 1.6–2.7 times lower than the commercial baselines. They also report 0.87 cross-participant consistency and 11.0 mm fingertip keypoint relative error during manipulation. The authors say the combined signals support object property estimation, slip detection, and recording hand-object interactions for contact-aware policies and world models. The abstract does not establish how well systems trained on the resulting data perform beyond the reported tests. ## Original source arXiv cs.AI · Article Original source date: Oct 29, 2026 (2026-10-29). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-13T03:31:32.712Z. First observed: 2026-10-30T06:42:56.048Z. Retrieved: 2026-10-31T08:47:45.393Z. - [Tactus: Tactile & Pose Sensing Glove for Natural Dexterity Data Collection](https://arxiv.org/abs/2610.39076) ## Limits The evidence is the paper’s abstract; it does not provide the detailed methods, test conditions, or results needed to assess the reported performance. The abstract says the signals support contact-aware policies and world models, but does not establish how well systems trained on the resulting data perform beyond the reported tests. ## Open question How well do Tactus-recorded demonstrations support contact-aware manipulation policies and world models beyond the reported tests? ## Source-stated The authors present Tactus as a glove designed to collect tactile force and hand-pose data during natural manipulation. > We present Tactus, a textile-integrated sensing glove - [Supporting source](https://arxiv.org/abs/2610.39076) ## Source-stated The authors say the glove combines calibrated fingertip force sensing and occlusion-resilient hand tracking in a form intended to preserve natural manipulation behavior. > no existing wearable platform provides both calibrated fingertip force sensing and occlusion-resilient hand tracking in a form factor that preserves natural manipulation behavior - [Supporting source](https://arxiv.org/abs/2610.39076) ## Source-stated The authors report a minimum detectable force of 0.59 N at 95% confidence, 1.6–2.7 times lower than the commercial baselines they tested. > Tactus resolves a minimum detectable force of 0.59 N at 95% confidence, 1.6 - 2.7x lower than the commercial baselines - [Supporting source](https://arxiv.org/abs/2610.39076) ## Source-stated The authors report 0.87 cross-participant consistency and 11.0 mm fingertip keypoint relative error during manipulation. > with 0.87 cross-participant consistency, while our pose tracking sustains 11.0 mm fingertip keypoint relative error during manipulation - [Supporting source](https://arxiv.org/abs/2610.39076) ## Source-stated The authors say the combined signals support object property estimation, slip detection, and recording hand-object interaction for contact-aware policies and world models. > We show that these combined signals support object property estimation, slip detection, and representative recording of hand-object interaction - [Supporting source](https://arxiv.org/abs/2610.39076) ## Source and contributor context Guan, Li, Granberry, Rachael, Aseeri, Sahar, Rostami, Sahar, Niehues, Taylor, Marri, Sai Dhanooshi, Huang, Mia, Zhang, Dingtian, Kou, Ziyi, Pursel, Jason, Whitmire, Eric, Mehta, Vatsal, Deshpande, Ashish, Trutna, Tristan, Belcher, Reed, Black, Emelia, Blake, Julian, Boldu, Roger, Bonaventure, Sean, Boyovich, Einar, Brennan, Jordan, Casillas, Elizabeth, Christensen, David, Christensen, Tyler, Colonnese, Nicholas, Dealey, Stuart, DeHaven, Gaige, Doran, Katie, Doughty, Chris, Drury, Robert, Espino, Silvia, Geyne, Gabriel, Goodrich, Beth, Gould, Kyra, Grisso, Joe, Guerrero, Susana, Gwozdziowski, Andrew, Ha, Monica, Hajiesmaili, Ehsan, Healy, Kate, Hope, Rose, Hurst, Michael, Ikemiya, Claire, Isibor, Daisy, James, Rick, Joginapally, Yashasvi, Johnson, Samuel, Jorgensen, Tyler, Jost, Kristy, Kamath, Ben, Karulin, Slava, Keener, Alexander, Kim, Woo, Krauskopf, Alison, Kumar, Ankit, Lang, Joshua, Li, Lizzy, Lynch, Cameron, Ma, Daniel, Martinon, Hector, Matos, Christina, McCurdy, Keagan, Mincey, Lee, Nezlek, Stuart, Nguyen, Tiffany, O'Hara, James, Osborne, Jason, Pai, Jacky, Pardakhtim, Roham, Perek, David, Poulin, Alexandre, Raines, Nick, Rasheed, Salwa, Resneck, Leah, Rhodes, Kevin, Ristani, Ergys, Roberts, Douglas, Sarker, Anik, Scarborough, Harrison, Schmidtler, Jonas, Shieh, Brian, Shuman, Benjamin, Simard, Garrett, Smith, Forrest, Somarathna, Udara, Sosin, Eliot, Stewart, Jonathan, Strom, Kristine, Swett, Adam, Talbot-Mantooth, Hal, Tanner, Julia, Thirunagari, Sai Sharan, Torrey, Curtis, Uzun, Simge, Wachs, Eileen, Wasson, Rylan, Wei, Henry, White, Jesse, Woodbyrne, Beau, Woods, Antonio · Article authors (partial credit list) 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. ## World model An AI system's internal model of how an environment behaves, used to predict what happens next or to plan. - [Definition reference](https://arxiv.org/abs/1803.10122) ## Benchmark A defined set of tests used to compare systems. Results depend on the tasks and measures selected. - [Definition reference](https://arxiv.org/abs/2211.09110) ## 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: 38f3a5fb-eb03-46c0-8c1c-4d1897062ded. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: passed by anthropic/claude-opus-5.5 on 2026-11-13T03:31:32.712Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: 13ea62c2acab6bb73ac448222662ea333446fd6b5e2dd346650a6de0cdcdc8c0. Source text hash: 2cc608a9633af0ae80b7902409ad23050b7a8d61121441eaa2a53122fd2687e2. ### passed — 2026-11-13T03:31:32.712Z — anthropic/claude-opus-5.5 Publication: 38f3a5fb-eb03-46c0-8c1c-4d1897062ded attribution — pass: The reading attributes the glove design, motivation, benchmarks and performance figures to the Tactus authors. It keeps the results as reported claims rather than independently verified facts. Labeling the publisher as arXiv cs.AI matches the feed metadata, and the paper is cross-listed in Artificial Intelligence. > We present Tactus, a textile-integrated sensing glove summary — pass: The summary states the actor, the development and the gap the authors claim to address. It explains IMUs and occlusion in plain words. Remaining terms such as world models, policies and keypoint error suit tooltips rather than blocking understanding. The figures, conditions and stated uses match the abstract. The question names a genuine unresolved issue rather than a leading claim. The source's priority claim of a first cross-platform evaluation is omitted, but that does not distort the meaning. > Tactus resolves a minimum detectable force of 0.59 N at 95% confidence, 1.6 - 2.7x lower than the commercial baselines, and with 0.87 cross-participant consistency date — pass: The original publication date of 2026-10-29 matches the arXiv v1 submission date. The later feed date and observation date are not used as the publication date. > \[Submitted on 29 Oct 2026\] support — pass: Each claim is grounded in the abstract, including the 1.5 mm stackup, the 0.59 N figure, the 1.6-2.7x comparison, the 0.87 consistency, the 11.0 mm error and the stated applications. The limitations correctly note that the evidence is an abstract only. They also note that downstream policy performance is not established. No benefits or costs are invented. > yielding demonstration data with the fidelity required for contact-aware manipulation policies and world models media — pass: The selected media is the verified direct arXiv abstract page for this paper, which is the most direct access point to the source. ## 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/9da9bfa355dac13eb2ec4c642f1f5df15629fed125673a4028e4738171318a4b/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/arxiv-cs-ai-tactus-authors-present-a-glove-for-dexterity-data-9da9bfa3.json)