cs.ROOct 15, 2026

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

Authors: Bing Cheng, Yiran Mao, Zewei Ye, Yifan Jin, Xinyi Chen, Ruifeng Tu, Shuhuai Wang, Tao Lin, 

Organizations: School of AI, Shanghai Jiao Tong University · Nanjing University · Central South University · Beijing University of Posts and Telecommunications · 

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

General-purpose dexterous manipulation requires policies to transfer knowledge from related tasks to new tasks with only a few target demonstrations. Existing benchmarks often provide too many target demonstrations. They also leave the differences between source and target tasks uncontrolled. As a result, they cannot clearly distinguish knowledge transfer from task memorization. Our key insight is the Core-Gen principle. Core tasks are related tasks that share a functional interaction and provide source knowledge, while a Gen task is a novel target within the same functional family that has only a few demonstrations. This setup tests whether a policy can reuse functional interaction knowledge. Based on this principle, we propose Evo-DexBench, which controls both source-task relatedness and target-task exposure. It comprises 30 tasks in six functional families and a VR teleoperation pipeline. The resulting dataset covers three everyday scenes per task. It contains 1,248 multimodal demonstrations collected with simulated Shadow Hands equipped with 123 taxels. Experiments show that even the best-performing method achieves an average success rate of only 38.5%, highlighting the challenges existing methods face in transferring functional interaction knowledge. Further analysis shows that, across the two tasks, joint Core+Gen training improves performance by 12% and 38%, respectively, over training on Core followed by fine-tuning on Gen. This indicates that joint training makes better use of Core knowledge and that functionally matched Core data yields greater gains in success rate than mismatched data. We release the Evo-DexBench dataset and benchmark to support the evaluation of dexterous manipulation policies' generalization across related tasks.

Figures & tables

Figure 1: Overview of the Evo-DexBench. The benchmark comprises six functional task categories, each containing four Core tasks and one Gen task. Each task randomly switches among three scenes within its corresponding Kitchen or Studio environment during data collection and evaluation. The bottom-right panel reports the performance of representative models.
BenchmarkArm–handBimanualForceScenes/taskCore-GenData Collection
LIBERO✗✗✗1✗SpaceMouse teleoperation
RoboTwin 2.0✗✓✗1✗MLLM-based generation
Bi-DexHands✗✓✗1ML4RL policy
DexArt✓✗✗1✗RL policy
DexCraft✓✗✗1✗Keyboard teleoperation
HANDFUL✓✗✗1✗RL policy
Table 1: Comparison of embodied manipulation benchmarks.
Figure 2: Overall framework of the benchmark. The environment covers six task categories, three Kitchen/Studio scenes, 123 hand taxels, and variations in layout, cameras, lighting, and physical properties, while decoupling task logic from scene layout. Policies and keyboard/mouse or Quest 3 teleoperation share a unified interface.
FamilyShared interaction
Button-drivenDisplace a localized control to change device state.
Rotational manipulationApply directed rotation to a functional component.
Access-and-transferOpen an access mechanism, transfer an object, and restore its state.
Structured configurationConstruct a spatial arrangement under ordering or stability constraints.
AssemblyAlign complementary objects with two hands for insertion or fitting.
Articulated configuration controlStabilize one part while changing another’s relative configuration.
Table 2: Interaction families and shared functional interaction patterns.
Figure 3: Core-Gen task examples. The first and second rows show single-arm rotational manipulation and access-and-transfer tasks, respectively, while the third row shows bimanual articulated configuration control tasks. The first four columns show Core tasks, and the last column shows Gen tasks.
Figure 4: Task visualizations. The first column shows short-horizon tasks: pressing a desk lamp button and stapling paper. The second and third columns show long-horizon tasks: opening a cabinet with one hand and placing a tennis ball on the lower shelf, and assembling a sandwich with both hands.
Figure 5: Family-level baseline SR for the evaluated policies, with 95% confidence intervals.
FamilyTaskπ0.5\pi_{0.5}DP-TDP-CKineDexSmolVLA
Button- drivenLampButton30% / 35.715% / 12.626% / 28.724% / 23.241% / 35.2
WallSwitch60% / 74.135% / 57.868% / 79.776% / 85.666% / 78.0
PhoneHomeButton16% / 24.41% / 2.28% / 13.75% / 14.88% / 12.8
ToasterSlider42% / 75.249% / 79.664% / 84.063% / 84.037% / 74.2
DisplayPowerButton20% / 39.433% / 52.89% / 31.518% / 40.818% / 36.0
Rotational manipulationFaucetHandle21% / 29.124% / 27.219% / 23.016% / 22.012% / 22.2
Table 3: Standard evaluation on 15 tasks. Each entry is SR (%) / StageScore .
StrategyDisplayPowerButtonDishwasherKnob
Gen only15% / 29.736% / 55.1
Full Core →\rightarrow Gen8% / 23.68% / 29.3
Early Core →\rightarrow Gen9% / 24.622% / 41.2
Core + Gen joint20% / 39.446% / 62.5
Table 4: Few-shot adaptation strategies at 50 epochs. Each entry is SR (%) / StageScore .
Figure 6: Transfer heatmap and tactile analysis. Left: Change in π0.5\pi_{0.5} SR relative to Gen-only training when mixing each Core family with a Gen target. The diagonal contains the functionally matched pairs. Right: Task-wise SR change from adding tactile observations to KineDex.
MethodGen onlyButton-drivenRotational manip.Access-and-transfer
π0.5\pi_{0.5}36% / 55.140% / 61.046% / 62.532% / 52.4
DP-T28% / 42.127% / 51.041% / 56.641% / 56.0
KineDex42% / 57.939% / 48.955% / 67.951% / 63.9
Table 5: Core-source relatedness on DishwasherKnob across three models. Each entry is SR (%) / StageScore ; the functionally matched Rotational source is bold.
Appendix figures & tables13 assets

