Structured benchmark directory
Tactile robotics benchmarks compared
Compare tactile robotics benchmarks by task, sensor, robot, modality, metric, split protocol, access, and evidence boundary.
Published | Updated by RoboSkin.ai Editorial Team
- 5
- sections
- 4
- questions
- 8
- next routes
Reported results / Protocol first
Experimental evidence, with the conditions attached
Inspect author-reported results by study, test setup, metric, and sample. These are extracted research records; RoboSkin.ai has not independently reproduced the experiments. Compare scores only when the evaluation conditions match.
Showing 6 of 6 records · Reviewed September 12, 2026
Swipe the table horizontally to read the complete evidence.
| Study and test setup | Result and reference | Sample and interpretation | Limits and access | Sources |
|---|---|---|---|---|
| TacPrint / Robot graspingarXiv 2607.29231v1; July 31, 2026; preprint 24 capacitive taxels predict a 35 × 26 depth map. Orange-model grasping; success requires a 50 mm lift and a 5 s hold. | Grasp success Dense depth: 35/40 (87.5%) Raw taxels: 27/40 (67.5%); contact-only: 15/40 (37.5%) | 40 tested positions per feedback condition; the 20 edge-contact trials are a subset. Dense depth exceeds raw-taxel success by 20 percentage points in this experiment. | One tested object geometry; estimates averaged over 5 s and correction count uncapped. The result does not establish faster control or general object transfer. No official downloadable code/data release identified in the reviewed paper as of September 11, 2026; not independently reproduced by RoboSkin.ai. | Primary source ↗ Section IV-C, Experiment 3 and Figure 11 Read evidence review |
| TacPrint / Robot graspingarXiv 2607.29231v1; July 31, 2026; preprint Same orange-model grasp, feedback conditions, 50 mm lift and 5 s hold as the all-position experiment; edge-contact subset. | Edge-contact grasp success Dense depth: 17/20 (85%) Raw taxels: 9/20 (45%); contact-only: 4/20 (20%) | 20 edge-contact trials per condition, already included in the 40-position record; do not add these as independent trials. Dense depth exceeds raw-taxel success by 40 percentage points on this subset. | Subset result under the same object and correction protocol. It does not establish a separate 60-trial evaluation or a universal slip-recovery rate. No official downloadable code/data release identified in the reviewed paper as of September 11, 2026; not independently reproduced by RoboSkin.ai. | Primary source ↗ Section IV-C, Experiment 3 and Figure 11 Read evidence review |
| TacPrint / Contact reconstructionarXiv 2607.29231v1; July 31, 2026; preprint Measured capacitive inputs; predicted depth evaluated at a guide-calibrated physical contact center. | Center-depth mean absolute error 0.085 ± 0.057 mm (mean ± SD) Reference: prescribed physical indentation depth; no matched alternative sensor baseline in this record. | All 40 controlled trials. The separate primary position-error statistic uses 37 trials, excluding three boundary-truncated reference regions. Evaluates local depth consistency at the reference center under controlled indentation. | Not a force error, full-surface physical depth error, or simulation-reference RMSE. SD describes spread, not a confidence interval. No official downloadable code/data release identified in the reviewed paper as of September 11, 2026; not independently reproduced by RoboSkin.ai. | Primary source ↗ Section IV-C, Experiment 2 Read evidence review |
| TacVerse / Cross-sensor classificationarXiv 2606.25877v1; June 24, 2026; preprint Shape classification trained on GelSightMarker; evaluated directly on GelSightNoMarker. | Classification accuracy GelSightNoMarker cross-sensor: 98.1% GelSightMarker within-sensor: 95.5% | Table 2 does not state a per-cell test denominator. The full dataset has 106,800 records across tasks and sensors; that is not the denominator of this score. A favorable transfer pair in the shape task; source and target test sets differ. | Not a matched-test-set improvement, a generic hardware ranking, or evidence that other sensor pairs transfer equally well. Paper/code split correspondence remains unverified without the raw data. Public code repository; dataset listing is gated. No root code license identified in the September 11 review. RoboSkin.ai did not download the gated data or reproduce the score. | Primary source ↗ Table 2, GelSightMarker source rows Read evidence review |
