Tactile Datasets Index: What Each Dataset Contains

Tactile dataset refers to a collection of recorded or simulated contact observations, often aligned with images, robot state or actions. This tactile dataset index contains 26 records organized by sensor type, task and license. Each record keeps its source-review date, download route and access limitations; a linked dataset is not necessarily a verified or unrestricted download.

Last updated:

Published | Updated by

November 8, 2026: added TouchScale release-scope, license and action-label evidence. Public metadata review only; no gated payload access or training.

Organized robot skin learning library with technical cards, tactile sensor samples, and research screens.
Resource-library visual for public learning routes and technical references.

Tactile dataset index by sensor type

This page indexes 27 records. Last updated: 2026-11-08. Individual source-review dates are retained.

Vision-based / optical

Magnetic

    Other, mixed or simulated

    Tactile dataset index by task

    This page indexes 27 records. Last updated: 2026-11-08. Individual source-review dates are retained.

    These groups match the stated task fields. They do not infer a slip or insertion label from a sensor name.

    grasping

    slip

    No record explicitly labels this task in the current index. Check the original release before assuming labels exist.

    insertion

    Tactile dataset index by license

    This page indexes 27 records. Last updated: 2026-11-08. Individual source-review dates are retained.

    Data, code and weights can have different terms. A repository license does not automatically cover a dataset. Unverified terms remain unresolved.

    Dataset reuse terms and primary access routes
    DatasetDataset license and limitsDownload / access sourceSource reviewed
    EgoTouch tactile datasetA separate dataset-file license was not verified. The official GitHub repository is MIT licensed, but that project-code license is not generalized here to the hosted EgoTouch data files.EgoTouch access source2026-08-22
    SoftVTBench tactile datasetApache-2.0 as shown on the current Hugging Face dataset card; upstream simulator, robot, tactile-runtime, and third-party asset terms may differSoftVTBench access source2026-08-22
    RCT: Robotic Contact Tactile tactile datasetCC BY 4.0 dataset; Apache-2.0 codeRCT: Robotic Contact Tactile access source2026-08-16
    ObjectFolder 2.0 tactile datasetCC BY 4.0 for ObjectFolder 2.0 according to the official download pageObjectFolder 2.0 access source2026-08-19
    TacVerse tactile datasetCC BY 4.0 in the official Hugging Face metadata; access is gated and asset-specific terms were not independently checkedTacVerse access source2026-09-11
    Humanoid Visual-Tactile-Action Dataset tactile datasetDataset access terms not stated on the reviewed paper pageHumanoid Visual-Tactile-Action Dataset access source2026-08-16
    TouchScale human visual-tactile dataset tactile datasetDataset card: CC BY-NC 4.0; noncommercial conditions apply. Gated access requires acceptance of host conditions and is not unrestricted reuse permission.TouchScale human visual-tactile dataset access source2026-11-08
    EgoExplorer human vision-motion-touch mini release tactile datasetDataset card: CC BY-NC 4.0. Noncommercial conditions apply; approval or download access is not commercial permission.EgoExplorer human vision-motion-touch mini release access source2026-10-26
    Sparsh-X Multisensory Touch Resource tactile datasetDataset license and download URL not stated on the reviewed paper pageSparsh-X Multisensory Touch Resource access source2026-08-16
    Touch and Go tactile datasetDataset license not stated on the reviewed paper pageTouch and Go access source2026-08-19
    TactiDex tactile datasetDataset license not stated on the reviewed project pageTactiDex access source2026-08-16
    TVL: Touch, Vision, and Language tactile datasetDataset license not stated on the reviewed project pageTVL: Touch, Vision, and Language access source2026-08-19
    CLAMP multimodal haptic dataset tactile datasetDataset license unverified: the official Dataverse route could not be inspected. The code license does not grant data reuse rights.CLAMP multimodal haptic dataset access source2026-10-27
    RoboTacDex: A Dexterous Visual-Tactile-Action Dataset for Humanoid Manipulation tactile datasetDataset-file license not stated. The arXiv article license does not license unreleased dataset files.RoboTacDex: A Dexterous Visual-Tactile-Action Dataset for Humanoid Manipulation access source2026-08-22
    PRISM: Precision and contact-rich Real-world Industrial Skill dataset with Multimodal sensing tactile datasetDataset-file license not verified. The linked ModelScope page did not expose readable terms in this check; paper and code licenses are not data permission.PRISM: Precision and contact-rich Real-world Industrial Skill dataset with Multimodal sensing access source2026-08-22
    HT-Bench tactile datasetDataset, model-weight, and code licenses not stated; the paper is CC BY 4.0, which does not establish a license for unreleased artifact filesHT-Bench access source2026-08-22
    OpenRoboCare expert caregiving demonstrations tactile datasetDocumentation names BSD-3-Clause but additionally limits use to research and education, prohibits re-identification, requires citation and participant privacy, and disallows redistribution without permission. AWS describes academic/non-commercial use. Do not treat this as unrestricted commercial permission.OpenRoboCare expert caregiving demonstrations access source2026-10-27
    T-Rex Tactile-Reactive Dexterous Manipulation Dataset tactile datasetMIT according to the current Hugging Face dataset card. Repository code is also MIT, but upstream robot assets, dependencies, and any files outside the public package retain their own terms.T-Rex Tactile-Reactive Dexterous Manipulation Dataset access source2026-08-22
    UniVTAC Benchmark Dataset tactile datasetMIT as displayed by the current Hugging Face dataset card. The official GitHub repository root LICENSE is separately Apache-2.0; its README’s MIT statement conflicts with that root license and is not used as the repository-license authority.UniVTAC Benchmark Dataset access source2026-09-19
    FreeTacMan tactile datasetMIT in the provider dataset card. The code repository root LICENSE and README identify Apache-2.0 separately from the dataset card. Do not generalize either label to all artifacts or third-party dependencies.FreeTacMan access source2026-08-16
    VTDexManip tactile datasetMIT License for the benchmark code; verify dataset-file terms separatelyVTDexManip access source2026-08-19
    Bench2Dex simulation demonstrations tactile datasetRepository root MIT applies to code; no separate dataset-file license or collection-wide model/asset license was verified.Bench2Dex simulation demonstrations access source2026-09-18
    ObjectFolder Real tactile datasetReview the license bundled with the downloaded ObjectFolder releaseObjectFolder Real access source2026-08-19
    TacClaw-Infinity / Daimon-Infinity public release tactile datasetThe hosting card labels the dataset CC BY-NC-SA 4.0: attribution, noncommercial use and share-alike conditions apply. Public download visibility is not a commercial-use permission.TacClaw-Infinity / Daimon-Infinity public release access source2026-10-16
    UniVTAC Encoder Pretraining Corpus tactile datasetThe provider dataset card labels the repository MIT. This is separate from the code root Apache-2.0 license and does not resolve third-party asset or checkpoint terms.UniVTAC Encoder Pretraining Corpus access source2026-09-19
    DexSuite scripted Lift tactile subset tactile datasetThis pinned dataset card labels the subset Apache-2.0. This is not a blanket license verification for the full corpus or simulator assets.DexSuite scripted Lift tactile subset access source2026-10-31
    ROMI-2K object-interaction dataset tactile datasetUnknown: no dataset-specific reuse license was identified in the reviewed official card or repository. A model-subdirectory Apache-2.0 license does not establish data reuse rights.ROMI-2K object-interaction dataset access source2026-10-12

    Structured dataset explorer

    Filter tactile robotics datasets

    Every row separates reported facts from unavailable fields. “Not stated” means the reviewed primary source did not provide enough evidence to fill that field.

    Latest record review: 2026-11-08 / 27 records

    Free newsletter / signup statusFree sample reportDiscuss a dataset research question

    Download selection checklist

    Showing 27 of 27 datasets

    Select up to three records to compare sensors, tasks, file format, dataset license, download route, intended use and limitations. Your selection stays when filters change. Exports preserve evidence notes; they do not redistribute dataset files.

    Scroll the table horizontally for all fields, or select records to read the comparison above.

    DatasetInstitution / yearRobot / sensorModalities / scaleTasks / objectsFormat / licensePrimary links
    TouchScale human visual-tactile datasetUse the authors’ citation instructions on the primary paper or project page.Registration / approval requiredAccess checked 2026-11-08Checked 2026-11-08: public card and host metadata only. The card still describes 100 hours and an earlier full-release plan; that plan does not verify 500-hour availability. No conditions accepted, gated data downloaded, episode decoded or model trained.Reviewed 2026-11-08TouchScale authors and 2077AI Data Foundation
    2026

    First submitted 2026-10-07

    arXiv v3 (2610.10288), November 2; host API revision 74914cc2e11d0a7f37ddbe5d57db4fe4e14a0a7e; revision-specific README inaccessible

    Data origin: real

    Robot: Human wearable collection; separate downstream robot experiments
    Sensor: Bimanual tactile gloves; Head RGB-D camera; Wrist RGB cameras
    Human tactile observations; Egocentric RGB-D; Wrist RGB; Language instructions
    Reported / previously documented scale: Paper: 500-hour study corpus. Reviewed public card: 100 hours / 15,324 episodes. Neither count was independently recounted from payloads.

    Public-file audit: Public card describes 100 hours and 15,324 episodes, distinct from the paper’s 500-hour study corpus; no payload recount.

    Independent check: Public page and API metadata inspected; host revision 74914cc2e11d0a7f37ddbe5d57db4fe4e14a0a7e. Revision-specific raw README returned HTTP 401. No authenticated files, episode decoding or training.

    Visual-tactile representation learning; Human contact supervision for later robot adaptation
    Objects: Everyday human manipulation; exact contents of gated payloads not inspected.

    Use limitations (editorial): Human contact supervision is not robot-action supervision. The paper limits robot evaluation to one platform and four tasks; adaptation to another embodiment remains a separate experiment.

    Generalization: Human contact supervision is not robot-action supervision. The paper limits robot evaluation to one platform and four tasks; adaptation to another embodiment remains a separate experiment.

    Public card describes WebDataset shards and per-episode HDF5; no payload decoding.
    Dataset-file license: Public card labels data CC BY-NC 4.0; noncommercial conditions and host access terms apply. No conditions accepted in this review.

    Code / source terms: Dataset card: CC BY-NC 4.0; noncommercial conditions apply. Gated access requires acceptance of host conditions and is not unrestricted reuse permission.

    Code license: Separate code terms not verified in this review.

    Asset license: Check object-asset terms separately before reuse.

    Model license: No model-weight terms verified.

    Train / test split: No released split manifest inspected.

    DexSuite scripted Lift tactile subsetUse the authors’ citation instructions on the primary paper or project page.Public file manifestAccess checked 2026-10-31Checked 2026-10-31: public manifest, card, info.json, action_layout.json and collection_qa.json fetched at the pinned revision. Metadata inspected only; no Parquet/video payload download, loader execution, trajectory replay, training or hardware test.Reviewed 2026-10-31DexSuite authors, led by Mila and McGill University collaborators
    2026

    arXiv v1 (2610.37799); subset revision f4be6e95af40843da39cd5a88d8bd736393dd59c

    Data origin: simulation

    Robot: Simulated Franka arm with UMI-style tactile parallel gripper
    Sensor: Simulated FlexiTac-style fingertip pads
    Simulated tactile force grids; Front and wrist RGB video; Robot and object state; Controller actions
    Reported / previously documented scale: Metadata declares 200 episodes and 15,030 frames for this Lift subset. The paper-wide 4,700 episodes are a different scope; payload rows were not independently recounted.

