# Proof-of-Validation (PoV)

**Proof-of-Validation (PoV)** is a decentralized scoring framework used to evaluate and rank inference nodes in the **Orby** AI network. Validators compute performance metrics for AI agents or inference nodes and must prove — via zero-knowledge proofs — that scoring was executed correctly.

This zk-verified scoring ensures a reliable reputation system, which is essential for both network health and RWA-linked models that generate income based on performance.

**🔧 How It Works:**

1. **Validator retrieves performance data** for a model or inference node.
2. **Scoring function is executed locally**, encapsulated in a zk-circuit.
3. **A zk-proof of execution is generated** and submitted on-chain.
4. **Scores and proofs are recorded and validated** via Ethereum smart contracts.

**🔐 Security Guarantees:**

* Prevents score manipulation or duplication across validators.
* Validates that scoring was computed honestly using on-chain verifiable circuits.
* Publishes score hashes and proofs to Ethereum for auditability and downstream automation.

**🔁 Flow:**

```
Performance Metrics → Local Evaluation (zk-circuit) → zk-Proof + Scores → Ethereum Verifier → On-Chain Reputation Update
```

**📊 Use Cases:**

* **Node Reputation:** Reliable scoring ensures high-performing nodes are prioritized for critical inference tasks.
* **Model Quality Tracking:** Models with better usage metrics and inference success rates can receive more traffic and fees.
* **RWA Revenue Optimization:** RWA-linked models or agents can be scored transparently, with zk-proven logic determining payouts to token holders based on usage and performance quality.

**🌍 RWA Relevance:**

In the RWA context, PoV is the accountability engine. It provides cryptographic proof that a model or AI agent is performing as expected — forming the basis for royalty distribution, dynamic pricing, or trust-weighted access. High-score agents can command premium rates, attract capital flows, and participate in enterprise or institutional deployments.