Supplementary material from the paper’s appendix.

Appendix

TaskDescriptionStage success conditions
LampButtonPress the lamp button and keep the lamp on.30%: index-finger force ≥\geq 0.15 N; 60%: button travel ≥\geq 70%; 100%: lamp remains lit for ≥\geq 1 s after triggering.
WallSwitchFlip the wall switch on.30%: hand tip within 0.20 m; 60%: angle ≥\geq 0.061 rad; 100%: switch angle ≥\geq 0.122 rad for 1 s.
PhoneHomeButtonPress and release the phone Home button.30%: correct contact force ≥\geq 0.12 N; 70%: contact held while travel ≥\geq 70%; 100%: button released, return travel ≤\leq 20%, and phone remains safely on the table.
ToasterSliderPress the toaster slider.30%: hand tip within 0.40 m of toaster; 60%: hand tip within 0.35 m of slider point; 100%: toaster slider displacement ≥\geq 0.062 m.
DisplayPowerButtonPress the display power button.30%: fingertip within 0.05 m; 60%: contact force ≥\geq 0.15 N; 100%: contact followed by travel ≥\geq 70% and lasted for 3 control steps.
Appendix
Table 6: Task definitions for the Button-driven family.
TaskDescriptionStage success conditions
FaucetHandleTurn the faucet handle open.30%: handle contact force ≥\geq 0.40 N; 60%: opening travel ≥\geq 50%; 100%: opening travel ≥\geq 80%.
MouseWheelHold the mouse and scroll its wheel.25%: mouse contact force ≥\geq 0.15 N or wheel rotation >0.03>0.03 rad; 50%: wheel contact force ≥\geq 0.15 N or rotation >0.03>0.03 rad; 100%: wheel rotation ≥\geq 1.5 rad, mouse linear speed ≤\leq 0.04 m/s, and angular speed ≤\leq 0.60 rad/s.
CoffeeKnobTurn the coffee-machine knob to its designated target setting.30%: knob contact force ≥\geq 0.40 N; 50%: turning toward the target at ≥\geq 0.03 rad/s with valid knob contact has been recorded; 75%: angle error ≤\leq 20 ∘ ; 100%: angle error ≤\leq 6 ∘ , knob angular speed ≤\leq 0.20 rad/s.
KettleLidTabRotate the kettle-lid tab to the notch.30%: thumb and index contacts, each ≥\geq 0.35 N; 55%: positive rotation ≥\geq 5 ∘ at ≥\geq 0.03 rad/s; 80%: angle in 60 ∘ –90 ∘ ; 100%: angle error ≤\leq 9 ∘ , tab angular speed ≤\leq 0.50 rad/s, kettle displacement ≤\leq 0.01 m, and tilt ≤\leq 3 ∘ .
DishwasherKnobTurn the dishwasher knob to its target setting.30%: knob contact force ≥\geq 0.40 N; 60%: angle error ≤\leq 45 ∘ ; 100%: effective turn recorded, angle error ≤\leq 9 ∘ , and angular speed ≤\leq 1.00 rad/s.
Appendix
Table 7: Task definitions for the Rotational manipulation family.
TaskDescriptionStage success conditions
CupFromCabinetOpen the cabinet, remove the cup, and place it upright on the counter.25%: door angle >45∘>45^{\circ} ; 50%: valid cup-body/handle grasp (force ≥\geq 0.40 N); 75%: cup clears the cabinet opening; 100%: cup is on the counter, tilt ≤\leq 15 ∘ , released, with linear speed ≤\leq 0.05 m/s and angular speed ≤\leq 0.60 rad/s for 0.40 s.