    Public-file audit: 21 listed paths. Metadata declares 200 episodes and 15,030 frames; payload rows were not independently recounted.

    Independent check: Pinned public file manifest, dataset card and JSON metadata inspected. No Parquet/video payload download, loader run, trajectory replay, training or hardware validation.

    Scripted simulated lifting; Contact-conditioned imitation learning research
    Objects: Lift-task objects with randomized initialization, according to the subset card.

    Use limitations (editorial): Simulated contact is not physical sensor measurement. The card records pre-action observations and omits the final post-action state. Define terminal handling and a held-out episode split before training; the declared train split covers all 200 episodes.

    Generalization: Simulated contact is not physical sensor measurement. The card records pre-action observations and omits the final post-action state. Define terminal handling and a held-out episode split before training; the declared train split covers all 200 episodes.

    JSON schema inspected: LeRobot v3.0, 20 Hz, [12,32] float32 left/right force grids and [224,224,3] RGB videos. Action layout has 7 manipulator pose-controller values plus 2 gripper joint targets. Payload decoding is untested.
    Dataset-file license: The pinned subset card labels the data Apache-2.0. Simulator software and third-party asset rights are separate and were not verified.

    Code / source terms: This pinned dataset card labels the subset Apache-2.0. This is not a blanket license verification for the full corpus or simulator assets.

    Code license: Code license not verified; the repository route linked by PyPI returned 404 in this review.

    Asset license: Simulator and third-party asset terms require separate review.

    Train / test split: The inspected metadata declares train 0:200. No held-out evaluation partition or leakage-free split was validated.

    CLAMP multimodal haptic datasetUse the authors’ citation instructions on the primary paper or project page.Download route linked; files not verifiedAccess checked 2026-10-27Checked 2026-10-27: official project and repository link a Dataverse dataset. The data landing route was inaccessible in this review; no verification challenge was solved. File manifest, dataset license and payload remain unverified. No download, installation or experiment performed.Reviewed 2026-10-27Cornell University EmPRISE Lab and paper collaborators
    2025

    CoRL 2025; arXiv v1 (2505.21495)

    Data origin: real

    Robot: Human-operated sensorized reacher-grabber
    Sensor: Active and passive thermal sensors; Normal-force sensing resistor; Contact microphone for vibration; Two IMUs
    Thermal; Normal-force signal; Vibration; Proprioception; Object images; Spoken annotations and derived language labels
    Reported / previously documented scale: Authors report 12.3 million samples across 5,357 household objects from 16 devices and 41 participants. These are paper totals, not independently recounted released files.

    Public-file audit: No object manifest or payload independently inspected; counts are source-reported.

    Independent check: Official documentation review only; no access request, agreement acceptance, dataset download or reproduction.

    Material recognition; Compliance recognition; Multimodal object perception
    Objects: Household objects in the authors' collection.

    Use limitations (editorial): Human collection and author-reported robot fine-tuning are different evidence. Sensor-specific preprocessing and embodiment adaptation are required; RoboSkin has not reproduced perception or manipulation results.

    Generalization: Human collection and author-reported robot fine-tuning are different evidence. Sensor-specific preprocessing and embodiment adaptation are required; RoboSkin has not reproduced perception or manipulation results.

    The repository describes raw data and a filtered NPZ with precomputed visual predictions. The paper uses processed sensor features; its force input is an uncalibrated signal, not a verified Newton-valued release. File schema and payload completeness remain unverified.
    Dataset-file license: Dataset license unverified: the official Dataverse route could not be inspected. The code license does not grant data reuse rights.

    Code / source terms: Dataset license unverified: the official Dataverse route could not be inspected. The code license does not grant data reuse rights.

    Code license: Repository code is labeled BSD-3-Clause; dataset and third-party terms must be checked separately.

    Model license: Model-weight terms not verified in this documentation review.

    Train / test split: Released file-level splits and leakage not verified.

    OpenRoboCare expert caregiving demonstrationsUse the authors’ citation instructions on the primary paper or project page.Download route linked; files not verifiedAccess checked 2026-10-27Checked 2026-10-27: official docs and AWS registry document public unsigned S3 access. This review inspected documentation only, not a live object manifest or payload; no data downloaded and no terms accepted.Reviewed 2026-10-27Cornell University EmPRISE Lab and paper collaborators
    2025

    IROS 2025

    Data origin: real

    Robot: Occupational therapists demonstrating on two manikins; not recorded robot trajectories
    Sensor: Manikin pressure skin: 44 readings per frame; RGB-D cameras; Gaze and pose tracking
    Pressure JSON; RGB-D; Eye gaze; 2D/3D pose; Hierarchical action annotations; First/third-person video
    Reported / previously documented scale: Official sources report 21 occupational therapists, 15 tasks, 315 sessions and 19.8 hours. Counts are provider-reported, not independently audited.

    Public-file audit: No object manifest or payload independently inspected; counts are source-reported.

    Independent check: Official documentation review only; no access request, agreement acceptance, dataset download or reproduction.

    Caregiving activity perception; Human demonstration analysis; Multimodal contact and pose research
    Objects: Caregiving activities performed on manikins, not clinical trials with patients.

    Use limitations (editorial): Manikin demonstrations are not clinical efficacy or safety evidence, and human activity annotations are not robot action supervision. Robot embodiment mapping and patient-context validation remain separate work; RoboSkin has not run these data.

    Generalization: Manikin demonstrations are not clinical efficacy or safety evidence, and human activity annotations are not robot action supervision. Robot embodiment mapping and patient-context validation remain separate work; RoboSkin has not run these data.

    Documentation describes pressure JSON, RGB JPEG, depth PNG, gaze/action CSV and pose JSON. Not every session contains every modality; inspect timestamps and missing files before alignment. Some 3D poses are reconstructed from 2D annotations and may contain noise.
    Dataset-file license: Documentation names BSD-3-Clause but additionally limits use to research and education, prohibits re-identification, requires citation and participant privacy, and disallows redistribution without permission. AWS describes academic/non-commercial use. Do not treat this as unrestricted commercial permission.

    Code / source terms: Documentation names BSD-3-Clause but additionally limits use to research and education, prohibits re-identification, requires citation and participant privacy, and disallows redistribution without permission. AWS describes academic/non-commercial use. Do not treat this as unrestricted commercial permission.

    Code license: Separate software licensing not verified; the dataset documentation is not a blanket code license.

    Model license: Model-weight terms not verified in this documentation review.

    Train / test split: Released file-level splits and leakage not verified.

    EgoExplorer human vision-motion-touch mini releaseUse the authors’ citation instructions on the primary paper or project page.Registration / approval requiredAccess checked 2026-10-26Checked 2026-10-26: the public card describes approval-gated access to 18 episodes / 66 minutes and a separate full-collection waitlist. Toolkit commit 0de464fd576939c053356df9be8be1cabc41c8cc exposes capture and reconstruction code. No access request, agreement acceptance, payload download, toolkit execution or training was performed.Reviewed 2026-10-26InternRobotics and EgoExplorer paper collaborators
    2026

    First submitted 2026-10-13

    arXiv v1 (2610.18872)

    Data origin: real

    Robot: Human wearable collection rig; no verified robot-action training release
    Sensor: Wuji tactile arrays; Egocentric camera; Body trackers and hand tracking
    Egocentric video; Tracker poses; Hand skeletons; Wuji tactile observations; Reconstructed SMPL-X; English step captions
    Reported / previously documented scale: The paper reports a 100-hour corpus. The current card describes an approval-gated mini release of 18 episodes / 66 minutes; these are not the full paper corpus or independently recounted payloads.

    Public-file audit: Card-described mini release: 18 episodes / 66 minutes. Payload files gated and not inspected; distinct from the paper 100-hour corpus.

    Independent check: Public card, paper abstract and toolkit commit inspected. No gated access request, payload download, training or hardware run.

    Human loco-manipulation observation; Multimodal representation and retargeting research
    Objects: Ten task templates across laboratory, household and industrial activities, according to the card.

    Use limitations (editorial): Human observations require embodiment, action-space and sensor mapping before robot policy use. Retargeting demonstrations do not establish robot-action labels or validated transfer to another hand.

    Generalization: Human observations require embodiment, action-space and sensor mapping before robot policy use. Retargeting demonstrations do not establish robot-action labels or validated transfer to another hand.

    The card describes LeRobot v3.0 at 30 Hz, video, five tracker poses, two 21-joint hand skeletons, 24 x 31 Wuji tactile arrays per hand and refined SMPL-X. Reconstructed body geometry is distinct from measured tactile signals. Archive alignment and completeness were not tested.
    Dataset-file license: Dataset card: CC BY-NC 4.0. Noncommercial conditions apply; approval or download access is not commercial permission.

    Code / source terms: Dataset card: CC BY-NC 4.0. Noncommercial conditions apply; approval or download access is not commercial permission.

    Code license: Toolkit code license not verified; do not inherit the dataset license.

    Model license: Model-weight and third-party body-model terms not verified.

    Train / test split: No file-level train/test partition or leakage audit verified.

    TacClaw-Infinity / Daimon-Infinity public releaseUse the authors’ citation instructions on the primary paper or project page.Public file manifestAccess checked 2026-10-16Checked 2026-10-16: public file listings are visible at revision 27b2420cf9f516b3456d01418e0ed01b74c8a2b5. The card describes the first 1,000 hours, while the paper reports a larger corpus. The project page still marks release links Coming soon, so public-host files do not establish full paper-corpus release. We inspected documentation and directory listings only; no bulk download, checksum audit, training or hardware experiment was performed.Reviewed 2026-10-16Daimon Robotics and the TacClaw-Infinity paper collaborators
    2026

    First submitted 2026-10-14

    arXiv v1 (2610.20320)

    Data origin: real

    Robot: TacClaw handheld collection interface; public host also contains DM-DataClaw and DM-DataDex material
    Sensor: Fingertip tactile sensors and visual observation channels described by the authors
    Vision; Tactile observations; Language annotations; End-effector trajectories
    Reported / previously documented scale: The paper reports over 10,000 hours, 1 million episodes and 1 billion frames across more than 1,000 task types. The hosting card separately describes a first 1,000-hour release. Neither the paper totals nor displayed storage bytes are an independently verified count of downloadable, complete episodes.

    Public-file audit: Public directories observed: DM-DataClaw, DataDex_Datasets and TacExo. First 1,000 hours are claimed in the card; complete unique episode count and equivalence to the paper corpus remain unverified.

    Independent check: Paper abstract, hosting card, root and subset directory listings inspected. No payload download, checksum verification, split audit, training or hardware run.

    Contact-rich manipulation demonstrations; Vision-tactile-language-action learning; Cross-embodiment transfer research
    Objects: Diverse manipulation tasks reported by the authors; a deduplicated object inventory for the inspected public files was not verified.

    Use limitations (editorial): The authors report transfer across three heterogeneous robots. RoboSkin has not reproduced that result. Human-collected end-effector trajectories require task, action-space and sensor compatibility checks before use on a different robot.

