Towards Human-Like Manipulation through RL-Augmented Teleoperation and Mixture-of-Dexterous-Experts VLA

9 min read

Keep a pretrained VLA in charge of broad visual-language behavior, then route force and fingertip contact through sparse experts that add action-specific corrections exactly where contact begins to matter.

Backboneπ₀ / OpenPI-0Contact experts8 MLPsRoutingTop-1 per tokenAction horizon50 stepsInference10 Euler stepsViews3 × 224²

the design choice

Do not make contact sensing pretend to be vision

MoDE keeps the pretrained visual-language pathway intact and gives fast physical signals a dedicated route into action generation.

Naive concatenation

One state vector

Arm torque, fingertip contact, and proprioception are flattened into one interface despite having different meanings and time scales.

The paper reports that this conflation can degrade the pretrained policy instead of helping it.

Dedicated contact pathway

Attend, route, correct

1

Read VLA context

2

Choose an expert

3

Add a residual

mode-vla architecture

The backbone proposes; MoDE corrects

Force and tactile tokens read the same task context as the action expert, but only their residual corrections reach the final velocity field.

1

Pretrained proposal

3 camera views

Task prompt

Robot state

Noisy actions

π₀ action expert

PaliGemma context plus a Gemma-300M suffix produces 50 flow-matching action slots.

2

Contact refinement

Force f ∈ R¹⁴

Arm joint torques: 7 joints × 2 arms

Tactile g ∈ R⁶⁰

6D wrench: 5 fingertips × 2 hands

Self-attention + sparse router

top-1

Contact tokens attend to the visual-language prefix and denoising suffix, then each token selects one of eight MLP experts.

E1E2E3E4E5E6E7E8
3

Residual outputs

Force route

Arm action adjustment

Contact onset and insertion resistance

Tactile route

Hand action adjustment

Fingertip slip and grasp stability

Base suffix + residual → velocity field

why token-level routing matters

One sensor snapshot becomes 50 routing decisions

The same current reading is copied across the action horizon and given a different step encoding at every position. Experts can therefore change inside one predicted chunk.

h1

reach

E2

h2

reach

E2

h3

contact

E6

h4

insert

E3

track

E3

h50

settle

E7

These regime labels illustrate the specialization the architecture enables. The paper motivates contact-onset and steady-state experts, but does not publish expert-usage traces that prove each expert learned a named role.

the small hierarchical piece

MoDE refines control until the VLA delegates the hand

IMCopilot matters at inference, but it is a separate RL skill rather than one of the eight dexterous experts.

Option 1: VLA drives the hand

trigger ≤ 0.5

Flow matching generates arm and hand actions. MoDE adds force-guided arm residuals and tactile-guided hand residuals.

Option 2: IMCopilot drives the hand

trigger > 0.5

The RL rotation primitive overrides the VLA's hand actions. The VLA still produces the arm motion, refined by MoDE's force route.

what the experiments support

Contact sensing helps, with force carrying the larger average gain

Each result uses 20 real-robot trials. Apple peeling combines MoDE-VLA with the separate IMCopilot skill.

Taskπ₀Ours

Apple peeling

73% peel completion

0%30%

Tube rearranging

Bimanual

15%30%

Gear assembling

Force-sensitive

40%60%

Charger plugging

Force-sensitive

5%15%
Average success15% → 34%

Remove force

-11 pts

Average success falls from 34% to 23%. Contact onset and compliant insertion lose their strongest cue.

Remove tactile

-8 pts

Average success falls to 26%, mainly through more slip in grasp-intensive phases.

Important limit

No reported ablation isolates sparse MoE routing from the rest of the sensor-fusion pathway.

takeaways

What is worth carrying forward

Preserve the prior

Use residual corrections when a new sensor is most useful locally and the pretrained policy is already strong elsewhere.

Route by future step

Per-horizon tokens let one action chunk move through several contact regimes without paying for every expert at every position.

Separate semantics

Arm torque corrects arm motion; fingertip wrench corrects hand motion. The output structure mirrors what each sensor measures.

The mental model

MoDE-VLA is best understood as a contact-aware refinement layer on top of π₀, not as a collection of complete VLA policies. The pretrained backbone still interprets the camera views, instruction, robot state, and noisy action chunk. MoDE gives force and tactile sensing a separate route so those signals do not have to compete with vision and language inside an interface learned mostly from parallel-gripper data.

A single current force reading and a single tactile reading are each projected into the backbone width, copied across all 50 future action positions, and tagged with step encodings. After attention lets them read the VLA context and current denoising state, a top-1 router chooses one of eight expert MLPs independently for each token. This gives different points in the same action chunk room to specialize for free motion, contact onset, steady force, or slip.

The outputs remain deliberately narrow. Force-conditioned tokens correct arm actions, while tactile-conditioned tokens correct hand actions. Both are added residually to the action expert's suffix, so the side path can become quiet in free space instead of overwriting the useful behavior already stored in π₀. For the hardest in-hand rotation, the VLA can also emit a trigger that hands control to IMCopilot, while MoDE continues refining arm motion.

The evidence is promising but early. Across 20 trials on each of four tasks, the full system raises average success from 15% to 34%. Removing force costs 11 points and removing tactile costs 8, but the paper does not report a clean ablation that removes only the sparse expert block while retaining both sensor streams. Apple peeling also mixes MoDE with IMCopilot, so its improvement should not be credited to the VLA fusion module alone.

Sources

Built by Suveen.

www.suveenellawela.com v.2026.5