MicrowaveBowlOpen the microwave, place the bowl inside, and close the door.25%: door opening ≥\geq 45%; 50%: two-finger bowl grasp, each force ≥\geq 0.30 N, held 0.12 s; 75%: bowl fully inside, support ≥\geq 0.20 N, tilt ≤\leq 35 ∘ , released, linear speed ≤\leq 0.05 m/s and angular speed ≤\leq 0.60 rad/s for 0.35 s; 100%: placement confirmed and the same bowl conditions hold with the door closed for 0.30 s. Closed condition is at opening ≤\leq 5% and speed ≤\leq 0.08 rad/s, and exits above 8%.
TennisBallTransferMove the tennis ball from shelf 1 to shelf 2 and close the cabinet.20%: door opening ≥\geq 55%; 45%: two-finger grasp with force ≥\geq 0.40 N per finger; 75%: ball in shelf-2 target, supported, released, linear speed ≤\leq 0.03 m/s, angular speed ≤\leq 0.60 rad/s for 5 steps; 100%: target was placed, the door is confirmed closed with speed ≤\leq 0.01 m/s for 5 steps, and the ball satisfies the placement conditions.
BananaFromDrawerOpen the drawer, place the banana on the desk, and close the drawer.25%: drawer opening ≥\geq 45%; 40%: valid grasp held 0.15 s; 60%: banana fully clears drawer; 85%: desk support ≥\geq 0.20 N, released and still for 0.50 s; 100%: banana remains released and stable on the desk while the drawer is confirmed closed with speed ≤\leq 0.01 m/s for 0.15 s.
CandleToCabinetOpen the upper glass cabinet, place the candle inside, and release it.25%: an upper door angle >60∘>60^{\circ} ; 50%: at least two fingers, each ≥\geq 0.30 N; 75%: candle bounding box inside cabinet; 100%: candle remains inside and upright with base-height error ≤\leq 0.025 m, released, tilt ≤\leq 20 ∘ , linear speed ≤\leq 0.03 m/s, and angular speed ≤\leq 0.50 rad/s for 1 s.
Appendix
Table 8: Task definitions for the Access-and-transfer family.
TaskDescriptionStage success conditions
TablewarePlacementPlace the plate in the center, fork left, and knife right.30%: any one piece correctly placed; 60%: any two pieces correctly placed; 90%: all three correctly placed. Correct placement requires position error ≤\leq 0.12 m, tilt ≤\leq 35 ∘ , utensil yaw error ≤\leq 35 ∘ , release, linear speed ≤\leq 0.04 m/s and angular speed ≤\leq 0.60 rad/s; 100%: all three have valid placements, left/right relations and clearance for 1 s.
HammerBlockAssemblyArrange three blocks into the reference hammer shape.25%: one block position/height error ≤\leq 0.015 m, yaw/tilt ≤\leq 8 ∘ , released; 50%: at least two blocks have valid poses and releases, with at least one prescribed adjacency pair (center error ≤\leq 0.020 m, edge angle ≤\leq 12 ∘ ); 75%: all three target poses and releases valid; 100%: all three target poses and prescribed adjacency pairs valid, all blocks released, linear speed ≤\leq 0.03 m/s, and angular speed ≤\leq 0.50 rad/s for 0.50 s.
SandwichAssemblyPlace the patty and top bread on the pre-stacked bottom bread and cheese.40%: either movable ingredient correctly placed; 80%: both movable ingredients correctly placed. Placement requires position error ≤\leq 0.050 m, height error ≤\leq 0.018 m, tilt ≤\leq 25 ∘ , valid support and overlap, release, linear speed ≤\leq 0.08 m/s and angular speed ≤\leq 1.20 rad/s; 100%: all four layers satisfy the ordered support, pose, overlap, release and speed conditions for 0.25 s.