    Generalization: The authors report transfer across three heterogeneous robots. RoboSkin has not reproduced that result. Human-collected end-effector trajectories require task, action-space and sensor compatibility checks before use on a different robot.

    The card documents a LeRobot-style layout. Public folders include DM-DataClaw, DataDex_Datasets and TacExo. Payload completeness, alignment and file-level train/test membership were not validated.
    Dataset-file license: The hosting card labels the dataset CC BY-NC-SA 4.0: attribution, noncommercial use and share-alike conditions apply. Public download visibility is not a commercial-use permission.

    Code / source terms: The hosting card labels the dataset CC BY-NC-SA 4.0: attribution, noncommercial use and share-alike conditions apply. Public download visibility is not a commercial-use permission.

    Code license: Not verified in this data-access review; dataset terms do not establish code terms.

    Model license: No model-weight license verified in this review.

    Train / test split: No file-level split manifest validated. Do not infer leakage-free splits from task folder names.

    ROMI-2K object-interaction datasetUse the authors’ citation instructions on the primary paper or project page.Announced; release not verifiedAccess checked 2026-10-12Checked 2026-10-12: the official dataset card marks ROMI-2K and ROMA Bench as Coming soon. Checkpoints and one recorded demo scene are visible, but a complete dataset release is not confirmed. A checked release item in the code README conflicts with its coming-soon text. No full dataset download, inference or hardware run was performed.Reviewed 2026-10-12Gaoling School of Artificial Intelligence, Renmin University of China; Beijing Key Laboratory of Research on Large Models and Intelligent Governance; Beijing Academy of Artificial Intelligence; Beijing Jiaotong University; State Key Laboratory of Multimedia Information Processing, Peking University; AresoX
    2026

    Original authors: Ruoxuan Feng; Yutong Chen; Ruihua Song; Huan Yang; Zhongyuan Wang; Guocai Yao; Di Hu

    First submitted 2026-10-03

    arXiv v2 (2610.06955v2)

    Data origin: real

    Robot: Pika Sense handheld collector; xArm 6 with Robotiq 2F-140 gripper for tabletop collection
    Sensor: Two GelSight Mini optical tactile sensors; RGB cameras and microphone; Six-axis force/torque sensor on the robotic tabletop setup only; none on the handheld collector
    RGB; Optical tactile images; Audio; Force/torque in the robotic tabletop subset; Initial scene depth in the tabletop setup; Object-property and content annotations
    Reported / previously documented scale: Paper v2 reports 1,657 handheld object–content combinations and 500 tabletop scenes (400 training, 100 test). These are different counting units, not a verified total of unique trajectories or objects.

    Public-file audit: Complete ROMI-2K data files not confirmed. The reviewed Hugging Face manifest lists model files; the GitHub tree contains one recorded demo scene.

    Independent check: Paper v2, official card, Hugging Face manifest and complete GitHub tree inspected. Coming-soon text conflicts with a checked README release item. No complete payload download, inference, training or hardware run.

    Object-centric active perception; Material and content inference; Multisensory property reasoning
    Objects: Everyday objects and object–content combinations. Most tabletop collection objects are reused from the handheld subset; the test set adds 56 unseen objects according to the paper.

    Use limitations (editorial): The paper separates handheld training interactions from tabletop evaluation, with six predefined primitives on one arm–gripper setup. ROMA uses mean gravity force as text, not the full recorded force/torque stream. Dataset coverage is not proof of general manipulation-policy transfer.

    Generalization: The paper separates handheld training interactions from tabletop evaluation, with six predefined primitives on one arm–gripper setup. ROMA uses mean gravity force as text, not the full recorded force/torque stream. Dataset coverage is not proof of general manipulation-policy transfer.

    Visual, optical tactile and audio interactions; force/torque only in robotic tabletop collection. Full archive schema and completeness unverified.
    Dataset-file license: Unknown: no dataset-specific terms identified. The Apache-2.0 text in a model subdirectory does not license the dataset.

    Code / source terms: Unknown: no dataset-specific reuse license was identified in the reviewed official card or repository. A model-subdirectory Apache-2.0 license does not establish data reuse rights.

    Code license: Unknown: no license file was identified in the reviewed GitHub tree.

    Model license: Apache-2.0 text exists in the hosted ROMA-Qwen2.5-Omni-7B model directory; it does not license ROMI-2K or establish terms for every component.

    Train / test split: Paper reports 400 training and 100 test tabletop scenes, with 56 additional unseen test objects. Released file-level split membership and leakage are unverified.

    Bench2Dex simulation demonstrationsUse the authors’ citation instructions on the primary paper or project page.Public file manifestAccess checked 2026-09-18Public, ungated provider manifest verified on 2026-09-18. Dataset payloads were not downloaded; completeness, checksums, schema validity and permission to reuse data remain unverified. Paper-scale demonstrations and hosted file counts are different units.Reviewed 2026-09-18Shanghai Jiao Tong University; Fudan University; The University of Hong Kong; Inspire Robots; Zhongguancun Academy; COWARobot Co. Ltd; Nanyang Technological University
    2026

    Original authors: Zhenjie Yang; Yideng Zhang; Dongjie Zhang; Chenyu Jiang; Xianshuai Liu; Yufeng Li; Zuhao Ge; Xingyu Jiao; Zheng Zhang; Kaiyu He; He Wang; Yuwen Zhong; Yi Deng; Muyun Jiang; Xianliang Huang; Haisheng Su; Donghang Zhang; Jian Zhang; Xue Yang; Hongyang Li; Zuxuan Wu; Yu-Gang Jiang; Xiaosong Jia; Junchi Yan

    First submitted 2026-09-14

    arXiv v1 (2609.15726v1)

    Data origin: simulation

    Robot: 12 simulated bimanual dexterous arm–hand embodiments; 26 task–embodiment settings
    Sensor: Surface-aligned simulated contact geometry; not measured output from a physical tactile sensor
    Multi-view RGB; Joint state; Commanded actions; Object state; Surface-aligned tactile maps; Depth, 2D/3D boxes and occupancy when recorded
    Reported / previously documented scale: About 1,300 human-teleoperated simulation demonstrations in the v1 paper. Public hosting separately lists 5,200 HDF5 file paths; file paths are not a verified count of unique demonstrations.

    Public-file audit: 5,201 listed paths: 5,200 HDF5 files plus .gitattributes. Unique demonstrations, frames and bytes were not independently recounted.

    Independent check: Official project, v1 paper, repository root MIT license and provider manifests inspected. No payload download, checksum or HDF5 validation, simulation run or independent reproduction.

    Bimanual tool use; Articulated-object interaction; Multi-stage manipulation; Controlled scene perturbations
    Objects: Task-specific tools, containers, articulated objects and assembly items; a deduplicated released object count is unknown.

    Use limitations (editorial): Simulation demonstrations support controlled perturbation studies. Multiple supported hands do not establish zero-shot cross-hand or physical-sensor transfer.

    Generalization: Four scene-perturbation channels (None, Equi., Inv., Full). Multiple hand support is not zero-shot cross-hand transfer; no physical-sensor or sim-to-real validation claimed.

    HDF5 episodes. The documented schema aligns frame_index, timestamp_ns and sim_step; action commands precede the next simulator step and observations follow it. Actions include names, control mode, validity and source. Supported fields are not guaranteed in every payload; README says depth is omitted in the current storage workflow.
    Dataset-file license: Unknown: no dataset card or dataset-file license found in the reviewed provider manifest. Code MIT is not a dataset license.

    Code / source terms: Repository root MIT applies to code; no separate dataset-file license or collection-wide model/asset license was verified.

    Code license: MIT at repository revision f96a8b2b4eb475483af66e9e03916b35bc43f1be.

    Asset license: Unknown collection-wide terms; upstream robot and object asset licenses require separate review.

    Model license: Unknown: no top-level weight license or model card verified; base-model terms may differ.

    Train / test split: Paper defines deterministic task-partitioned evaluation seeds and None / Equi. / Inv. / Full perturbation channels, with 50 rollouts per task/channel/policy. Released training/validation/test split membership and leakage remain unknown; a directory name is not a verified split.

    UniVTAC Encoder Pretraining CorpusUse the authors’ citation instructions on the primary paper or project page.Public file manifestAccess checked 2026-09-19Public contact-file listing verified on September 19, 2026. No standalone download establishing exact equivalence to the 205,826-sample paper corpus was verified; HDF5 payloads were not downloaded.Reviewed 2026-09-19ScaleLab, Shanghai Jiao Tong University; D-Robotics; ViTai Robotics; The University of Hong Kong; Nanjing University; Shenzhen University; Wuhan University; Fudan University; Tsinghua University
    2026
    Robot: UniVTAC simulated gripper contact-generation setup; not a physical-robot trajectory corpus
    Sensor: Simulation-synthesized marker-based visuo-tactile observations
    Raw tactile RGB image with markers; Marker-free tactile RGB image; Gelpad depth map; Projected 2D marker coordinates; Object pose as 3D translation plus quaternion
    Reported / previously documented scale: The paper reports 205,826 simulation samples across 14 geometric primitives. The September 19 provider inventory separately lists 638 contact-pretraining HDF5 files across 14 non-empty shape directories. File counts are not frame counts or proof of equivalence to the paper corpus; neither collection is the task benchmark, 400 paper policy-training trajectories or 450 physical demonstrations.

    Public-file audit: 638 listed contact HDF5 files across 14 non-empty shape directories. Equivalence to the paper's 205,826 contact samples is unverified; files are not frames.

    Independent check: Pinned provider file listing and code documentation inspected. No payload download, array validation, frame recount, training or hardware run.

    Shape reconstruction; Contact-deformation prediction; Marker-position prediction; Relative object-pose regression; Tactile representation pretraining
    Objects: Fourteen geometric primitives ranging from convex forms such as spheres and cones to non-convex forms such as stars and cross-shapes, mounted on a standardized prism base for simulated contact generation.

    Use limitations (editorial): Use for simulated contact representations. Contact samples are not policy trajectories; public file counts do not establish equivalence to the paper corpus.

    Public contact/<shape>/hdf5/ paths; the paper describes marked and marker-free RGB, depth, projected markers and 7D pose. Released payload arrays and paper-corpus equivalence remain unverified.
    Dataset-file license: MIT in the provider dataset-card metadata; third-party assets and checkpoint terms require separate review.

    Code / source terms: The provider dataset card labels the repository MIT. This is separate from the code root Apache-2.0 license and does not resolve third-party asset or checkpoint terms.

