Poke &Wiggle

Leaderboard

Model

Claude Opus 5.5

Anthropic

#2

Rank · PAW-GEN-10-Agentic

Model information

Website
Website⁠ (opens in a new tab)
Released
September 2026

PAW-GEN-10-Agentic

Share of episodes rated Ok or Excellent.

Summary

Overall Success
39%
Evaluations
900
NominalEvaluated on the placements the model was trained on.InterpolationEvaluated on unseen placements between the training placements.ExtrapolationEvaluated on unseen placements outside the training placements' convex hull.
EnvironmentNominal · D10Trained on ~10 demonstrations per environment.Interpolation · D10Trained on ~10 demonstrations per environment.Extrapolation · D10Trained on ~10 demonstrations per environment.
Basic skills
Sort screws into bins100%#195% CI 89%–100%100%#195% CI 89%–100%73%#295% CI 56%–86%
Datum-corner alignment43%#495% CI 27%–61%47%#395% CI 30%–64%0%#495% CI 0%–11%
Pour screws into the funnel97%#295% CI 83%–99%93%#295% CI 79%–98%77%#395% CI 59%–88%
High precision
Connect DC jack47%#395% CI 30%–64%57%#395% CI 39%–73%13%#395% CI 5%–30%
Route cable through closed hoops0%#395% CI 0%–11%0%#395% CI 0%–11%0%#295% CI 0%–11%
Put tools in standing toolbox3%#295% CI 1%–17%7%#295% CI 2%–21%0%#295% CI 0%–11%
Bimanual
Place loaded boxes in container (bimanual)3%#395% CI 1%–17%13%#295% CI 5%–30%3%#395% CI 1%–17%
Place spray bottles upright, nozzle forward (bimanual handover)97%#295% CI 83%–99%90%#395% CI 74%–97%63%#395% CI 46%–78%
Tool use
Open the toolbox with a screwdriver17%#295% CI 7%–34%3%#195% CI 1%–17%7%#195% CI 2%–21%
Clamp electrical component33%#295% CI 19%–51%70%#295% CI 52%–83%20%#295% CI 10%–37%
Pooled44%#295% CI 38%–50%48%#295% CI 42%–54%26%#295% CI 21%–31%

Spatial Generalization

Spatial generalization for Claude Opus 5.5: SuccessClaude Opus 5.5: Nominal 44% (95% CI 38%–50%), Interpolation 48% (95% CI 42%–54%), Extrapolation 26% (95% CI 21%–31%).Success rate (%)04080Claude Opus 5.5: 44% (95% CI 38%–50%) on NominalClaude Opus 5.5: 48% (95% CI 42%–54%) on InterpolationClaude Opus 5.5: 26% (95% CI 21%–31%) on ExtrapolationNominalInterpolationExtrapolation

Per Environment (Success)

Model performance by environmentClaude Opus 5.5 overall: Sort screws into bins 91%, Datum-corner alignment 30%, Pour screws into the funnel 89%, Connect DC jack 39%, Route cable through closed hoops 0%, Put tools in standing toolbox 3%, Place loaded boxes in container (bimanual) 7%, Place spray bottles upright, nozzle forward (bimanual handover) 83%, Open the toolbox with a screwdriver 9%, Clamp electrical component 41%.Sort screws into binse01Datum-corner alignmente02Pour screws into the funnele03Connect DC jacke04Route cable through closed hoopse05Put tools in standing toolboxe06Place loaded boxes in container (bimanual)e07Place spray bottles upright, nozzle forward (bimanual handover)e08Open the toolbox with a screwdrivere09Clamp electrical componente10Basic skillsHigh precisionBimanualTool use

Rollouts

Mid-training dataM0
Task-specific dataD10
SpatialInterpolation

One clip per environment from this entry's evaluation, beside where each evaluation episode placed the environment's objects, coloured by its rating. Use the button beside each environment to open its evaluation dataset in the dataset viewer.

  1. Sort screws into bins
    Success
    100%
    Progress
    100%
  2. Pour screws into the funnel
    Success
    93%
    Progress
    93%
  3. Connect DC jack
    Success
    57%
    Progress
    83%
  4. Route cable through closed hoops
    Success
    0%
    Progress
    30%
  5. Open the toolbox with a screwdriver
    Success
    3%
    Progress
    50%
  6. Held-out benchmark environment
    Datum-corner alignment
    Success
    47%
    Progress
    47%
  7. Held-out benchmark environment
    Put tools in standing toolbox
    Success
    7%
    Progress
    25%
  8. Held-out benchmark environment
    Place loaded boxes in container (bimanual)
    Success
    13%
    Progress
    57%
  9. Held-out benchmark environment
    Place spray bottles upright, nozzle forward (bimanual handover)
    Success
    90%
    Progress
    91%
  10. Held-out benchmark environment
    Clamp electrical component
    Success
    70%
    Progress
    81%

Success by spatial configuration

e01 · Sort screws into bins
Nominal: 100% (95% CI 89%–100%)Interpolation: 100% (95% CI 89%–100%)Extrapolation: 73% (95% CI 56%–86%)NominalInterp.Extrap.
e02 · Datum-corner alignment
Nominal: 43% (95% CI 27%–61%)Interpolation: 47% (95% CI 30%–64%)Extrapolation: 0% (95% CI 0%–11%)NominalInterp.Extrap.
e03 · Pour screws into the funnel
Nominal: 97% (95% CI 83%–99%)Interpolation: 93% (95% CI 79%–98%)Extrapolation: 77% (95% CI 59%–88%)NominalInterp.Extrap.
e04 · Connect DC jack
Nominal: 47% (95% CI 30%–64%)Interpolation: 57% (95% CI 39%–73%)Extrapolation: 13% (95% CI 5%–30%)NominalInterp.Extrap.
e05 · Route cable through closed hoops
Nominal: 0% (95% CI 0%–11%)Interpolation: 0% (95% CI 0%–11%)Extrapolation: 0% (95% CI 0%–11%)NominalInterp.Extrap.
e06 · Put tools in standing toolbox
Nominal: 3% (95% CI 1%–17%)Interpolation: 7% (95% CI 2%–21%)Extrapolation: 0% (95% CI 0%–11%)NominalInterp.Extrap.
e07 · Place loaded boxes in container (bimanual)
Nominal: 3% (95% CI 1%–17%)Interpolation: 13% (95% CI 5%–30%)Extrapolation: 3% (95% CI 1%–17%)NominalInterp.Extrap.
e08 · Place spray bottles upright, nozzle forward (bimanual handover)
Nominal: 97% (95% CI 83%–99%)Interpolation: 90% (95% CI 74%–97%)Extrapolation: 63% (95% CI 46%–78%)NominalInterp.Extrap.
e09 · Open the toolbox with a screwdriver
Nominal: 17% (95% CI 7%–34%)Interpolation: 3% (95% CI 1%–17%)Extrapolation: 7% (95% CI 2%–21%)NominalInterp.Extrap.
e10 · Clamp electrical component
Nominal: 33% (95% CI 19%–51%)Interpolation: 70% (95% CI 52%–83%)Extrapolation: 20% (95% CI 10%–37%)NominalInterp.Extrap.