The Case for a Validator Network

2. How It Works Today

Two separate processes run in parallel. On the bank side, the trading desk produces a mark using an internal model. A separate team inside the same bank, the price verification desk, checks that mark against dealer quotes from other banks and pricing service data. This is supposed to happen monthly. On the fund side, the fund hires a valuation agent (Houlihan Lokey, Kroll) to independently estimate what the assets are worth. The agent delivers a report quarterly, typically 60 to 90 days after the reporting date. That report goes to the fund's auditor for annual review.

The bank has its number. The fund has its number. Most of the time they are close enough. When they diverge, dispute resolution kicks in: third party dealers submit two way quotes, midpoint resolves it.

Multiple checks on both sides. On paper, thorough. In practice, every layer has a problem.

The bank's price verification desk is inside the bank. Same employer, same building, same P&L ultimately. The fund's valuation agent is independent, but the fund hires the agent, pays the agent, and can fire the agent. Whether the agent's other clients have exposure to the same asset is not programmatically verifiable. Independence on both sides is organizational, not economic.

Dispute resolution fails when it matters most. The process gives 8 to 16 hours to gather quotes, but for illiquid assets, dealers refuse because there is no trade incentive. When too few come back, the fallback is the Calculation Agent's own estimate. The independence requirement collapses into the conflict it was designed to prevent.

Verification is periodic. The bank's price check is monthly. Valuation agents are quarterly. Auditors are annual. Between checkpoints, the Calculation Agent's mark is the only mark. And the marks themselves are weak: for Level 3 assets, firm executable quotes are rare. Most of the infrastructure runs on indicative quotes. “Roughly 88ish, don't hold me to it.”

3. Proof of Stake, Not Proof of Paperwork

Ravariant built Liquidity Adjusted Valuation to produce defensible marks for assets that have no price. But a mark from one source is another Calculation Agent opinion. The question is how to make it verifiable through independent consensus. The answer is a validator network. Not validators in the blockchain sense, running nodes and ordering transactions. Validators here are independent pricing agents: risk firms, credit funds, data providers, and secondary buyers who submit marks on the value of an asset, backed by staked capital, with no exposure to the position they are marking.

A 50 page valuation report from Houlihan Lokey exists because traditional finance has no other way to verify that independent work was done. The pages create the appearance of independent work. The actual independence is unverifiable. Every report ends with: “this opinion should not be relied upon as a guarantee of value.”

Two things matter for a credible mark. Did the person producing it have any reason to want it higher or lower? Are they willing to put something behind it? On chain eligibility verification answers the first. Staked capital answers the second. The report was the best approximation of trust that the old infrastructure could produce. Stake is the actual thing.

Methodology documentation still gets published to IPFS, linked to each submission by a content hash. But the protocol does not need it to function. The consensus and dispersion do the work that the 50 pages used to do.

4. How It Works

Pricing runs in epochs. Validators must have zero exposure to the asset they are marking: no token holdings, no lending positions, no LP exposure, no vault positions. Unlike traditional finance, where independence is self reported on a form, here it is verifiable against actual on chain state.

Eligibility is optimistic. The protocol assumes the validator is clean. Anyone can challenge by submitting on chain proof of exposure. If the challenge succeeds, marks get thrown out and stake gets slashed. Same pattern as optimistic rollups: assume valid, prove fraud.

Validators submit a bid and an offer. Sealed. Hashed. Committed before anyone can see anyone else's mark. After the commit window closes, all marks are disclosed simultaneously and the contract verifies each hash.

Aggregation computes a stake weighted median of midpoints. Three outputs get published on chain: the consensus mark, the dispersion score, and the confidence level. Any lending protocol reads these and sets margin accordingly. Wide validator disagreement means higher collateral requirements, automatically. The same independent price verification that banking regulators require, running continuously, with no human in the loop.

5. The Bid/Offer

The midpoint is the mark. The width is the confidence. A validator who says “I'd buy at 90 and sell at 91” is making a very different statement from one who says “somewhere between 84 and 96.”

Accounting standards require that price quotes “represent a firm commitment to transact.” A number without capital behind it is a guess. In traditional finance, the two way quote enforces this: dealers provide both a bid and an offer, and if hit, they transact. The two way structure forces honesty.

Tighter spread means higher fee but more risk. If the realized value falls outside the submitted spread, the stake gets slashed proportionally to the tightness. Quote tight and be wrong, lose more. Quote wide and be wrong, lose less.