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    UniVTAC Benchmark DatasetUse the authors’ citation instructions on the primary paper or project page.Public file manifestAccess checked 2026-09-19Public file listing verified on September 19, 2026. Match main/Isaac Sim 4.5 to isaac45, or isaac51/Isaac Sim 5.1 to isaac51. The official repository states task data are not cross-compatible and released policy checkpoints target 4.5. See the research guide for the downloadable 30-row inventory and payload-verification limits.Reviewed 2026-09-19ScaleLab, Shanghai Jiao Tong University; D-Robotics; ViTai Robotics; The University of Hong Kong; Nanjing University; Shenzhen University; Wuhan University; Fudan University; Tsinghua University
    2026
    Robot: Simulated Franka Panda with a parallel-jaw gripper and bilateral simulated GelSight Mini sensors
    Sensor: Bilateral simulated GelSight Mini sensors through UniVTAC and TacEx
    Head RGB image; Wrist RGB image; Robot end-effector and joint state; Robot action and episode metadata; Bilateral tactile RGB; Bilateral marker-overlaid tactile RGB; Bilateral tactile depth; Bilateral marker coordinates; Bilateral tactile-sensor pose
    Reported / previously documented scale: September 19 file inventory at revision e1aee7b0c95543b535e0146b2de3ee1bc6ddaabd lists 800 HDF5 paths under isaac45 and a separate 800 under isaac51, 100 per task. The historical August revision 172331dbbce95bc04c3e59b22f32dc72ba5561ae contained 800 HDF5 episodes, approximately 125.43 GB, with metadata marking 763 successful and 37 non-success episodes. Historical size/outcome counts were not revalidated for the current versions. Task files are distinct from encoder samples, policy-training trajectories and physical demonstrations.

    Public-file audit: 2,344 paths in the entire provider repository, shared with the contact and checkpoint collections. Task-only inventory: 800 HDF5 paths in isaac45 and 800 in isaac51, 100 per task in each version. Bytes, frames and unique demonstrations were not recounted.

    Independent check: Official hosting API and file manifest inspected at the pinned revision. No dataset payloads downloaded; no checksums, schema validation or training run.

    Lift Bottle; Pull-out Key; Lift Can; Put Bottle in Shelf; Insert Hole; Insert HDMI; Insert Tube; Grasp Classify
    Objects: Task-specific bottles, cans, keys and locks, shelf-placement objects, peg-and-hole components, HDMI connectors and ports, tube fixtures, and tactile-classification shapes as defined by the eight benchmark tasks.

    Use limitations (editorial): Use the matching simulator version and task package. Simulated GelSight channels do not establish performance on a physical GelSight sensor.

    Versioned isaac45/<task>/hdf5/ and isaac51/<task>/hdf5/ paths with task metadata. Documentation describes head/wrist observations, robot state and bilateral tactile RGB, marker, depth and pose. Version compatibility and payload validity were not executed; the prior Dataset Viewer error was not retested.
    Dataset-file license: mit in the provider dataset-card metadata at the pinned revision; individual asset terms were not inspected.

    Code / source terms: MIT as displayed by the current Hugging Face dataset card. The official GitHub repository root LICENSE is separately Apache-2.0; its README’s MIT statement conflicts with that root license and is not used as the repository-license authority.

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    T-Rex Tactile-Reactive Dexterous Manipulation DatasetUse the authors’ citation instructions on the primary paper or project page.Public file manifestAccess checked 2026-09-13The 5,464-episode, approximately 50-hour LeRobot v3.0 subset is publicly browsable and downloadable on Hugging Face. The official repository provides a dataset quickstart and states that the full 100-hour pretraining and midtraining corpus is not part of the current release.Reviewed 2026-08-22UC Berkeley; NVIDIA; Stanford University; Panasonic; Sapienza University of Rome; ItalAI
    2026
    Robot: Fixed-base bimanual Dexmate Vega-1 with two 22-DoF Sharpa Wave dexterous hands
    Sensor: ZED X Mini head camera; Two ZED X One S wrist cameras; Ten fingertip image-based tactile sensors, five per hand
    Head RGB video; Bilateral wrist RGB video; Robot state; Current and target joint positions; Raw fingertip grayscale tactile video; Estimated tactile deformation maps; Estimated 6D fingertip wrenches; Frame and episode metadata
    Reported / previously documented scale: The paper and official repository report a complete 100-hour collection with 7,700-plus trajectories, 22 motor primitives, and more than 200 objects. The public Hugging Face card specifies 5,464 episodes, 5,473,459 frames at 30 FPS (approximately 50 hours), 5,370 language-annotated trajectories, 22 motor primitives, and 207 objects. The public subset must not be described as the complete 100-hour corpus.

    Public-file audit: 7,903 paths in the provider file manifest, including metadata. Bytes, frames and episodes were not independently recounted.

    Independent check: Official hosting API and file manifest inspected at the pinned revision. No dataset payloads downloaded; no checksums, schema validation or training run.

    Bimanual tactile-reactive motor primitives; Dexterous manipulation; Force-sensitive contact; Deformable-object interaction; Vision-language-action midtraining
    Objects: The current public dataset card reports 207 unique canonical objects and publishes composition metadata, but RoboSkin did not independently audit every object label or collapse them into an inferred category taxonomy.

    Use limitations (editorial): Use for the recorded bimanual embodiment and tactile streams. The public subset is distinct from the complete paper corpus; transfer to another hand requires evaluation.

    LeRobotDataset v3.0 with head and bilateral wrist videos, robot state, current and target joint positions, ten fingertip tactile streams, deformation maps, 6D wrenches, and episode metadata. The official repository provides selective download, inspection, and 3D replay examples.
    Dataset-file license: mit in the provider dataset-card metadata at the pinned revision; individual asset terms were not inspected.

    Code / source terms: MIT according to the current Hugging Face dataset card. Repository code is also MIT, but upstream robot assets, dependencies, and any files outside the public package retain their own terms.

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    EgoTouchUse the authors’ citation instructions on the primary paper or project page.Public file manifestAccess checked 2026-09-13The official Hugging Face repository exposes scene folders and split metadata. The official GitHub repository says the upload is in progress and currently accessible files may be incomplete, so completeness and the exact version must be checked before reuse.Reviewed 2026-08-22Harbin Institute of Technology, Shenzhen; Meituan Academy of Robotics; Tsinghua Shenzhen International Graduate School, Tsinghua University
    2026
    Robot: Human-wearable bimanual collection; no robot platform is used in the EgoTouch capture pipeline
    Sensor: Head-mounted wide-angle camera; Dual wrist cameras; Bimanual pressure-sensing gloves with dense palm pressure maps; Bimanual 3D hand-pose capture
    Head-mounted egocentric RGB; Dual wrist-mounted RGB; Bimanual 3D hand pose with 42 joints; Continuous bilateral tactile pressure maps; Camera and wrist poses; Frame timestamps
    Reported / previously documented scale: 208 manipulation tasks across 1,891 episodes in indoor and outdoor environments. The official repository reports approximately 2 million frames and more than 1,000 objects; those repository statistics are author-reported and the repository still warns that the dataset upload may be incomplete.

    Public-file audit: 22,897 paths in the provider file manifest, including metadata. Bytes, frames and episodes were not independently recounted.

    Independent check: Official hosting API and file manifest inspected at the pinned revision. No dataset payloads downloaded; no checksums, schema validation or training run.

    Bimanual hand-object manipulation; Egocentric vision-to-touch prediction; Contact prediction under visual occlusion; Multi-view tactile estimation
    Objects: More than 1,000 objects across home, office, outdoor, retail, and workbench scenes are reported by the official repository; a complete reviewed object taxonomy was not verified.

    Use limitations (editorial): Human glove pressure and hand pose can inform representation or retargeting research. They are not robot action labels or calibrated signals for a different sensor.

    The official repository documents variable-length 30 FPS HDF5 episodes containing three 640 x 480 RGB views, bimanual hand-pose arrays, bilateral pressure grids, camera poses, masks when available, metadata, and timestamps. The Hugging Face repository also exposes scene folders and split.json.
    Dataset-file license: Unknown: a dataset-file license has not been verified separately from paper or code terms.

    Code / source terms: A separate dataset-file license was not verified. The official GitHub repository is MIT licensed, but that project-code license is not generalized here to the hosted EgoTouch data files.

    Train / test split: Split-named files listed: split.json. Contents and leakage were not validated.

    PRISM: Precision and contact-rich Real-world Industrial Skill dataset with Multimodal sensingUse the authors’ citation instructions on the primary paper or project page.Download route linked; files not verifiedAccess checked 2026-11-04Access checked November 4, 2026: the official project links a ModelScope dataset route, superseding the August 22 disabled-button observation. The target page did not expose a readable manifest or dataset-file terms. Download route linked; files, payload completeness, reuse license and loader execution remain unverified. Scientific-source fields retain their original review date.Reviewed 2026-08-22State Key Laboratory of General Artificial Intelligence, School of Intelligence Science and Technology, Peking University; Delta Intelligence; PKU-Wuhan Institute for Artificial Intelligence; Hubei Humanoid Robot Innovation Center Co., Ltd.; China Academy of Information and Communications Technology
    2026
    Robot: Dual Franka Emika Panda tracker platform; Bimanual Realman RM75-6F exoskeleton platform; LEJU upper-body humanoid VR platform
    Sensor: Multi-view RGB-D cameras; 6DoF force/torque sensing; Unnamed visuotactile gripper and visuotactile dexterous hand on a subset of Franka episodes
    Multi-view RGB; Depth; Visuotactile image when available; 6DoF force/torque when available; Robot joint angle and torque; End-effector Cartesian pose; Gripper state; Calibration and timestamps; Human control signal
    Reported / previously documented scale: 5,000+ robot trajectories paired with 5,000 human demonstrations; 45+ hours; 25+ industrial tasks; approximately 27M images across visual and visuotactile streams. The paper does not disclose the number or share of tactile-equipped episodes.

    Public-file audit: Unknown: no independently recounted public-file total in this record.

    Independent check: Official project data link checked. Linked ModelScope page did not expose readable file manifest or terms; no payload download, checksum, loader, training or hardware test.

    Electronic component plug and unplug; Caliper packaging; Conveyor sorting; NIST Assembly Task Board operations; Industrial installation and assembly; Long-horizon compositional procedures
    Objects: Industrial components, assembly hardware, packaging items, conveyor-sorted materials, and task-board objects; the source does not provide one complete released object inventory.

    Use limitations (editorial): Contact-rich industrial collection is task and hardware dependent. Check the released action schema, force conventions and access conditions before planning policy training.

    The paper describes timestamped episodes in a cross-platform schema and conversion to LeRobot v3.0 for its experiments. The current linked host did not expose a readable manifest or payload in this review; final file layout and decoding remain unverified.
    Dataset-file license: Unknown: a dataset-file license has not been verified separately from paper or code terms.

    Code / source terms: Dataset-file license not verified. The linked ModelScope page did not expose readable terms in this check; paper and code licenses are not data permission.

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    SoftVTBenchUse the authors’ citation instructions on the primary paper or project page.Public file manifestAccess checked 2026-09-13The official repository links public code and the Hugging Face dataset mirror. The Hugging Face card reviewed on 2026-08-22 documents 4,000 hosted demonstrations, while the older GitHub README still lists 1,628 demonstrations and 33 assets. Pin the exact dataset revision because the first-party release documents currently disagree.Reviewed 2026-08-22Tuojing Intelligence; Tsinghua University; King's College London; Southeast University; Stevens Institute of Technology; Hong Kong University of Science and Technology (Guangzhou); University of Manchester; Simple AI; Imperial College London; Carnegie Mellon University; Zhejiang University; Beihang University; University of Hong Kong
    2026
    Robot: Simulated Franka arm with Panda parallel-jaw gripper in Isaac Sim and Isaac Lab
    Sensor: Simulated bilateral GelSight Mini profiles rendered through TacEx, Taxim, and FOTS
    Multi-view RGB; Dual-finger tactile RGB; Tactile marker motion; Proprioception; Language; Continuous and binary gripper actions; Evaluator-only FEM state
    Reported / previously documented scale: Latest paper: 4,000 expert demonstrations across 40 tasks and more than 50 assets. Hugging Face dataset-card revision fd2793a documents four subsets with 10 tasks and 100 successful demonstrations each, or 4,000 demonstrations total. The GitHub README last changed 2026-07-22 still lists an earlier 1,628-demonstration, 33-asset state, so the first-party release documents are not synchronized.

