One named evaluator
The server runs evaluation and authorizes payout. Signatures identify that evaluator. They do not prove that training occurred or make verification decentralized.
An assurance market for AI capability improvements.
Define what better means, let independent workers compete, and inspect the evidence before settlement.
Solana commits the economic terms. A named evaluator judges model quality.
GradientMine keeps those trust boundaries separate instead of pretending the chain can verify machine learning.
Only real run evidence will appear here.
No GPU purchase needed for the included digits task. A CPU trains a genuine low-rank update to a small neural network.
Python 3.11+ · Works locally · No custom token
Keep the identity file on your own machine. Never paste it into a website or chat.
Commands adapt to the selected bounty. In Devnet mode, use a dedicated, funded test identity and verify the program ID independently.
The server runs evaluation and authorizes payout. Signatures identify that evaluator. They do not prove that training occurred or make verification decentralized.
The included UCI digits split is reproducible public data. A held-out API boundary is not secrecy or protection against benchmark gaming. Reported accuracy is not general intelligence.
Solana Devnet SOL has no monetary value. Eligible does not mean paid. A payout is shown only after finalized transaction evidence. Escrow rent is not recovered by the timeout refund.
At the cutoff, the evaluator compares all registered candidates against the same frozen parent and held-out split. The measured improvement must reach the bounty threshold and the one-sided, multiple-candidate-adjusted paired-bootstrap lower bound must be positive. Eligible candidates are ranked by accuracy, then artifact hash. The statistical procedure is approximate under IID assumptions; it does not eliminate leakage or selection bias across repeated rounds.