Deepchecks tool integration

[thewebchap]

Decentralised Benchmarking

1. Open reviewers run the model against a shared benchmark dataset.
2. Scores (accuracy, F1, etc.) are written on-chain, tied to the model’s ID

Reproducibility Proofs

1. Using content-addressed datasets on Rubix (via IPFS)
2. Validators can re-run training or fine-tuning if reproducibility is required

Reputation-weighted Endorsements
Validators have reputation scores; endorsements from high-reputation validators carry more weight in ranking models.