| TacVerse / Cross-sensor classificationarXiv 2606.25877v1; June 24, 2026; preprint Grating classification with GelSightMarker fixed as the source sensor; zero-shot cross-sensor evaluation. | Classification accuracy GelSightNoMarker: 24.8%; MagicGripper: 5.4%; ViTacTip: 4.1% GelSightMarker within-sensor: 90.3% | Per-target test denominator is not stated in Table 3. Each target score uses its own sensor observations. The same source sensor transfers much less successfully on grating than on the selected shape pair. | Different task and test observations from the shape record. These accuracies cannot be averaged into one sensor-quality score or compared with a robot success rate. Public code repository; gated dataset listing. Access and license boundaries were reviewed September 11, 2026; no independent reproduction by RoboSkin.ai. | Primary source ↗ Table 3, fixed GelSightMarker source Read evidence review |
| TWINS / Robot trajectory trackingarXiv 2608.01733v1; August 3, 2026; preprint Isomorphic Robot executing learned policies; tracking commanded trajectories over 14 arm joints. | Joint tracking error MAE: 0.94°; 95th percentile: 3.43° Reference: commanded joint trajectories. Reported tracking delay is approximately 0.4–0.6 s. | 14 arm joints; the paragraph does not report the number of evaluated timestamps. The 40 collected demonstrations are not a task-success denominator. Quantifies tracking accuracy in the reported robot execution. | Policy outcomes are evaluated qualitatively. This record supplies neither a task-success rate nor a matched tactile-disabled uplift estimate. Hardware repository available under CERN-OHL-W-2.0; learning code marked Coming soon in the September 11 source review. No independent build or policy reproduction by RoboSkin.ai. | Primary source ↗ Section 4.4, qualitative evaluation of learned policies Read evidence review |
Short answer
What you need to know
- 1
A tactile robotics benchmark is a defined evaluation contract: task, sensor input, robot or collection platform, data split, metric, baseline, and version. A dataset becomes a benchmark only when it is paired with a reproducible protocol.
- 2
Benchmark scores are not portable across different tactile sensors, robots, objects, control rates, or success definitions. Compare systems only inside the same protocol or reproduce both under one shared setup.
- 3
For manipulation, always include an outcome-level baseline with touch disabled. Representation accuracy alone does not prove that tactile sensing improves robot behavior.
Topic 01
What makes a tactile benchmark credible?
A credible benchmark states what is held constant and what is allowed to vary. At minimum, it identifies the physical data unit, sensor and robot configuration, training information, split unit, metric, baseline, evaluation repetitions, and failure definition.
Tactile data are unusually vulnerable to leakage because adjacent frames from the same press, grasp, or trajectory can be nearly identical. The correct split unit follows the claim: hold out presses for new-contact evaluation, materials for material generalization, sensors for cross-sensor transfer, and objects or tasks for manipulation transfer.
- Input contract: raw tactile images, taxel arrays, force vectors, touch plus vision, or learned embeddings
- Independence unit: frame, contact sequence, object, material, task, sensor, robot, or laboratory
- Outcome: perception accuracy, prediction error, task success, recovery, latency, or safety event
- Reproducibility: released data, split manifest, evaluation code, hardware description, and version
Topic 02
Benchmark families are not one leaderboard
The directory includes representation suites, multimodal understanding tests, cross-sensor transfer protocols, full-hand benchmarks, and closed-loop manipulation challenges. They answer different questions and should not be collapsed into a single ranking.
| Benchmark family | Primary question | Typical unit | Useful output | Common mistake |
|---|---|---|---|---|
| Representation | Does an encoder preserve contact information that transfers to downstream tasks? | Tactile frame, clip, or contact sequence | Frozen-encoder task metrics and data-efficiency curves | Treating average improvement across unlike tasks as one universal score |
| Cross-sensor | Does a model survive a change in sensor design or instance? | Held-out sensor or sensor instance | Zero-shot gap and few-shot recovery | Mixing frames from the same physical contact across train and test |
| Multimodal | Can touch align with vision, language, audio, or robot state? | Aligned example or trajectory | Retrieval, classification, generation, or prediction metrics | Ignoring timestamp error or generated-label quality |
| Manipulation | Does tactile input improve closed-loop task behavior? | Episode or repeated real-robot trial | Success, recovery, time, damage, and intervention rate | Reporting perception accuracy without a touch-disabled task baseline |
Topic 03
Minimum evaluation contract for tactile manipulation
A useful task-level comparison fixes the gripper or hand, object set, initial-state distribution, controller budget, tactile preprocessing, vision inputs, and success rule. It then repeats trials across relevant disturbances and reports both aggregate results and failure categories.