This does three things. Prevents directional bias because you cannot mark everything low while also publishing an offer. Communicates uncertainty natively, which is the information regulators require for price verification adjustments. And creates the on chain equivalent of “firm commitment to transact” without requiring actual transactions. The stake is the commitment.

6. What Gets Slashed

Without slashing, validators have asymmetric payoff: get paid for showing up, no cost for being wrong. That is the valuation agent problem again.

But illiquid assets do not have a “right answer” in real time. Slashing for inaccuracy on a Level 3 asset would destroy the network. The distinction matters: dishonesty is when a validator hides exposure, freezes their mark, or tries to manipulate consensus. Inaccuracy is inherent to the asset class. Punish the first. Protect the second.

Single outlier marks are protected. If a validator marks a fund at 75 when everyone else says 95, that might be the most valuable signal in the network. The commit-reveal mechanism already prevents validators from seeing consensus before submitting. If an outlier is later proven right by a realized value, that validator gets a large retrospective bonus. The system actively encourages honest disagreement.

7. Who Participates and Why

In traditional finance, independent marks come from dealer desks, valuation agents, pricing services, and fund administrators. The on chain validator set maps to the same parties. The difference: the incentive is explicit, programmatic, and tied to accuracy rather than relationships.

The last row matters most. In traditional finance, the strongest price signal comes from a bids wanted in competition process, where holders ask buyers to submit sealed bids and the bidders who win take delivery. A dealer who bids 88 and wins owns the position. On chain, the same process runs through commit-reveal. When a position gets liquidated, these bids determine the sale price, which feeds back into retrospective scoring of every validator who marked the asset. The sealed bid closes the loop between ongoing marks and actual outcomes.

Validators get paid. The fee pool comes from protocols consuming the price, not from issuers. Three layers: base fee for showing up on time, spread bonus for tighter quotes, reputation multiplier for track record. A percentage of each epoch gets reserved for retrospective bonuses. When a realized value materializes, the pool distributes to validators whose marks were closest. Accuracy compounds into reputation. Reputation compounds into fee share.

There is also a reciprocity dynamic. Dealers quote for each other because they know they will need quotes in return. Same here. A risk firm that validates RAVA marks builds a public track record. When that firm tokenizes its own product and needs a price feed, the network it helped build is already there.

8. What It Replaces

The lending protocol replaces the bank. Goldman provides leverage, acts as Calculation Agent, determines margin calls, resolves disputes. On chain, Morpho provides leverage. The pricing comes from RAVA, produced by independent validators. The protocol consumes the mark. It does not produce it. The Calculation Agent conflict disappears.

9. A Concrete Example

A tokenized credit fund wants to be listed as collateral on a Morpho lending market. The Morpho curator cannot list it because there is no credible price feed. Chainlink provides spot prices for liquid tokens. For an illiquid fund with no exchange price, there is nothing.

The fund manager registers the asset in the RAVA Asset Registry. Pricing frequency is set to weekly. Five risk analytics firms, credit funds, and data providers register as validators. They stake capital. The protocol verifies each has zero exposure to the fund.

First epoch runs. Five validators submit sealed hashes of their bid/offer quotes. After the reveal, the aggregation contract computes a stake weighted median of 91.2, with a dispersion of 1.8. The Morpho curator subscribes to the feed. Based on the dispersion, margin is set at 1.15x the base rate. If dispersion tightens as validators converge, margin decreases. If it widens during stress, margin increases. No phone calls. No quarterly reports.

A borrower posts the fund token as collateral and borrows against the RAVA mark. Six months later, the fund processes a redemption at an NAV of 90.8. The settlement reporter submits the realized value on chain. The retrospective scoring module compares it against all historical marks. Validator A, who consistently submitted midpoints near 91 with tight spreads, gets a bonus. Validator D, who consistently marked at 95 with wide spreads, gets nothing slashed but their reputation decays. Next epoch, Validator A's marks carry more weight.

10. Ravariant

The principle behind independent pricing has existed for decades. The implementation has always been constrained by the infrastructure. You cannot verify independence programmatically when it is a relationship between legal entities. You cannot enforce commitment to a mark when the mark is an email. You cannot produce continuous verification when verification requires a human to review a document.

On chain infrastructure removes those constraints. Independence verified against wallet state, every epoch. Commitment backed by staked capital. Verification through consensus rather than periodic reviews. Dispersion feeding directly into margin calculations. When a position is finally liquidated, the sale price gets recorded on chain and every historical mark is scored against what actually happened.

The principle does not change. The infrastructure finally catches up to it. Every dollar of leverage against every illiquid tokenized asset needs this infrastructure to exist. Right now, it does not.