- Hackathons
- Feed
- Builders
- Organizations
- Learn
Onchain verifiable ML performance metrics for NFT valuations using merkle Proofs
On chain proof that whenever an NFT sells, I predicted X price before it happened, therefore I get an error rate Y on that sale without disclosing the rest of the predictions
- 0 Raised
- 436 Views
- 0 Judges
Team
Categories
Description
Selective ML-NFT Valuation Merkle Proofs by PunkPredictor
A Python toolkit and on-chain smart contract for building, storing, and verifying sparse Merkle-tree proofs over machine-learning–generated NFT valuation predictions to prove predictive accuracy metrics trustlessly.
📦 Features
- Create Merkle leaves from a DataFrame of ML-predicted NFT valuations
- Store leaves, salts, predictions & roots in a local SQLite database
- On-demand proof generation (O(log N) siblings) without keeping the full tree in memory
- Publish Merkle root on Base Sepolia via a `PunkPredictor` Solidity contract along with a timestamp as a state variable
- Verify leaf inclusion both off-chain (Python) and on-chain
The end result is that one can prove that they predicted that NFT with ID X would sell for Y before it actually sold, and therefore if it sells shorting after, it has a verifiable prediction error of Z. This is done without having to reveal the valuations of all other 10,000 NFTs