    Public-file audit: 16,060 paths in the provider file manifest, including metadata. Bytes, frames and episodes were not independently recounted.

    Independent check: Official hosting API and file manifest inspected at the pinned revision. No dataset payloads downloaded; no checksums, schema validation or training run.

    Deformable-object grasp and placement; Rigid-twin matched control tasks; Object-variation generalization; Spatial-variation generalization; Vision-only and visuo-tactile policy comparison
    Objects: Ten volumetric deformable manipulated-object designs and matched rigid twins; the latest paper reports more than 50 assets across the complete benchmark inventory.

    Use limitations (editorial): Soft-contact experiments depend on the reported objects, deformation and sensor setup. Do not extend reported success to arbitrary materials or contact forces.

    HDF5 trajectories with synchronized policy observations, robot actions, simulator state, and evaluator fields, plus third-person, wrist, and bilateral tactile MP4 videos documented by the current dataset card.
    Dataset-file license: apache-2.0 in the provider dataset-card metadata at the pinned revision; individual asset terms were not inspected.

    Code / source terms: Apache-2.0 as shown on the current Hugging Face dataset card; upstream simulator, robot, tactile-runtime, and third-party asset terms may differ

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    RoboTacDex: A Dexterous Visual-Tactile-Action Dataset for Humanoid ManipulationUse the authors’ citation instructions on the primary paper or project page.Announced; release not verifiedAccess checked 2026-08-22The June 30, 2026 v1 paper says the dataset will be open-sourced soon. No official project page, repository, downloadable package, or dataset-file license was verified on 2026-08-22; record access as announced, not released.Reviewed 2026-08-22Fudan University; ByteDance Intelligent Creation; Shanghai AI Laboratory; The Chinese University of Hong Kong
    2026
    Robot: Unitree G1 humanoid with dual arms totaling 14 DoF and two BrainCo Revo2 Tactile dexterous hands totaling 12 hand DoF as counted by the paper; lower limbs and waist fixed during collection
    Sensor: Four RGB-D cameras, including two Intel RealSense D435i cameras for the head and third-person views; Two BrainCo Revo2 Tactile dexterous hands measuring fingertip normal force, tangential force and direction, and self-capacitance proximity
    Four-view RGB at 640 x 480; Four-view depth at 640 x 480; Arm and finger joint state; Arm and finger joint action; Bilateral fingertip normal force; Bilateral fingertip tangential force and direction; Self-capacitance proximity; Natural-language semantic annotation
    Reported / previously documented scale: More than 6,000 physical-robot trajectories totaling approximately 25 hours, covering 19 tasks, an author-reported 23 skills, and 22 objects. Figure 4 exposes 22 discernible atomic-skill labels, so the 23-skill headline has an unresolved internal source discrepancy. Trajectories are recorded at 30 Hz; tactile and dexterous-hand joint-state messages are published at 100 Hz and recorded locally at 30 Hz with the other modalities.

    Public-file audit: Unknown: no independently recounted public-file total in this record.

    Independent check: Existing source review only. No new dataset-file download, checksum validation, training run, or hardware test in this audit.

    Basic grasping, placing, and pushing; Articulated-object manipulation; Dual-arm collaborative manipulation; Fine manipulation; Humanoid-interactive manipulation; Pick and place a pear; Turn a page; Insert a book into a document bag; Unscrew a bottle cap
    Objects: Twenty-two objects spanning articulated, constrained, container, functional, and graspable categories, including kitchen food and office supplies as well as rigid and deformable objects. The paper does not publish a complete machine-readable object inventory.

    Use limitations (editorial): A described collection is not a verified downloadable training package. Confirm release, action schema and dataset terms before committing to reproduction.

    The v1 paper describes 30 Hz synchronized trajectories containing four 640 x 480 RGB-D views, arm and finger states and actions, bilateral tactile signals, and semantic annotations. It does not state the public package format, directory schema, compression, checksums, or train-validation-test file layout.
    Dataset-file license: Unknown: a dataset-file license has not been verified separately from paper or code terms.

    Code / source terms: Dataset-file license not stated. The arXiv article license does not license unreleased dataset files.

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    HT-BenchUse the authors’ citation instructions on the primary paper or project page.Announced; release not verifiedAccess checked 2026-08-22The v2 preprint says the authors will release all data, evaluation protocols, pretrained weights, and training/testing scripts. No dedicated downloadable dataset package, official repository, or artifact license was verified on 2026-08-22; record access as announced, not released.Reviewed 2026-08-22Beihang University; Rimbot; ShanghaiTech University; Tsinghua University; Chinese Academy of Sciences; BUPT
    2026
    Robot: Single reported dexterous full-hand collection and downstream-evaluation platform; hardware model not stated in the reviewed HTML
    Sensor: Single reported full-hand tactile sensing pipeline; hardware model not stated in the reviewed HTML
    Egocentric RGB; Full-hand tactile pressure maps
    Reported / previously documented scale: 10M RGB frames; 7.8M tactile frames; 226 tasks

    Public-file audit: Unknown: no independently recounted public-file total in this record.

    Independent check: Existing source review only. No new dataset-file download, checksum validation, training run, or hardware test in this audit.

    Tactile similarity retrieval; Masked tactile inpainting; Vision-to-tactile synthesis; Tactile frame prediction
    Objects: Home, electronics workbench, chemistry lab, retail, workbench, outdoor, and other scenes; object count not stated.

    Use limitations (editorial): Full-hand prediction and retrieval tasks do not by themselves validate closed-loop control. Check whether data, splits and evaluation code are actually released.

    Synchronized egocentric RGB and full-hand tactile maps; the benchmark pipeline normalizes tactile maps to 224 x 224.
    Dataset-file license: Unknown: a dataset-file license has not been verified separately from paper or code terms.

    Code / source terms: Dataset, model-weight, and code licenses not stated; the paper is CC BY 4.0, which does not establish a license for unreleased artifact files

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    RCT: Robotic Contact TactileUse the authors’ citation instructions on the primary paper or project page.Download route linked; files not verifiedAccess checked 2026-08-16Dataset, split tools, and evaluation code are publicly linked by the official project page.Reviewed 2026-08-16TU Dresden; ScaDS.AI Dresden/Leipzig; LASR Lab
    2026
    Robot: Robot arm with rotating three-sensor adapter
    Sensor: DIGIT vision-based tactile sensor
    Tactile image; Material RGB image; Language descriptors; Normal force; Indentation depth
    Reported / previously documented scale: 29,279 tactile frames; 1,832 contact sequences; 122 materials; 3 DIGIT sensors

    Public-file audit: Unknown: no independently recounted public-file total in this record.

    Independent check: Existing source review only. No new dataset-file download, checksum validation, training run, or hardware test in this audit.

    Material generalization; Touch-to-text retrieval; Touch-to-vision retrieval; Sensor-disjoint evaluation
    Objects: 122 industrial reference materials in 7 categories.

    Use limitations (editorial): Use for material/contact representations and the documented sensor-disjoint protocol. Contact sequences do not supply a general manipulation-policy benchmark.

    Ordered contact sequences with material, category, sensor, position, depth, force, image, and descriptor metadata.
    Dataset-file license: CC BY 4.0 dataset; Apache-2.0 applies to code, according to the existing project review.

    Code / source terms: CC BY 4.0 dataset; Apache-2.0 code

    Train / test split: Existing source review documents split tools. File-level membership and train/test leakage were not independently validated.

    TactiDexUse the authors’ citation instructions on the primary paper or project page.Access unknownAccess checked 2026-08-16The official project page documents the benchmark and demonstrations; no separate dataset download URL was verified.Reviewed 2026-08-16ShanghaiTech University; InstAdapt
    2026
    Robot: Human demonstration capture; Bimanual Franka Inspire deployment
    Sensor: Whole-hand tactile glove
    Whole-hand pressure map; Hand kinematics; Object 6D pose; Language description; Task-phase annotation
    Reported / previously documented scale: Not stated on the reviewed project page

    Public-file audit: Unknown: no independently recounted public-file total in this record.

    Independent check: Existing source review only. No new dataset-file download, checksum validation, training run, or hardware test in this audit.

    Single-hand dexterous manipulation; Bimanual manipulation; Human-to-robot skill transfer
    Objects: Household and tool objects shown in project demonstrations; total object count not stated.

    Use limitations (editorial): Human demonstration and hand-object annotations require retargeting for robot control. Download availability and dataset-file terms must be established separately.

    Synchronized tactile pressure, hand-object kinematics, 6D pose, language, and hierarchical task annotations.
    Dataset-file license: Unknown: a dataset-file license has not been verified separately from paper or code terms.

    Code / source terms: Dataset license not stated on the reviewed project page

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    FreeTacManUse the authors’ citation instructions on the primary paper or project page.Public file manifestAccess checked 2026-09-13The official project page links the dataset, code, hardware guide, and mirror.Reviewed 2026-08-16Shanghai Innovation Institute; The University of Hong Kong; Shanghai Jiao Tong University; Fudan University
    2025
    Robot: Robot-free wearable gripper; Piper; Franka
    Sensor: Modular LED-based visuo-tactile sensor
    Wrist RGB video; Visuo-tactile video; Tool-center-point pose; Gripper distance
    Reported / previously documented scale: More than 3M visuo-tactile image pairs; more than 10K trajectories; 50 tasks

    Public-file audit: 24,915 paths in the provider file manifest, including metadata. Bytes, frames and episodes were not independently recounted.

    Independent check: Official hosting API and file manifest inspected at the pinned revision. No dataset payloads downloaded; no checksums, schema validation or training run.

    Contact-rich demonstrations; Imitation learning; Tactile pretraining; Visuo-tactile manipulation
    Objects: Fifty contact-rich task categories; examples include fragile handling, insertion, stamping, texture classification, and calligraphy.

    Use limitations (editorial): Human-collected tactile demonstrations need an explicit mapping to the target robot, action space and sensor. Collection convenience is not transfer evidence.

    MP4 wrist and tactile videos plus timestamped trajectory files containing TCP pose, quaternion, Euler angles, and gripper distance.
    Dataset-file license: mit in the provider dataset-card metadata at the pinned revision; individual asset terms were not inspected.

    Code / source terms: MIT in the provider dataset card. The code repository root LICENSE and README identify Apache-2.0 separately from the dataset card. Do not generalize either label to all artifacts or third-party dependencies.

    Code license: Apache-2.0 in the repository root LICENSE; the README agrees and qualifies this as unless specified otherwise. Documentation reviewed November 17, 2026; no code execution.

