HumanoidBehavior
Buyer Room →
TECHNICAL EVIDENCE ROOM

Inspect the proof.
Not the pitch.

A public, machine-readable sample demonstrates the current measured MuJoCo execution path. The artifact records the requested seeds, policies, metrics, validation and provenance.

Open Guided Buyer Demo →Open Raw JSON ↗Acquisition Room ↗
10Measured runs
5Explicit seeds
2Policy configs
1Persisted artifact

Run identity

Benchmarkhumanoid-stand-v1
Environmenthb-humanoid-v1
EngineMuJoCo
Measuredtrue
Simulationtrue
Requested / evaluated10 / 10
Seeds42 · 1337 · 2026 · 7 · 99
Policiespolicy-a-stabilizer · policy-b-stabilizer

Validation & provenance

ValidationPassed — every requested seed was evaluated against every policy.
WorkflowMuJoCo Experiment
Git SHA1a82f15df8e68af712f378f3eb39f9f00dcd8fd9
GitHub Actions run35433718562

Selected measured output

policy-a-stabilizer survival_rate.mean = 0.3264 policy-b-stabilizer survival_rate.mean = 0.3304 policy-a-stabilizer control_cost.mean = 0.163122 policy-b-stabilizer control_cost.mean = 0.147768 targetSeconds = 5.0 fallRate = 1.0 for both policy configurations

These values describe this simulator experiment only. They are not physical-robot performance claims and are not presented as verified metrics for every public behavior.

Reproducibility path

The current architecture separates behavior specification, experiment configuration, browser-side measured MuJoCo execution, persistence and reporting. Buyers can inspect the repository and run the supported workflow independently.

Open Experiment Lab →Read API Docs →Review GitHub ↗