RingTowerAssemblyStack the red, blue, and yellow rings from largest to smallest.30%: red ring correctly on bottom; 60%: red and blue rings correctly placed; each ring center within 0.028 m of post axis, height error ≤\leq 0.018 m, tilt ≤\leq 22 ∘ ; 100%: correct three-ring order and poses, with linear speed ≤\leq 0.06 m/s and angular speed ≤\leq 0.80 rad/s for 20 steps.
BoxPackingPack three shaped objects into the box and close the lid.20%: at least one object fully inside; 40%: at least two objects fully inside; 75%: all three objects fully inside; 100%: all three objects fully inside and maximum opening across all lids ≤\leq 4%.
Appendix
Table 9: Task definitions for the Structured configuration family.
TaskDescriptionStage success conditions
PenCappingRemove the pen, grasp its cap, and cap the pen.20%: either object previously grasped ( ≥\geq 0.15 N per parent link); 40%: both objects previously grasped; 70%: axis error ≤\leq 18 ∘ , lateral error ≤\leq 0.015 m, tip inside cap range; 100%: axis error ≤\leq 18 ∘ , lateral error ≤\leq 0.010 m, and insertion depth ≥\geq 0.015 m.
CupSleeveFitSlide a sleeve from the cup bottom to its middle.20%: either object previously grasped ( ≥\geq 0.18 N per parent link); 40%: both objects previously grasped; 70%: axis error ≤\leq 15 ∘ , lateral error ≤\leq 0.018 ×\times scale, center height in [ −0.055-0.055 , 0.0400.040 ] ×\times scale; 100%: center height in [ 0.0620.062 , 0.1000.100 ] ×\times scale, lateral error ≤\leq 0.010 ×\times scale, axis error ≤\leq 15 ∘ . Here scale =0.72=0.72 .
BottleFunnelInsertionAlign a funnel with a bottle and insert it.20%: either object previously grasped ( ≥\geq 0.18 N per parent link); 40%: both objects previously grasped; 70%: axis error ≤\leq 15 ∘ , lateral error ≤\leq 0.018 m, tip near mouth; 100%: axis error ≤\leq 15 ∘ , insertion depth ≥\geq 0.010 m, tip height in [0.1031, 0.1660] m in the bottle frame, lateral error ≤\leq 0.010 m, linear speed relative to bottle ≤\leq 0.06 m/s, and relative angular speed ≤\leq 1.00 rad/s for 3 steps.
PlugInsertionHold a plug and socket and fully insert the plug.25%: opposite hands hold both objects, at least two parent contacts per object at ≥\geq 0.20 N; 50%: both actual pins align with the socket channels, direction error ≤\leq 12 ∘ , tips near the openings; 75%: both aligned pins enter their channels; 100%: both pins remain aligned with depth ≥\geq 0.01575 m each, linear speed relative to socket ≤\leq 0.15 m/s, and relative angular speed ≤\leq 1.00 rad/s.
ScissorsSheathingInsert the tips of closed scissors into the sheath.20%: scissors previously grasped with at least two parent contacts, each ≥\geq 0.20 N, for 0.10 s; 60%: both real blade tips align with the channel cross-section, axis error ≤\leq 12 ∘ , axial distance from entrance ≤\leq 0.025 m; 100%: both blade tips inside the sheath with 0.002 m geometric tolerance, axis error ≤\leq 12 ∘ , maximum tip speed relative to sheath ≤\leq 0.08 m/s, and relative angular speed ≤\leq 1.00 rad/s for 0.10 s.