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    Humanoid Visual-Tactile-Action DatasetUse the authors’ citation instructions on the primary paper or project page.Access unknownAccess checked 2026-08-16The paper describes the dataset; no official public download URL was verified on 2026-08-16.Reviewed 2026-08-16Gwangju Institute of Science and Technology
    2025
    Robot: Humanoid teleoperation platform; Two Inspire RH56-DFX dexterous hands
    Sensor: 1,062 tactile sensors per hand; Piezoresistive tactile carpet
    Egocentric RGB; Third-person RGB; Dense tactile pressure; Arm and finger proprioception; Robot action; External pressure heatmap
    Reported / previously documented scale: 101.9K synchronized samples; approximately 77-80 episodes per task

    Public-file audit: Unknown: no independently recounted public-file total in this record.

    Independent check: Existing source review only. No new dataset-file download, checksum validation, training run, or hardware test in this audit.

    Towel strong pressure; Towel weak pressure; Sponge strong pressure; Sponge weak pressure
    Objects: Two deformable soft objects (towel and sponge), with rigid-object comparison data described in the paper.

    Use limitations (editorial): Full-hand coverage is tied to the collection embodiment. Pressure maps do not establish calibrated shear, fingertip geometry or another hand’s control performance.

    Synchronized visual, 2,124-channel hand tactile, proprioceptive, action, and external pressure signals; public file format not stated.
    Dataset-file license: Unknown: a dataset-file license has not been verified separately from paper or code terms.

    Code / source terms: Dataset access terms not stated on the reviewed paper page

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    Sparsh-X Multisensory Touch ResourceUse the authors’ citation instructions on the primary paper or project page.Access unknownAccess checked 2026-08-16The paper documents the training resource; a dedicated public dataset URL was not verified.Reviewed 2026-08-16FAIR at Meta; University of Washington; Carnegie Mellon University
    2025
    Robot: Robot manipulation platforms used for insertion and in-hand rotation
    Sensor: Digit 360
    Tactile image; Audio; Motion; Pressure
    Reported / previously documented scale: Approximately 1M unlabeled contact-rich interactions

    Public-file audit: Unknown: no independently recounted public-file total in this record.

    Independent check: Existing source review only. No new dataset-file download, checksum validation, training run, or hardware test in this audit.

    Physical-property inference; Plug insertion; In-hand rotation; Tactile adaptation
    Objects: Diverse manipulation interactions; object count and category inventory not stated on the reviewed paper page.

    Use limitations (editorial): Check which data, encoders and sensor modalities are included in the selected release. A multisensory resource does not guarantee a shared schema or cross-sensor accuracy.

    Four synchronized Digit 360 modalities; public file format not stated on the reviewed paper page.
    Dataset-file license: Unknown: a dataset-file license has not been verified separately from paper or code terms.

    Code / source terms: Dataset license and download URL not stated on the reviewed paper page

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    Touch and GoUse the authors’ citation instructions on the primary paper or project page.Download route linked; files not verifiedAccess checked 2026-08-19The paper and project describe a human-collected in-the-wild visuo-tactile dataset; verify the current download terms before reuse.Reviewed 2026-08-19University of Michigan; Carnegie Mellon University
    2022
    Robot: Human-operated GelSight collection device
    Sensor: GelSight
    Egocentric RGB video; GelSight tactile video; Material labels
    Reported / previously documented scale: Approximately 13.9K detected touches; approximately 3,971 object instances; 20 material categories

    Public-file audit: Unknown: no independently recounted public-file total in this record.

    Independent check: Existing source review only. No new dataset-file download, checksum validation, training run, or hardware test in this audit.

    Visuo-tactile feature learning; Material recognition; Tactile-driven image stylization; Future tactile prediction
    Objects: Rigid and deformable objects across indoor and outdoor environments; the paper reports 20 material categories.

    Use limitations (editorial): Use aligned human-collected visual and tactile observations for perception. They do not provide a standardized robot-action or closed-loop slip benchmark.

    Synchronized egocentric video and GelSight tactile recordings with detected-touch timing and material annotations.
    Dataset-file license: Unknown: a dataset-file license has not been verified separately from paper or code terms.

    Code / source terms: Dataset license not stated on the reviewed paper page

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    TVL: Touch, Vision, and LanguageUse the authors’ citation instructions on the primary paper or project page.Public file manifestAccess checked 2026-11-15The official repository links TVL-revise and reports that it fixes swapped tactile/image folders in the original upload. The revised host manifest is public; the correction has not been independently validated by loading image pairs. Dataset-file reuse terms remain unverified.Reviewed 2026-08-19UC Berkeley; Meta AI Research; TU Dresden; CeTI
    2024
    Robot: Handheld 3D-printed HCT collection device; Robot-collected SSVTP subset
    Sensor: DIGIT
    Tactile RGB image; RGB image; English tactile description
    Reported / previously documented scale: 43,741 in-contact image-touch pairs; HCT contributes 39,154 pairs and SSVTP contributes 4,587 pairs

    Public-file audit: 10 paths in the pinned TVL-revise host manifest: 8 split-archive parts, README.md and .gitattributes. Images, frames and bytes were not independently recounted.

    Independent check: Upstream correction notice, revised dataset card and public host manifest inspected at pinned revisions. The folder-swap fix is source-reported; no archive download, extraction, image-pair validation, loader execution or training run.

    Tactile-semantic classification; Touch-vision alignment; Touch-language alignment; Tactile description generation
    Objects: In-the-wild surfaces and objects plus the SSVTP laboratory subset; a total object-category inventory is not stated on the reviewed project page.

    Use limitations (editorial): Use touch, vision and language for multimodal representations. Language alignment does not establish manipulation success, force calibration or action supervision.

    Temporally aligned touch, vision, and open-vocabulary language examples; 10% of labels are human annotations and 90% are VLM-generated in the paper.
    Dataset-file license: Unknown: the inspected TVL-revise card and host metadata do not specify dataset-file terms. The upstream code license is not treated as a dataset license.

    Code / source terms: Dataset license not stated on the reviewed project page

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    ObjectFolder RealUse the authors’ citation instructions on the primary paper or project page.Download route linked; files not verifiedAccess checked 2026-08-19The official ObjectFolder site provides dataset and benchmark download routes.Reviewed 2026-08-19Stanford University; Carnegie Mellon University
    2023
    Robot: Franka Emika Panda
    Sensor: GelSight robotic finger
    3D mesh; HD RGB video; Impact audio; Tactile reading
    Reported / previously documented scale: 100 real-world household objects

    Public-file audit: Unknown: no independently recounted public-file total in this record.

    Independent check: Existing source review only. No new dataset-file download, checksum validation, training run, or hardware test in this audit.

    Cross-sensory retrieval; Contact localization; Material classification; Reconstruction; Manipulation benchmarks
    Objects: One hundred real-world household objects; detailed inventory is provided through the official ObjectFolder resources.

    Use limitations (editorial): Object-level real observations support multisensory perception. A fixed object collection is not a long-horizon robot-manipulation dataset.

    Multisensory measurements aligned to object surface locations, including mesh, visual, acoustic, and tactile assets.
    Dataset-file license: Unknown: a dataset-file license has not been verified separately from paper or code terms.

    Code / source terms: Review the license bundled with the downloaded ObjectFolder release

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    ObjectFolder 2.0Use the authors’ citation instructions on the primary paper or project page.Download route linked; files not verifiedAccess checked 2026-08-19The official download page provides the Object Files, rendering instructions, metadata, and license information.Reviewed 2026-08-19Stanford University; Carnegie Mellon University
    2022
    Robot: Simulation / neural object rendering
    Sensor: Simulated GelSight-style tactile rendering
    Visual appearance; Impact audio; Tactile RGB image; 3D geometry
    Reported / previously documented scale: 1,000 multisensory neural objects

    Public-file audit: Unknown: no independently recounted public-file total in this record.

    Independent check: Existing source review only. No new dataset-file download, checksum validation, training run, or hardware test in this audit.

    Object scale estimation; Contact localization; Shape reconstruction; Multisensory representation learning
    Objects: One thousand neural objects derived from real-object 3D assets across common household categories.

    Use limitations (editorial): Simulated object signals support controlled representation experiments. Rendering fidelity and transfer to real sensors require separate tests.

    Object File implicit neural representations with query parameters for view, lighting, impact, contact location, gel rotation, and indentation depth.
    Dataset-file license: CC BY 4.0 for ObjectFolder 2.0 according to the existing review of the official download page.

    Code / source terms: CC BY 4.0 for ObjectFolder 2.0 according to the official download page

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    TacVerseUse the authors’ citation instructions on the primary paper or project page.Registration / approval requiredAccess checked 2026-09-13Official code and gated Hugging Face task archives verified on 2026-09-11. The 106,800-image count is paper-reported; files were not downloaded or recounted. No root code license was identified at the reviewed revision.Reviewed 2026-09-11Imperial College London; Queen Mary University of London; King’s College London
    2026
    Robot: Controlled tactile data-collection platform
    Sensor: GelSightNoMarker; GelSightMarker; MagicGripper; MagicTac; TacTip; ViTac; ViTacTip
    Tactile RGB image; Shape label; Grating label; Force label
    Reported / previously documented scale: 106,800 tactile images from seven vision-based tactile sensors

    Public-file audit: 7 paths in the provider file manifest, including metadata. Bytes, frames and episodes were not independently recounted.

    Independent check: Official hosting API and file manifest inspected at the pinned revision. No dataset payloads downloaded; no checksums, schema validation or training run.

    Shape classification; Grating classification; Force regression; Zero-shot cross-sensor transfer; Few-shot adaptation
    Objects: Controlled shape and grating stimuli plus force-regression contacts; a general object-category inventory is not stated.

    Use limitations (editorial): Cross-sensor comparisons require the documented sensor and task splits. Do not compare accuracy numbers across different split or few-shot settings.

    Multi-sensor tactile images and task labels organized for within-sensor, zero-shot cross-sensor, and few-shot protocols.
    Dataset-file license: cc-by-4.0 in the provider dataset-card metadata at the pinned revision; individual asset terms were not inspected.

    Code / source terms: CC BY 4.0 in the official Hugging Face metadata; access is gated and asset-specific terms were not independently checked

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    VTDexManipUse the authors’ citation instructions on the primary paper or project page.Download route linked; files not verifiedAccess checked 2026-08-19The author project and repository link benchmark code, paper, and dataset access.Reviewed 2026-08-19Zhejiang University
    2025
    Robot: Human dexterous demonstration platform; Simulated dexterous manipulation benchmark
    Sensor: Sparse binary tactile sensing
    RGB image; Sparse binary touch; Proprioception; Manipulation trajectory
    Reported / previously documented scale: 10 daily manipulation tasks across 182 objects; trajectory and frame totals are not stated as text on the reviewed project page

    Public-file audit: Unknown: no independently recounted public-file total in this record.

    Independent check: Existing source review only. No new dataset-file download, checksum validation, training run, or hardware test in this audit.

    Visual-tactile pretraining; Bottle-cap turning; Faucet screwing; Lever sliding; Table reorientation; In-hand reorientation; Bimanual hand-over
    Objects: One hundred eighty-two objects used across ten daily manipulation tasks.

