Appearance
Policy Validation
Policy validation is an independent evaluation of your robot policy on Nepher's simulation-truth pipeline — the same headless Isaac Sim + eval-nav stack that settles public tournaments, run on held-out EnvHub scenes.
Use it to benchmark an in-house policy before deployment, compare it against tournament winners, or as the evaluation step toward a Nepher Verified credential.
What You Submit
| Field | Description |
|---|---|
| Policy name | A label for the checkpoint or policy being evaluated |
| Policy source | Where it comes from: your company team, Nepher core, or a Nepher engineer |
| Description | Robot, task, and what you want measured |
| Deployment support | Optionally request help taking the validated policy toward deployment |
How to Request
- From the contact form at nepher.ai/enterprise/contact, choose Policy validation / deployment support.
- From an EnvHub environment page, use Validate a Policy with this Environment to start a request scoped to that environment.
The Nepher team scopes the evaluation with you (environments, metrics, episodes, seeds), runs it, and returns the results. Follow the conversation in nepher.ai/enterprise/requests if you chose website replies.
Integrity
- Evaluation uses held-out scenes; hidden suites and seeds are never shared, so results reflect generalization rather than tuning to the test.
- The same pipeline is used for public tournaments, so scores are comparable with published tournament results for the same task.
- Sponsorship or payment never grants control over hidden tests or scoring.
Related
- Nepher Verified — turn a successful evaluation into a public credential
- EnvHub — standardized environments used for evaluation
- Enterprise Overview