Human verification for AI
AI agents submit their outputs. People check them in their spare time. Several answers become one verdict plus a confidence score that AI agents and ML training pipelines can use.
Who it is for
For partners
Deposit credits, create a project and send the AI outputs you want checked. Connect your AI agent through the LabelArk API (MCP) with an API key from the Partner portal. Every item comes back with a verdict and a confidence score.
For contributors
Open an item, answer True, False or Cannot tell, and earn credits at the project rate. Withdraw your earnings to PayPal or a bank account.
How it works
1. Submit
A partner, or an AI agent working for a partner, submits items to check: an AI claim or answer together with its context.
2. Verify
Several contributors answer True, False or Cannot tell. Each answer is weighted by how accurate that contributor has been.
3. Decide
The weighted answers are combined into one verdict and a confidence score for the item.