- No-touch or disabled-touch baseline on the same robot and task
- Repeated trials across objects, poses, contact conditions, and disturbances
- Latency from physical contact through sensing, inference, command, and actuation
- Calibration drift, sensor replacement, wear, and reinitialization procedure
- Failure taxonomy: miss, slip, jam, drop, excess force, timeout, or unsafe contact
Topic 04
How to use the directory
Filter by benchmark type, task, sensor, and year, then open the primary paper and official project or code page. Record the exact version you use. “Open source” on a project page does not automatically establish the license of every dataset, model weight, or bundled asset.
The evidence-boundary column is deliberate. It prevents a benchmark’s strongest reported result from being generalized beyond its sensor, robot, task, split, or publication status.
Topic 05
Bench2Dex: bimanual simulation and scene changes
The September 14, 2026 Bench2Dex preprint reports 26 task and embodiment settings spanning 12 dexterous hands, with about 1,300 human-teleoperated simulation demonstrations. Its tactile interface encodes local contact geometry into image-like maps; it does not reproduce a particular physical sensor or provide calibrated force measurements.
The main policy comparison separates matched scenes, visual-context changes, geometry changes, and independently sampled combined changes. It measures stable success and dependency-valid stage progress. The 50 evaluation rollouts per cell are not independent retraining runs, and the tasks are not fully crossed with all hands. The public code and data listings were checked on September 18; dataset, asset, and checkpoint licensing still require separate confirmation.
Structured benchmark explorer
Compare tactile robotics benchmarks
These suites evaluate different sensors, modalities, robots, and outcomes. A score is comparable only when the task, split, hardware, metric, and benchmark version match.
Source review: 2026-09-19 / 13 records
Showing 13 of 13 benchmarks
| Benchmark | Type / year | Tasks | Modalities / hardware | Metrics / protocol | Access / evidence boundary | Primary links |
|---|---|---|---|---|---|---|
| Bench2DexShanghai Jiao Tong University; Fudan University; The University of Hong Kong; Inspire Robots; Zhongguancun Academy; COWARobot Co. Ltd; Nanyang Technological UniversityReviewed 2026-09-18 | Simulation benchmark for bimanual dexterous manipulation 2026 | Bimanual tool use; Articulated-object interaction; Multi-stage manipulation; Controlled scene perturbations | Signals: Multi-view RGB; Joint state; Commanded actions; Object state; Surface-aligned tactile maps; Depth, 2D/3D boxes and occupancy when recorded Sensors: Surface-aligned simulated contact geometry; not measured output from a physical tactile sensor Robot: 12 simulated bimanual dexterous arm–hand embodiments; 26 task–embodiment settings | Stable terminal success rate (SR); Latched stage completion rate (LSCR); Time to stable success conditional on success; Configured drop and high-speed diagnostics Protocol: arXiv v1, first submitted 2026-09-14: 26 task–embodiment settings, 50 rollouts per setting/channel/policy, four policies and four channels (20,800 evaluation episodes). Success requires a terminal predicate to hold for the configured dwell, 0.5 s by default. None/Equi./Inv. use matched anchors; Full is sampled independently. | Official project and MIT code public. teleopdata revision b195787c65046083e6a43776b67bdf1389dfa3eb exposes HDF5 paths; dataset, model and collection-wide asset terms unknown. No payload validation or simulator run by RoboSkin. Boundary: Author-reported technical report; peer-reviewed acceptance and independent reproduction unverified. Main results are not a tactile-on/off ablation. Contact geometry is simulated, not a real sensor output; 12 hand support does not prove zero-shot cross-hand transfer. High-speed diagnostics are not measured forces. | |
| UniVTAC BenchmarkScaleLab, Shanghai Jiao Tong University; D-Robotics; ViTai Robotics; The University of Hong Kong; Nanjing University; Shenzhen University; Wuhan University; Fudan University; Tsinghua UniversityReviewed 2026-09-19 | Simulation-based visuo-tactile manipulation benchmark 2026 | Lift Bottle; Pull-out Key; Lift Can; Put Bottle in Shelf; Insert Hole; Insert HDMI; Insert Tube; Grasp Classify | Signals: Head RGB; Wrist RGB; Robot state and action; Bilateral simulated tactile RGB; Marker motion; Gelpad depth; Tactile-sensor pose Sensors: Bilateral simulated GelSight Mini sensors in the released collection and evaluation pipeline Robot: Simulated Franka Panda with a parallel-jaw gripper | Binary task success rate over evaluation rollouts; Per-task success rate; Eight-task