> For the complete documentation index, see [llms.txt](https://docs.arcv.network/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.arcv.network/0.-quickstart-guides/0.1-how-to-annotate.md). # 0.1 How to Annotate **Goal:** Learn the contributor workflow for evaluating AI responses and earning native USDC through funded, verified tasks. **Current implementation:** The repository's workbench demonstrates task submission, a four-stage audit animation, and session earnings. These are simulations, not evidence of paid on-chain work. A displayed balance or success banner is not a settlement receipt. Use the walkthrough to learn the interface; confirm a funded production campaign and its payout terms before treating participation as paid work. ## 1. Configure your wallet Open your wallet's network settings and add Arc using the following values: | Setting | Value | | --------------- | ---------------------------- | | Network | Arc | | Chain ID | 5042 | | RPC | | | Native currency | USDC | | Explorer | | Save and switch networks. These settings follow [Arc's official connection guide](https://docs.arc.io/arc/references/connect-to-arc); use the documented mainnet RPC rather than assuming `https://rpc.arc.io` is equivalent. Keep native USDC available for transactions you initiate. Receiving a payout does not itself require you to send a gas transaction. ## 2. Open the workbench Visit [the contributor portal](https://arcv.network/annotate). Open the wallet/sign-in control and inspect the account shown. The prototype sign-in and session balance are not proof of a connected payout wallet or an embedded wallet with custody. For a production task, confirm the actual wallet address, network, funded bounty, and gross/net reward terms. Use **Expert Mode** for detailed evaluation or **Rapid Mode** for a single overall rating. Review the bounty details and task guidelines before starting. Use **Flag Issue** for defective prompts or **Skip Task** when you cannot evaluate the task reliably. ## 3. Compare the responses Read the prompt and all supplied context, then inspect Response A and Response B side by side. Choose the response that best satisfies the task. For code, check runtime, invalid inputs, and correctness rather than merely preferring shorter code. Do not select an answer solely because an AI helper recommends it. ## 4. Rate quality and explain the decision In the current Expert interface, select scores from 1 to 5 for **Algorithmic Runtime & Efficiency**, **Edge-Case & Error Handling**, and **Syntax & PEP 8 Standards**. Enter a technical justification of at least 15 characters; use rationale chips only when they apply. Rapid Mode instead provides an **Overall Quality Rating** slider. For general-language campaigns, use these four assessment dimensions when the campaign rubric supports them: | Dimension | Check | | --------------------- | --------------------------------------------------------------------- | | Factuality | Are claims correct and grounded in supplied evidence? | | Reasoning rigor | Does the visible explanation support the answer without logical gaps? | | Instruction adherence | Are required constraints and requested outputs satisfied? | | Conciseness | Is the answer clear without repetitive or irrelevant content? | Four general-purpose sliders are not currently implemented. Follow the actual task controls and [detailed rubric guidance](/2.-contributor-and-annotator-guide/2.2-rubrics-and-golden-patch.md). ## 5. Optionally use your AI Copilot Open **Connect AI Copilot (BYO Key)**, select a provider, endpoint, and model, then test and save the configuration. Use critique assistance to inspect errors or draft a rationale; review and correct every generated suggestion before submitting. Credentials persist in browser localStorage until removed, and requests go to the selected provider endpoint. Local storage is not encrypted secret custody. Provider retention rules apply, and sending proprietary task content requires campaign permission. A model's opinion is not executed code or independently verified source evidence. See [Copilot setup and privacy](/2.-contributor-and-annotator-guide/2.3-byo-key-copilot.md). ## 6. Optionally submit a Golden Patch In Expert Mode, select a response and enable the **Golden Patch** option. Correct the answer while preserving the prompt's intent. Supply a complete correction rather than an explanation of changes alone. The current editor requires at least 15 characters and does not execute the code. The specified bonus is **+0.50 USDC per accepted patch**, subject to a funded campaign's terms. The prototype simulates this addition; it does not establish an earned payment. ## 7. Submit and verify the result Click **Submit Decision**, the current label for submitting an evaluation. The centered modal demonstrates structural checks, interaction telemetry, judge review, and settlement before advancing the queue. Do not interpret its programmed pass messages as actual audit evidence. For a production payout, verify a successful Arc transaction, the correct contributor address, and the native USDC amount against the campaign reward. If only the session counter changes, no on-chain payout has been established. Continue with the [complete workbench guide](/2.-contributor-and-annotator-guide/2.1-workbench-overview.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.arcv.network/0.-quickstart-guides/0.1-how-to-annotate.md?ask=&goal= ``` `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.