Appendix
Table 10: Task definitions for the Assembly family. For PenCapping, CupSleeveFit and BottleFunnelInsertion, a grasp requires the same hand to contact an object through at least two distinct tactile parent links, each at or above the listed force threshold, continuously for 0.10 s.
TaskDescriptionStage success conditions
LaptopClosingStabilize the laptop base and close the screen to about 25 ∘ .30%: opposite hands contact base/screen at ≥\geq 0.40 N, base displacement ≤\leq 0.040 m and tilt ≤\leq 5 ∘ ; 60%: screen reaches 60 ∘ ±\pm 8 ∘ ; 100%: screen reaches 25 ∘ ±\pm 5 ∘ , base remains stable, and absolute screen angular speed ≤\leq 1.00 rad/s for 0.25 s.
StaplerPaperMarkInsert paper, press the stapler, then move it away.20%: paper force ≥\geq 0.50 N or displacement ≥\geq 0.03 m; 50%: paper enters staple area with tilt ≤\leq 35 ∘ ; 70%: paper remains in area, stapler press-arm qpos ≤\leq 0.10 rad, mark created; 100%: after a mark is created, the current stapling point and the paper mark remain ≥\geq 0.06 m apart for 5 steps.
ScissorsOpeningStabilize scissors and open the moving handle.40%: support-handle contact ≥\geq 0.15 N or base linear speed <0.30<0.30 m/s has been recorded, and current base linear/angular speeds are ≤\leq 0.30 m/s and ≤\leq 6.00 rad/s; 70%: moving-handle contact ≥\geq 0.15 N or proximity to target has been recorded, and current angle error ≤\leq 0.50 rad; 100%: opening angle 1.263 ±\pm 0.20 rad, scissors-base linear speed ≤\leq 0.30 m/s, and angular speed ≤\leq 6.00 rad/s for 10 steps.
BoxLidOpeningStabilize a box and open its hinged lid.30%: task contact ≥\geq 0.40 N with box displacement ≤\leq 0.10 m, tilt ≤\leq 7 ∘ , linear speed ≤\leq 0.05 m/s and angular speed ≤\leq 0.50 rad/s has been recorded; 60%: opening past ≥\geq 10 ∘ with a stable box, lid contact ≥\geq 0.40 N and positive opening velocity ≥\geq 0.02 rad/s has been recorded; 100%: lid angle 90 ∘ –105 ∘ , box stable, and lid angular speed ≤\leq 0.25 rad/s for 1 s.
GlassesFoldingFold both eyeglass temples into the storage position.40%: left temple angle ≥\geq 0.55 rad has been recorded; 80%: right temple angle ≥\geq 0.45928 rad has been recorded; 100%: right/left angles satisfy either ( ≥\geq 0.91, ≥\geq 0.55) rad or ( ≥\geq 0.45928, ≥\geq 1.08975) rad.
Appendix
Table 11: Task definitions for the Articulated configuration control family.
GroupingThumbIndexMiddleRingLittlePalmTotal
By finger/area292118181423123
RegionDistalMiddleProximalPalmSub-palm Total
By anatomy622414212 123
Appendix
Table 12: Distribution of the 123 taxels on one Shadow Hand.
PolicyNative modelPolicy interfaceModel-specific adaptation
π0.5\pi_{0.5}OpenPI JAX; pretrained flow-matching VLAThree RGB views, language, 31-D/62-D state; 50-step chunksPad state/action interfaces to 32-D/64-D and discard padded action dimensions
SmolVLALeRobot 0.4.0; SmolVLM2-500M with flow-matching expertThree/four RGB views, language, 31-D/62-D state; 50-step chunksPad to 32-D/64-D and expand four projection tensors for bimanual control