macro-average success rate Protocol: The paper trains every policy on 50 automatically collected full trajectories per task, or 400 trajectories across eight tasks, and evaluates each method on 100 test rollouts per task. Table I reports eight-task averages of 30.9% for vision-only ACT, 40.5% for VITaL, and 48.0% for ACT with the UniVTAC Encoder; 30.9% to 48.0% is an increase of 17.1 percentage points. | Public code and Hugging Face listings now separate 800 task HDF5 files for Isaac Sim 4.5 and 800 for 5.1; the official repository associates the released policy checkpoints with 4.5 only. The current collection/evaluation TODO still limits working sensor support to simulated GelSight Mini. See the research review for pinned revisions and an inventory; no benchmark rerun was performed. Boundary: An author-defined simulation benchmark, not an independent leaderboard or physical-robot safety test. Neither versioned 800-file task release is the paper's 400 policy-training trajectories or evaluation rollouts. Logs inspected in the August 22 audit differ from Table I; the September 19 file inventory does not reproduce or re-evaluate those results. | |
| TouchWorld Real-Robot Evaluation ProtocolHarbin Institute of Technology, Shenzhen; PHANES AIReviewed 2026-08-22 | Author-defined real-robot manipulation evaluation protocol 2026 | Water Flower; Tabletop Clearing; Cup Insertion; Power Plug Insertion; Pot Wiping; Tissue Pulling | Signals: Multi-view RGB; Proprioceptive state; Left and right tactile pressure observations; Natural-language task instruction; Robot actions Sensors: JQ-Industries tactile glove mounted on Wuji dexterous hands Robot: One unnamed humanoid platform with Wuji dexterous hands | Per-task manipulation success rate in a clean setting; Per-task manipulation success rate under human perturbation; Six-task average success rate Protocol: The v2 preprint defines six contact-rich tasks, collects 200 teleoperated training trajectories per task, and conducts 100 real-robot evaluation rollouts per task. Each task is evaluated in clean and human-perturbation settings; the paper does not state that this is an independent shared leaderboard. | The preprint and official project page are public. No standalone benchmark package, evaluation scripts, fixed test-set release, or benchmark artifact license was verified on 2026-08-22. Boundary: This is an author-defined evaluation on one unnamed humanoid, one Wuji-hand and tactile-glove configuration, six tasks, and source-specific baselines. The reported 65.0% clean and 53.7% perturbed averages are author-reported TouchWorld results, not independent validation or a universal cross-model benchmark. | |
| SoftVTBenchTuojing 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 KongReviewed 2026-08-22 | Simulated deformation-aware visuo-tactile manipulation benchmark 2026 | Object-Soft manipulation; Spatial-Soft manipulation; Object-Rigid matched control; Spatial-Rigid matched control; In-distribution evaluation; Out-of-distribution evaluation | Signals: Simulated multi-view RGB; Simulated bilateral tactile RGB; Simulated marker motion; Proprioception; Language; Robot actions; Evaluator-only FEM deformation state Sensors: Simulated GelSight Mini profiles through TacEx, Taxim, and FOTS Robot: Simulated Franka arm with Panda parallel-jaw gripper | Task Success Rate (TSR); Deformation-aware Success Rate (DSR); Peak normalized deformation; TSR-DSR gap Protocol: Uses policy-independent, object-specific probing to fix deformation tolerances before training; DSR credits a rollout only when it completes the task and remains within the calibrated FEM deformation tolerance. | The official project and GitHub repository link public training and evaluation code plus a Hugging Face dataset mirror. Hugging Face dataset-card revision fd2793a documents 4,000 hosted demonstrations, but the older GitHub README still lists 1,628 demonstrations and 33 assets; users should pin a revision because the first-party release documents disagree. Boundary: All robot, tactile, object, and FEM signals are simulated; no SoftVTBench-specific physical-sensor or sim-to-real validation is reported. DSR is defined by this paper and is not an industry standard, safety certification, or universal damage metric. | |