    Use limitations (editorial): Separate visual-tactile pretraining from the simulator policy benchmark. Neither alone demonstrates transfer to a physical hand or a new sensor.

    Visual-tactile pretraining data and an Isaac Gym policy benchmark; inspect the official repository for the current file layout.
    Dataset-file license: Unknown: a dataset-file license has not been verified separately from paper or code terms.

    Code / source terms: MIT License for the benchmark code; verify dataset-file terms separately

    Train / test split: Unknown: no train/test split manifest verified in this audit.

    Short answer

    What you need to know

    1. 1

      A useful tactile dataset is defined by more than frame count. Check the physical collection event, sensor and robot state alignment, object and task diversity, split unit, access terms, and downstream evaluation.

    2. 2

      Contact sequences matter because adjacent tactile frames from the same press or trajectory are strongly related. Random frame splits can leak near-duplicate contact evidence into both training and test sets.

    3. 3

      Use the searchable directory below to find resources such as TactiDex, VTDexManip, EgoTouch, and Touch and Go. Filter by sensor, robot, task, or modality, then open the primary source to check downloads and reuse terms. A listed resource may still have incomplete files or an unstated license.

    Topic 01

    TouchScale: study scale, access and action labels

    The November 2 TouchScale v3 paper describes a 500-hour human visual-tactile corpus. On November 8, the public host card still described a 100-hour release, with gated access and CC BY-NC 4.0 terms. Its earlier full-release plan does not establish that the full corpus is available. The new directory record preserves the reviewed card revision and separates documentation from uninspected payloads.

    The paper states that TouchScale does not include action labels. Use human contact recordings for an explicitly defined representation or prediction objective; do not treat them as ready-made robot action targets. Plan any robot demonstration, embodiment mapping and controller adaptation as separate inputs to the experiment.

    Topic 02

    What a directory export does and does not establish

    Keep three quantities separate when planning data use: what the paper reports collecting, what the host lists, and what your loader successfully decodes. They can differ without implying that data is lost. A file listing can include metadata, duplicates, auxiliary media or only one released subset; a successful metadata read does not count every episode inside a payload.

    Related collection hubs are excluded from the dataset-record count displayed by the explorer. The fixed research index is a separate paper selection, not a second dataset inventory. CSV and JSON exports describe the directory evidence; they neither redistribute the underlying data nor certify a working training loader.

    Evidence stageRecord it explicitlyNext unresolved question
    Paper collectionUnit, collection protocol and exact paper versionWhich part is actually released?
    Public route or manifestHost, revision, observed paths and access conditionsCan the intended payload be read under its terms?
    Decoded episodeLoader version, complete boundaries, field meanings and missing valuesDoes it match the target model and evaluation split?
    Training/evaluation runData selection, transformations, split and actual outcomeWhat transfers beyond that experiment?

    A concrete access-state change: PRISM

    On November 4, 2026, the official PRISM project linked a ModelScope dataset route. The target page did not expose a readable file manifest or reuse terms in this check. This supersedes the earlier disabled-button observation, but it does not establish payload completeness, a dataset license or successful decoding. The record therefore distinguishes a linked route from verified files.

    For tactile adaptation, also preserve the paper's modality boundary: tactile sensing covers an unspecified subset. The project's broad multimodal description should not be used to fill missing episode-level touch fields. No dataset payload was downloaded for this update.

    Topic 03

    A simulated tactile subset: inspect the action contract

    The DexSuite scripted Lift record now links a pinned public subset with inspected JSON metadata. It separates simulation from physical sensing and the subset from the paper-wide corpus. The data card places observations before actions and omits the final post-action state, so a next-observation loader must handle the terminal boundary explicitly.

    RoboSkin inspected metadata and the file listing, not trajectory payloads. Before training, check one complete episode, tactile values, time alignment, action semantics and a held-out split. A declared feature shape does not establish successful decoding or policy performance.

    Topic 04

    Choose tactile datasets by the task you want to learn

    The best dataset is the one whose observations, labels and action contract match your experiment. Use this task shortlist to narrow the explorer, then check the dated access and license notes on each record. These are research-fit suggestions, not a benchmark ranking or a fresh verification of every download.

    Learning taskStart with these recordsCheck before training
    Materials, touch descriptions or cross-modal retrievalTouch and Go and TVLTouch and Go supports human-collected visuo-tactile research; TVL aligns touch, vision and language from its constituent collections. Do not infer robot action trajectories from paired observations. Keep contact sequences together and inspect label provenance.
    Simulated grasping or insertionUniVTAC Benchmark DatasetSelect the task and matching simulator version. This is simulation data; encoder pretraining samples and physical demonstrations are separate resources. Inspect actions and complete episodes before reuse.
    Deformation-aware manipulation in simulationSoftVTBenchKeep simulated visual/tactile observations and deformable-object tasks separate from claims about physical sensor transfer. Check the released files and dataset terms.
    Humanoid visual-tactile-action learningRoboTacDexInspect robot embodiment, tactile placement, action semantics and synchronization. Follow its dated access record; a paper or project link does not establish a verified training download.
    Slip detection and recoveryRequire event-labeled contact sequences; use the SlipSense research brief to define the evidence neededDo not substitute generic grasp frames for independent slip-onset labels. Check false alarms, detection delay, load variation and held-out objects; recovery also requires actions and outcomes.

    Topic 05

    Does a paper describe data you can actually obtain?

    A result table can justify reading a method without establishing a usable download. SlipSense describes a labeled slip collection; Touch2Trace describes pretraining and cable demonstrations; Visible Touch describes real-robot demonstrations and magnetic hardware. The original papers and available official project links checked on September 19, 2026 did not establish verified downloads and reuse licenses for these collections. They remain method references rather than new downloadable dataset records.

    When evaluating a release, distinguish the observation data from its task labels, action contract and pretrained weights. For slip, request independent onset labels and object-level splits. For cable policies, request ordered observations, joint commands and excluded-start rules. For overlays, request raw readings, sensor geometry, camera transforms and normalization statistics. Unknown fields remain unknown until the released assets support them.

    Topic 06

    How to read this directory

    The entries below are research resources with different goals; they are not interchangeable rows in one leaderboard. Some emphasize tactile-language and material understanding, others whole-hand contact, data collection, humanoid action alignment, or multisensory representation learning.

    Before use, open the primary source and project page. Verify the actual downloadable files, license, sensor hardware, collection protocol, annotations, train-test splits, and version. A paper saying that a resource is open does not replace checking the current repository terms.

    Keep geometry-derived pseudo-contact separate from measured touch. HuTact pretrains on human hand-object interaction geometry: reconstructed hand and object meshes provide a proximity-like cue, not measured physical contact properties. Optional midtraining uses unlabeled robot play, and downstream fine-tuning still uses real robotic tactile data. The authors evaluate only a Revo2-based bimanual setup; transfer to other hands and sensing modalities is untested. This is a complementary pretraining source, not evidence that tactile sensors or robot data can be removed. Budget geometry preparation, robot adaptation and labeled task data separately. Code, weights, dataset release and reproduction have not been independently verified by RoboSkin, so this method example is not a downloadable dataset entry.

    Source reviewed October 22, 2026: HuTact v1 methodology and limitations. The paper reports the experiments; RoboSkin has not reproduced them.

    If a modality label is unfamiliar, read what tactile sensing measures before interpreting an image, taxel value or inferred force as interchangeable data.

    Topic 07

    From a dataset shortlist to a loading plan

    Start with the task and collection setup. Human demonstrations can support perception or representation learning without providing an executable robot action contract. Simulated touch can support controlled experiments without establishing calibration or transfer to a physical sensor.

    Select up to three records in the explorer, then export the comparison. The CSV preserves the recorded availability, license, source-review date, primary paper, and official links. It is a snapshot of directory notes, not a redistribution of the underlying datasets or a grant of reuse rights.

    Use the downloadable selection checklist to document sensor compatibility, action semantics, synchronization, splits, current file access, reuse terms, and a minimum loading run. Write down what you actually downloaded and decoded. Leave unresolved fields explicit instead of treating a project announcement as a ready-to-train release.

    Check optical sensor differences in the GelSight versus DIGIT comparison so preprocessing and calibration choices match the collection hardware.

    • Define the required observations, labels, actions, and target robot before choosing a resource.
    • Inspect one complete sequence: units, timestamps, coordinate frames, missing values, and action meaning.
    • Check the official host and dataset-file terms separately from the paper and code licenses.
    • Record the version and review date; use the directory citation for these notes and cite the original authors for their research.

    Topic 08

    Tactile dataset and resource comparison

    This comparison records the main research unit and the limit a user should preserve. Counts are included only where the primary source states them clearly.

    ResourceSignals and scaleBest-fit questionEvaluation unitEvidence boundary
    PRISM5,000+ robot trajectories, 5,000 paired human demonstrations, 45+ hours, and approximately 27M images across visual and visuotactile streams; tactile sensing covers an unspecified subset.Industrial contact-rich manipulation across multiple robots, grippers, and teleoperation interfaces.Split by complete episode, task, operator, hardware configuration, object, and modality availability.November 4, 2026: the official project links a ModelScope dataset route; file inventory, payload and dataset-file license remain unverified.
    EgoTouch208 human bimanual manipulation tasks across 1,891 episodes with head and dual-wrist RGB, 3D hand pose, and wearable pressure maps; the repository reports about 2M frames.Egocentric vision-to-touch prediction and human contact representation pretraining.Split by complete episode, task, object, scene, and subject or collection condition where exposed.This is human-wearable interaction data, not robot action trajectories. Hosted files exist, but the official repository warns that upload may be incomplete and a separate dataset-file license was not verified.
    HT-Bench10M egocentric RGB frames and 7.8M full-hand tactile frames collected across 226 tasks.Full-hand tactile representation learning, cross-modal alignment, and unseen-task evaluation.Split by task and trajectory; test held-out tasks, objects, sensor units, or embodiments for the claimed transfer.A 2026 preprint. Scale does not make temporally adjacent frames independent, and reported results are specific to the benchmark setup.
    RCT29,279 tactile frames from 122 industrial reference materials in 7 categories, collected with 3 DIGIT sensors; paired touch, image, language, and force context.Material understanding and tactile-language retrieval.Keep full press or contact sequences together; test held-out materials where possible.A 2026 preprint. Reported performance is specific to its sensors, materials, models, and splits.
    TactiDexWhole-hand tactile observations aligned with multi-granularity kinematic and object states for single-hand and bimanual tasks.Contact-rich dexterity and transfer across manipulation settings.Use the standardized task and transfer protocol described by the source.A 2026 preprint and project resource; inspect the released tasks, files, and license before reuse.
    FreeTacManPaired visuo-tactile observations and interaction trajectories collected with a portable, human-operated workflow.Scaling contact-rich demonstrations without occupying a robot arm for every collection session.Split by task, object, trajectory, and operator conditions that match the transfer claim.A 2025 preprint; human-device data still needs validation on the target robot embodiment.
    Humanoid visual-tactile-action datasetSynchronized vision, tactile observations, and action context for humanoid contact-rich manipulation.Learning policies that need touch aligned with the action that produced it.Keep synchronized trajectory segments and embodiment conditions intact.A preprint; transfer depends on robot geometry, sensor placement, action space, and task distribution.
    Sparsh-X research resourceDigit 360 tactile images, audio, motion, and pressure used for self-supervised multisensory touch representations.Learning reusable tactile features across physical-property and manipulation tasks.Evaluate downstream tasks and held-out conditions, not only pretraining loss.A 2025 preprint tied to a multisensory sensor stack; cross-sensor transfer still requires evidence.
    Touch and GoApproximately 13.9K detected touches across indoor and outdoor scenes with egocentric vision and GelSight recordings.In-the-wild visuo-tactile representation learning and future-touch prediction.Split by object, scene, video, or collection sequence for the intended claim.The object-instance count is estimated in the paper and the collection is human-operated rather than robot-action data.
    TVL43,741 in-contact image-touch pairs with English tactile descriptions.Touch-vision-language alignment and tactile description.Keep HCT and SSVTP origins, contact events, and human versus generated labels visible.The paper reports that 90% of labels are VLM-generated and documents occasional label errors.
    ObjectFolder Real / 2.0100 real household objects plus 1,000 neural objects with visual, acoustic, tactile, and geometric data.Multisensory object recognition, reconstruction, and manipulation.Separate real measurements from simulated neural-object rendering.Synthetic and real results are not interchangeable; use the exact release and task protocol.
    TacVerse106,800 paper-reported tactile images from seven vision-based tactile sensors.Within-sensor learning, zero-shot transfer, and force-only few-shot adaptation.The paper specifies whole-trial chronological 60/20/20 splits; sample_id-to-trial mapping remains unverified.Official code and gated Hugging Face archives verified on 2026-09-11; raw data not downloaded or independently recounted.
    VTDexManipVisual-tactile data from 10 daily manipulation tasks across 182 objects plus a six-task dexterous benchmark.Visual-tactile pretraining and policy evaluation.Use task, object, modality, and policy splits that match the transfer claim.The official code is MIT-licensed; dataset-file rights need separate verification.