DP-CConvolutional Diffusion Policy with temporal U-NetTwo observation steps; 30-step external action horizonUse a 32-step internal target for U-Net divisibility and discard the final two steps
DP-TTransformer Diffusion PolicyTwo observation steps; 30-step action horizonMap benchmark compact state and action channels to the released workspace
KineDexMulti-view ResNet-18, per-finger tactile CNN, conditional U-NetTwo observation steps; 16-step horizon; motion plus five force targets per handMap 123 taxels to 100 finger taxels, expand the action head, and add bounded force feedback
KineDex No-TouchBackbone-matched KineDex variantVisual/proprioceptive input; 30 motion values per handFeed zero tactile input and remove force supervision and force feedback
Appendix
Table 13: Policy implementations and Evo-DexBench compatibility adaptations.
PolicyEpochsPeak LRMinimum LRBatchsize
π0.5\pi_{0.5}25 †5 ⁣× ⁣10−65\!\times\!10^{-6}5 ⁣× ⁣10−75\!\times\!10^{-7}32
SmolVLA25 ‡1 ⁣× ⁣10−41\!\times\!10^{-4}2.5 ⁣× ⁣10−62.5\!\times\!10^{-6}32
DP-T6001 ⁣× ⁣10−41\!\times\!10^{-4}1 ⁣× ⁣10−61\!\times\!10^{-6}32
DP-C6001 ⁣× ⁣10−41\!\times\!10^{-4}1 ⁣× ⁣10−61\!\times\!10^{-6}32
KineDex6001 ⁣× ⁣10−41\!\times\!10^{-4}1 ⁣× ⁣10−51\!\times\!10^{-5}32
KineDex No-Touch6001 ⁣× ⁣10−41\!\times\!10^{-4}1 ⁣× ⁣10−51\!\times\!10^{-5}32
Appendix
Table 14: Training hyperparameters. KineDex No-Touch uses the same settings as Full KineDex.
PolicyTaskOverallNom.Cam.Light.Phys.Spat.
π0.5\pi_{0.5}LampButton30% / 35.74015354020
WallSwitch60% / 74.16050755560
PhoneHomeButton16% / 24.4302510015
ToasterSlider42% / 75.24055502045
DisplayPowerButton20% / 39.42515152025
FaucetHandle21% / 29.1151554030
Appendix
Table 15: Complete baseline results. Overall is SR (%) / StageScore; Nom., Cam., Light., Phys., and Spat. are condition-specific SR (%).
Gen targetEpochsTraining dataSR / StageScore
DisplayPowerButton50Gen only15% / 29.7
 50Button-driven Core + Gen20% / 39.4
 50Rotational manip. Core + Gen17% / 34.7
 50Access-and-transfer Core + Gen12% / 33.0
DishwasherKnob50Gen only36% / 55.1
 50Button-driven Core + Gen40% / 61.0
Appendix
Table 16: Complete π0.5\pi_{0.5} Core-source relatedness results. Entries are SR (%) / StageScore; bold marks the strongest retained result for each target.
Gen targetTraining dataSR / StageScore
DisplayPowerButtonGen only15% / 29.7
 LampButton + ToasterSlider + Gen17% / 32.9
 WallSwitch + PhoneHomeButton + Gen20% / 39.3
 Four matched Core + Gen20% / 39.4
DishwasherKnobGen only36% / 55.1
 MouseWheel + CoffeeKnob + Gen35% / 53.6
Appendix
Table 17: Core-set breadth/composition ablation for π0.5\pi_{0.5} . Entries are SR (%) / StageScore; bold marks the best result for each target.
TaskNo-TouchFull
FaucetHandle15% / 16.316% / 22.0
MouseWheel46% / 67.053% / 72.5
CoffeeKnob25% / 63.933% / 68.9
KettleLidTab54% / 70.762% / 75.3
DishwasherKnob60% / 73.555% / 67.9
LaptopClosing72% / 77.769% / 75.9
Appendix
Table 18: Task-wise KineDex tactile ablation. Entries are SR (%) / StageScore.
More in this topic