| HT-BenchBeihang University; Rimbot; ShanghaiTech University; Tsinghua University; Chinese Academy of Sciences; BUPTReviewed 2026-08-22 | Full-hand representation benchmark 2026 | Tactile similarity retrieval; Masked tactile inpainting; Vision-to-tactile synthesis; Tactile frame prediction; Downstream board cleaning; Downstream pear picking; Downstream water pouring; Downstream sand shoveling | Signals: Egocentric RGB; Full-hand tactile pressure maps Sensors: Single reported full-hand tactile sensing pipeline Robot: Dexterous full-hand platform | Recall@5; Full-map and hole-region RMSE; Full-map and hole-region cIoU; Real-world task success rate Protocol: Four offline tracks use standard-test and task-level OOD splits. Version 2 also reports board cleaning, pear picking, water pouring, and sand shoveling with 15 trials per method for each task. | The v2 preprint says the authors will release all data, evaluation protocols, pretrained weights, and training/testing scripts; no dedicated downloadable package or official repository was verified on 2026-08-22. Boundary: A 2026 preprint tied to one full-hand tactile system and embodiment pipeline; it does not cover fingertip optical tactile sensors, force/torque sensors, skin-like taxel arrays, or non-hand embodiments. The four-task downstream table reports success rates without confidence intervals or statistical-significance testing and is not a universal cross-sensor or cross-robot standard. | |
| TacBenchFAIR at Meta; University of Washington; Carnegie Mellon UniversityReviewed 2026-08-19 | Tactile representation evaluation suite 2024 | Force estimation; Slip detection; Pose estimation; Grasp stability; Textile recognition; Dexterous manipulation | Signals: Vision-based tactile images Sensors: DIGIT; GelSight 2017; GelSight Mini Robot: Task-specific manipulation platforms | Task-specific regression; Task-specific classification; Manipulation outcome Protocol: Frozen tactile encoders are paired with lightweight downstream decoders across six touch-centric tasks. | Official project page links the paper, code, datasets, and model weights. Boundary: The suite targets vision-based tactile sensing; average gains across different task metrics should not be read as one universal score. | |
| ObjectFolder BenchmarkStanford University; Carnegie Mellon UniversityReviewed 2026-08-19 | Multisensory object benchmark suite 2023 | Cross-sensory retrieval; Contact localization; Material classification; 3D shape reconstruction; Visuo-tactile cross-generation; Grasp stability; Contact refinement; Surface traversal; Dynamic pushing | Signals: Vision; Audio; Touch; 3D geometry Sensors: GelSight robotic finger for ObjectFolder Real Robot: Franka Emika Panda for ObjectFolder Real collection; Simulated object interactions | Task-specific retrieval; Classification; Reconstruction; Manipulation metrics Protocol: Ten tasks are organized around recognition, reconstruction, and manipulation with neural and real objects. | The official project page links open datasets, papers, downloads, and the benchmark suite. Boundary: Results across simulated neural objects and real-object measurements are not interchangeable; each task keeps its own protocol. | |
| Touch-Vision-Language BenchmarkUC Berkeley; Meta AI Research; TU Dresden; CeTIReviewed 2026-08-19 | Multimodal tactile-language benchmark 2024 | Tactile-semantic description; Touch-vision alignment; Touch-language alignment | Signals: DIGIT tactile image; RGB image; English tactile description Sensors: DIGIT Robot: Handheld collection device; Robot-collected SSVTP subset | Tactile-semantic classification accuracy; Cross-modal alignment accuracy; LLM-rated description consistency Protocol: TVL combines HCT and SSVTP data and evaluates multimodal models on tactile-semantic understanding and alignment. | The official project page links the paper, code, dataset, and models. Boundary: Most HCT language labels are VLM-generated; the benchmark paper documents label errors and should be read with that boundary. | |
| TacVerseImperial College London; Queen Mary University of London; King’s College LondonReviewed 2026-09-11 | Cross-sensor vision-based tactile benchmark 2026 | Shape classification; Grating classification; Force regression | Signals: Vision-based tactile images; Force labels Sensors: GelSightNoMarker; GelSightMarker; MagicGripper; MagicTac; TacTip; ViTac; ViTacTip Robot: Controlled tactile data-collection platform | Classification accuracy; Force-regression error; Cross-sensor transfer change Protocol: Within-sensor and zero-shot transfer on 106,800 reported images: seven-by-seven shape comparison, fixed-source grating and force tests. Few-shot adaptation is force-only and also uses labelled target validation. The paper describes whole-trial chronological 60/20/20 splits. | Official benchmark code and project verified on 2026-09-11; the linked Hugging Face dataset is gated. Code was inspected but not run, and raw files were not downloaded. Boundary: A 2026 preprint, not a closed-loop manipulation benchmark. The code partitions sample_id values; their mapping to whole physical trials remains unverified without the gated data. | |
