Data Submission

Submitters produce the raw material that trains the network’s foundation model. Each epoch, the network generates targets — points in embedding space that represent what data the network needs. Submitters compete to submit data closest to these targets, earning rewards when they win.

How It Works

1. Find a target — Each epoch, validators generate new targets. Open targets are waiting for submissions.
2. Score your data — Download model weights, run them on your data locally, and compute the embedding distance to the target.
3. Submit — Post your scored submission on-chain with a bond as a security deposit.
4. Win and claim — The submission closest to each target wins. After a challenge window, the winner claims their reward.

Rewards are split evenly between the submitter and the model whose embeddings were used. Bonds are returned to honest participants; forfeited bonds from fraudulent submissions are recycled into the rewards pool.

For the full theory, see Data Submission and Targets.

Guides

Submit Data End-to-end workflow from finding a target to claiming rewards.