    Topic 09

    Why contact-sequence splits matter

    RCT reports 29,279 frames but also preserves full contact sequences. Frames from one press are correlated, so a random frame-level split can place nearly the same physical event in training and test data. The paper reports that removing contact-sequence overlap reduces tactile-to-text Recall@1 by 17.7 percentage points.

    The broader rule is to split at the level of the claim. For unseen-material performance, hold out materials. For unseen-object manipulation, hold out objects. For transfer across robots or sensors, hold out the target hardware. A large test set is not independent if the same contact event, object instance, or trajectory appears on both sides.

    • Record the physical unit: press, grasp, trajectory, object, task, operator, robot, and sensor
    • Group correlated frames before creating train, validation, and test sets
    • Publish split manifests or deterministic split code
    • Report results for the hardest held-out condition relevant to the claim

    Topic 10

    Dataset selection checklist

    Start from the deployment mismatch you need to measure. A material dataset may be rich enough for tactile-language learning but unsuitable for action-conditioned robot control. A whole-hand trajectory dataset may support dexterity research but still mismatch a fingertip sensor, gripper geometry, or action space.

    • Access: files, repository status, license, citation, version, and checksum
    • Hardware: sensor model, serial variation, placement, sampling rate, calibration, and units
    • Alignment: timestamps for touch, vision, force, pose, joint state, and actions
    • Coverage: objects, materials, tasks, contact types, operators, robots, and disturbances
    • Splits: sequence, object, material, task, sensor, or embodiment independence
    • Outcome: retrieval, classification, prediction, imitation, or real-robot task success

    Topic 11

    What dataset size does not prove

    Frame count does not establish diversity, independent evaluation, target-robot transfer, or production readiness. Ten thousand adjacent frames from a small set of presses can contain less independent evidence than a smaller collection spread across objects, sensors, and trajectories.

    Use this directory to locate sources, then document the exact dataset version and split used in your experiment. Do not compare headline metrics across resources unless the sensor inputs, tasks, models, and evaluation protocols are genuinely aligned.

    For a concrete example, read the N0-Foundation explanation of NeoData and OpenNeoData and distinguish the reported corpus from the open subset and individual model training claims.

    Topic 12

    Inspect a dataset before adapting it for training

    After checking access and license at the original source, inspect episode boundaries, feature meanings, timestamp conventions and missing observations. Container names do not guarantee compatible actions, camera timing or tactile inputs.

    The LeRobot tutorial explains v3 storage and supplies a small numeric checker with valid and broken synthetic fixtures. It complements the tactile CSV exercise; neither is an original research dataset or a complete training-readiness certificate.

    For time-aligned acquisition and replay, use the ROS 2 tactile tutorial and preserve the source dataset’s units and validity semantics when designing an adapter.

    Topic 13

    ROMI-2K: active perception data and release boundaries

    ROMI-2K is the object-interaction dataset introduced with ROMA (arXiv:2610.06955). Paper v2 reports 1,657 handheld object–content combinations and 500 robotic tabletop scenes, spanning lift, press, collide, shake, rotate and squeeze interactions. These are different counting units; they must not be added and labeled as unique trajectories or unique objects.

    The handheld Pika Sense subset captures vision, audio and optical touch without a force sensor. The robotic tabletop setup adds a six-axis force/torque sensor. ROMA itself uses mean gravity force as text, so recorded force/torque trajectories and the model’s consumed input are different claims.

    Access checked October 12, 2026: the official Hugging Face card still marks ROMI-2K and ROMA Bench as coming soon. Public checkpoints and one recorded demo scene do not establish a complete dataset release. A checked release item in the GitHub README conflicts with its coming-soon text. No dataset-specific reuse license was identified; a license in a model subdirectory must not be applied to the data.

    Use the ROMI-2K structured record for collection details and source links; revisit the official card before attempting reuse.

    Place these interactions in the active sensing explanation to distinguish information gathering from downstream manipulation.

    Recomputable directory analysis · v2026.11.17

    Public availability and reproduction conditions

    Compiled 2026-11-17 from the 27 records in this tactile directory. This is a selected research sample, not an industry census. Individual source-review and access-check dates appear in each record. A public manifest is evidence of hosted paths, not proof that all payloads download or a result can be reproduced.

    Counts include unknowns in the denominator (27); no missing value is treated as a negative result.
    Evidence categoryRecordsInterpretation
    Registration / approval required3 / 27As recorded on each access-check date; access can change.
    Public file manifest10 / 27As recorded on each access-check date; access can change.
    Download route linked; files not verified8 / 27As recorded on each access-check date; access can change.
    Announced; release not verified3 / 27As recorded on each access-check date; access can change.
    Access unknown3 / 27As recorded on each access-check date; access can change.
    Dataset-specific license documented13 / 2714 unknown; a paper or code license does not fill this field.
    Split files listed1 / 27Listing only; split membership and leakage not checked.
    Split tools / protocol described3 / 27Distinct from inspecting released splits; 23 records remain unknown.

    The practical constraint is the evidence gap between a paper-scale claim and a usable training package. This audit inspected 12 provider manifests without downloading dataset payloads. Start by pinning a release, reading dataset-file terms, checking the schema and units, and validating the train/test boundary before running a baseline.

    Specific evidence still to verify

    Directory change log

    — Added the upstream TVL folder-swap correction notice and a pinned TVL-revise download route. Inspected the public file manifest and card; dataset-file terms, archive contents and image-pair alignment remain unverified. Previous audit snapshot.

    — Added intended-use and limitation guidance to comparison and exports, with format, download route and verification boundary in the three-record comparison. This edition reuses the dated source/access evidence; no new file download or research reproduction is claimed. Previous audit snapshot.

    — Added ROMI-2K for active perception. The official card still says Coming soon; complete dataset files and dataset-specific reuse terms remain unverified. Preserved the source discrepancy and separated handheld optical touch from tabletop force/torque. Older records retain their own check dates.

    — Added Bench2Dex simulation demonstrations with a pinned public manifest, separate license fields and original authors. STAR remains a paper-linked candidate outside the catalog count and Dataset schema. Published a new audit snapshot; older records retain their own verification dates.

    — Added separate access evidence, provider revisions, public-manifest scale, dataset-file license status and split evidence to existing records, with computable counts and JSON export. Original audit snapshot.

    Paper-associated data

    Data candidates: access still unverified

    These papers describe research data, but we have not verified a downloadable release and its reuse terms. Candidates are separate from the dataset catalog, its counts and filters, and its Dataset structured data. The JSON export preserves them in a separate candidates field.

    STAR paper-associated dexterous manipulation data

    Paper-linked candidate · Access checked 2026-09-18

    Paper-linked data candidate only. Official project reopened on 2026-09-18; no dataset download, code release or dataset reuse license verified. Excluded from the dataset catalog count, available-file filters and Dataset structured data.

    Original authors
    Xiangcheng Liu; Tianhao Wu; Le Zheng; Yidong Wang; Bowen Jiang; Mingjie Pan; Xinlin Ren; Yi Liu; Jianlan Luo
    Source version
    First submitted 2026-09-11; arXiv v1 (2609.12549v1)
    Institutions / data origin
    Shanghai Innovation Institute; Agibot / real
    Reported scale
    Author-reported 200 hours, 10,576 trajectories and 65 tasks. No hosted-file count verified.
    Robot and sensors
    Bimanual wheeled mobile robot; two hands with 10 actuated DoF and 16 total DoF each; commercial model name unknown; 268-dimensional piezoresistive tactile sensing per hand; Two wrist RGB cameras and one head RGB camera
    Tasks and objects
    Dexterous multi-finger manipulation; Pick-and-place and insertion; Long-horizon manipulation. Task-specific household and manipulation objects; exhaustive inventory and unique object count unknown.
    Modalities, actions and synchronization
    Three-view RGB; Bilateral tactile signals; Robot state; Command actions; Task-level language instructions. Paper describes timestamp alignment to 30 Hz; this is not the raw tactile sampling rate. Command actions use end-effector poses and hand joints (34 active dimensions, padded to 64 for the policy). Public package, file schema and train/validation/test manifests unknown.
    Splits and generalization
    Evaluation objects are held out from 200-hour pretraining but used in 100 task-specific post-training trajectories per task; 20 evaluation trials per task use held-out initial configurations. Not unseen-object zero-shot generalization.
    Code / data / model / asset terms
    Unknown: no dataset-file, code, weight or asset license verified. The arXiv article license does not license the described dataset.
    Verified and unresolved
    Paper v1 and official project inspected. Download entry, released file inventory, payload integrity, split files and dataset license remain unknown. No independent reproduction by RoboSkin.

    Common questions

    FAQ for this topic

    01

    What is the best tactile dataset for robot learning?

    There is no universal best dataset. Match the resource to the learning goal, sensor signals, robot embodiment, task, and evaluation split required by the deployment claim.

    02

    Why are random frame splits risky for tactile data?

    Adjacent frames from the same press, grasp, or trajectory can be near duplicates. If they appear in both training and test sets, the metric can overstate generalization.

    03

    Does open source mean unrestricted commercial use?

    No. Open access to a paper, project page, code, or files does not define commercial rights. Check the license for the exact dataset version and every bundled asset.

    04

    Should tactile datasets include robot actions?

    They should when the goal is policy learning, action-conditioned prediction, or replay of manipulation. Material recognition or representation learning may use different labels and collection units.

    Related