| RCT Generalization BenchmarkTU Dresden; ScaDS.AI Dresden/Leipzig; LASR LabReviewed 2026-08-19 | Contact-sequence and material generalization protocol 2026 | Touch-to-text retrieval; Touch-to-vision retrieval; Material and category probes; Cross-sensor generalization | Signals: DIGIT tactile images; Material RGB; Language; Normal force; Indentation depth Sensors: Three DIGIT sensors Robot: Robot arm with rotating sensor adapter | Recall@1; Category-probe accuracy; Generalization gap Protocol: Preserves full robot presses and supports held-out sequences, contact positions, sensors, categories, and materials. | Official project page links the dataset, split tools, code, and preprint. Boundary: The benchmark focuses on industrial reference materials and contact sequences rather than closed-loop manipulation success. | |
| ManiSkill-ViTac 2025Challenge organizing team listed in the primary paperReviewed 2026-08-19 | Simulation-to-real manipulation challenge 2025 | Tactile manipulation; Vision-tactile fusion manipulation; Tactile sensor structure design | Signals: Vision; Marker-based visuo-tactile sensing; Robot state Sensors: Simulated and real marker-based visuo-tactile sensors; GelSight Mini integration resources Robot: Challenge simulation and real-robot environments | Standardized task metrics defined per challenge track Protocol: Three independent tracks evaluate learned contact-rich manipulation and sensor design in simulated and real-world settings. | The official repository provides environments, leaderboard links, and real-robot resources under Apache-2.0. Boundary: Challenge scores depend on a specific annual track, environment version, task definition, and evaluation rules. | |
| TactiDexShanghaiTech University; InstAdaptReviewed 2026-08-19 | Real-world tactile-guided dexterous manipulation benchmark 2026 | Single-hand dexterous manipulation; Bimanual manipulation; Human-to-robot skill transfer | Signals: Whole-hand pressure; Hand kinematics; Object 6D pose; Language; Task phase Sensors: Whole-hand tactile glove Robot: Bimanual Franka Inspire platform; Human demonstration capture | Standardized task and transfer metrics defined by the source Protocol: Aligns whole-hand tactile signals with multi-granularity kinematic and object states for standardized real-world evaluation. | The official project page documents the benchmark and demonstrations; a separate public download was not verified. Boundary: A 2026 preprint; inspect the current release, task definitions, splits, and access terms before comparison. | |
| VTDexManipZhejiang UniversityReviewed 2026-08-19 | Visual-tactile dexterous policy benchmark 2025 | Six dexterous manipulation tasks; Cross-task generalization; Modality ablation | Signals: Vision; Sparse binary touch; Proprioception Sensors: Sparse binary tactile sensing on the manipulation platform Robot: Dexterous manipulation platform described by the authors | Task success; Generalization; Noise and viewpoint robustness Protocol: Compares pretrained and non-pretrained visual-tactile policies across six tasks using a dataset spanning 10 daily activities and 182 objects. | The author project page links the ICLR paper plus code and dataset. Boundary: Reported policy gains are specific to the paper platform, sparse tactile representation, task suite, and RL training setup. |
Paper routes
Start with source-backed RoboSkin briefs
Tactile datasets / 2026-09-11TacVerse tests whether touch transfers between sensorsTacVerse compares shape, grating, and force perception across tactile sensors. Its public code and gated dataset are now identifiable, but transfer quality depends strongly on the task.
Tactile AI / 2026-08-22HT-Bench full-hand tactile benchmark for robot manipulationHT-Bench v2 pairs egocentric vision with millions of full-hand tactile frames, corrects the vision-to-tactile metric split, and adds four real-robot evaluations.Common questions
FAQ for this topic
What is a tactile robotics benchmark?
It is a reproducible evaluation contract for a tactile perception, representation, or robot-control question. It specifies inputs, hardware, data splits, metrics, baselines, and evaluation conditions.
Can benchmark scores be compared across tactile sensors?
Only when both systems use a shared protocol that controls sensor mounting, calibration, data, robot, task, metric, and evaluation version. Otherwise the numbers describe different experiments.
Is a tactile dataset automatically a benchmark?
No. A dataset supplies observations. A benchmark adds defined tasks, splits, metrics, baselines, and evaluation code or instructions.
Which baseline matters most for tactile manipulation?
Use the same robot, task, controller budget, and visual inputs with the tactile pathway disabled. That isolates whether touch improves the task outcome.