> For the complete documentation index, see [llms.txt](https://docs.trueopen.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.trueopen.ai/protocol/v1/tasks/task-04-verification.md).

# Verification

Retrieve Task data, independently verify the full output, and close round 1.

Verification begins when the round-1 Verifier set is finalized. Each selected Verifier independently downloads the same committed Task data and evaluates the full output.

```mermaid
sequenceDiagram
    participant V as Selected Verifiers
    participant B as Task Builders
    participant K as Chain

    V->>B: Fetch input, output, and required evidence
    B-->>V: Authorized committed data
    opt Complete verified data cannot be retrieved
        V->>K: Signed data-unavailable report
    end
    V->>V: Verify commitments and full generated output
    V->>B: Upload own evidence bundle
    V->>B: Submit result commitment
    B->>K: Relay commit transaction or batch
    K->>K: Close commit phase when all arrive or deadline expires
    alt Commit quorum is available
        K-->>V: Reveal phase begins
        V->>B: Submit result reveal
        B->>K: Relay result transaction or batch
        K->>K: Recompute each expected result
        K->>K: Apply quorum gate, then value-consensus gate
        K-->>B: Round 1 closed with facts commitment
    else Commit quorum is unavailable
        K->>K: Classify verification or data availability failure
    end
```

## Retrieve and validate

Each selected Verifier authenticates with the fixed Task Builders and retrieves the input, output, Worker evidence, and Profile material required by the Task's evidence schema.

Before model computation, it checks hashes, sizes, both Worker evidence commitments, Profile version, tokenizer, generation parameters, and evidence types. Material for another Task, Profile version, or round is invalid even if it is otherwise well formed.

If no Builder can provide a complete valid set, a selected Verifier may directly submit a signed unavailability report before the commit deadline. One report or a service log is not enough to assign responsibility; the chain applies the defined threshold and Builder coverage rules.

## Compute the result

For text-generation verification, the Verifier evaluates every committed generated token position using the frozen prefill or teacher-forcing procedure. It does not sample random positions or rerun unconstrained autoregressive generation.

The Verifier produces per-token comparison leaves, a typed metric summary, an aggregate proof, and an evidence-bundle commitment. It stores its own signed material before relay and uploads its evidence bundle to the Task Builders.

## Commit phase

The Verifier first submits a commitment to its complete result payload and private salt. This hides the decision while other selected Verifiers are still computing.

The reveal phase may begin early only when every selected Verifier has committed. A threshold-sized subset is not enough for the fast path because an uncommitted Verifier could otherwise observe revealed answers and copy them. If not all commits arrive, the protocol waits until the commit deadline and proceeds only if the required quorum exists.

## Reveal and two gates

After reveal begins, each Verifier submits the result payload and salt. The chain rejects a reveal that does not match the commitment. It derives the sample verdict and work-unit count from the accepted metric summary under the frozen Profile rules; the Verifier does not submit either value as its own verdict.

Closing the round uses two gates:

1. **Quorum gate:** enough valid selected Verifier results must exist.
2. **Value-consensus gate:** a sufficiently large cluster must agree on the same verdict and work-unit count.

A bare `PASS` or `REJECT` is insufficient. It must be supported by a valid committed result and its required metric and evidence commitments.

## Close round 1

Round 1 closes when all valid reveals arrive or the reveal deadline expires. The chain records the outcome, verdict, participant facts, and immutable round-facts commitment.

Closing round 1 moves no funds. The verdict remains challengeable until the challenge window closes, and no service earnings become claimable.

See [Verification](/protocol/v1/tasks/verification.md) and [Challenges and settlement](/protocol/v1/tasks/challenges-and-settlement.md).


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.trueopen.ai/protocol/v1/tasks/task-04-verification.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.