Dexterous Manipulation

Dexterous manipulation studies how robots use multi-fingered hands to grasp, reorient, and interact with objects through contact.

See all 143 papers

Explore similar work

CardsList
Jul 9, 2026

DexVerse: A Modular Benchmark for Multi-Task, Multi-Embodiment Dexterous Manipulation

Building general-purpose dexterous manipulation policies requires benchmarks that go beyond isolated tasks to systematically evaluate policies across diverse interaction modes, sensory conditions, and robot embodiments. However, existing benchmarks remain limited in task and data diversity, embodiment coverage, or controllable visual variation, hindering studies of cross-task and cross-embodiment generalization. We present DexVerse, a large-scale and modular benchmark for dexterous manipulation. DexVerse includes 100 tasks spanning a broad range of manipulation skills, including object grasping and relocation, articulated-object interaction, functional tool use, bimanual coordination, non-prehensile control, contact-rich behaviors, multi-goal execution, and long-horizon multi-stage task completion. It supports 3 robot arms and 6 dexterous hands, and is extensible to new tasks, assets, and embodiments. To evaluate visuomotor generalization, DexVerse provides configurable visual variations in textures, background, lighting, and camera viewpoints. We further provide a VR-based teleoperation interface and 3,180 demonstrations with synchronized proprioceptive, RGB, depth, point-cloud, and state observations. We benchmark representative methods, including Diffusion Policy, DP3, OpenVLA, and π0.5π_{0.5}, across 19 tasks. Results reveal substantial challenges in task generalization and visuomotor robustness, establishing DexVerse as a promising testbed for general-purpose dexterous manipulation. Project page: https://ycyao216.github.io/DexVerse.site

Yunchao Yao, Zhuxiu Xu, Tianqi Zhang +12

UNC-Chapel Hill · The University of Hong Kong · UC Berkeley

Dexterous ManipulationLong-Horizon Robotic ManipulationBimanual Robotic ManipulationGeneralization in Robotic ManipulationCross-Embodiment Robot LearningMulti-Task Robotic Manipulation+1
AskPDF
May 15, 2026

DexJoCo: A Benchmark and Toolkit for Task-Oriented Dexterous Manipulation on MuJoCo

Achieving human-level manipulation requires dexterous robotic hands capable of complex object interactions. Advancing such capabilities further demands standardized benchmarks for systematic evaluation. However, prior work does not focus on task-oriented dexterous manipulation. Existing benchmarks also often lack high quality human demonstrations and an easy-to-use evaluation pipeline. In this paper, we present DexJoCo, a benchmark and toolkit for task-oriented dexterous manipulation, comprising 11 functionally grounded tasks that evaluate tool-use, bimanual coordination, long-horizon execution, and reasoning. We develop a low-cost data collection system and collect 1.1K trajectories across these tasks, with support for domain randomization to assess robustness. We benchmark modern models under diverse settings, including visual and dynamics randomization, multi-task training, and action-head adaptation. Through extensive empirical analysis, we identify several important insights and common limitations of current policies in dexterous manipulation, highlighting key challenges for future research in dexterous hand robot learning. Code is available at: https://dexjoco.github.io

Hanwen Wang, Weizhi Zhao, Xiangyu Wang +11

NLPR & MAIS, CASIA · SJTU · MBZUAI +3

Dexterous ManipulationLong-Horizon Robotic ManipulationBimanual Robotic ManipulationRobot Manipulation Benchmarks
AskPDF
Oct 19, 2026

HandsOn: A Large-Scale Simulation Benchmark for Generalizable Dexterous Manipulation

General-purpose manipulation requires reliable behavior across embodiments, objects, and task demands. We introduce HandsOn, a composable simulation platform and benchmark for cross-embodiment dexterous manipulation. It supports five robotic arms, six dexterous hands, and approximately 200 rigid, articulated, and deformable assets in randomized scenes. The benchmark pairs 18,000 curated human demonstrations with 66 human-verified tasks: Core-50 supports demonstration-based learning and execution tests, while Transfer-16 evaluates asset and relation transfer, skill composition, and deformable manipulation without target-task training demonstrations. Qualitative real-robot rollouts illustrate sim-to-real transfer. Across diffusion, vision-language-action, world-action, and prompted VLM-based methods, the best trained multi-task policy achieves 29.1% Core-50 success. Without benchmark finetuning, GPT-6 using move_eef for palm-and-finger control achieves 5.2% on Core-50 and 1.9% on Transfer-16 over 660 trials. Trained-policy stage diagnostics expose failures in contact establishment, action coordination, and maintaining achieved goals, with additional sequencing errors in composed tasks. HandsOn provides a shared basis for investigating manipulation generalization and the limits of reliable physical interaction.

Tianwei Luo, Yiming He, Hongjin Wu +14

Tsinghua University · NatureWill · Beijing University of Posts and Telecommunications +1

Robotic ManipulationRobot Manipulation BenchmarksGeneralization in Robotic ManipulationCross-Embodiment Robot LearningRobot Manipulation Datasets
AskPDF