Documentation
How BenchProof is put together
BenchProof is a verification layer for AI benchmark claims. The Intelligent Contract is the source of truth for lifecycle and verdicts. The application index stores evidence metadata and transaction state so a human can read a case without decoding calldata.
Architecture
- A claimant publishes a structured claim plus evidence hashes.
- Reviewers file categorized challenges.
request_evaluationruns a GenLayer non-comparative LLM consensus over untrusted data.- The verdict is stored on-chain and mirrored in the case file.
Intelligent Contract
Python contract contracts/benchproof.py. States: DRAFT → OPEN → CHALLENGED → EVALUATING → FINALIZED. Evidence freezes at publish. Challenges stop when evaluation begins. Finalized claims cannot reopen.
- Network
- studionet (61999)
- RPC
- https://studio.genlayer.com/api
- Contract
- 0x0E6A637e78241D5005245281Bc92d7C3aA09e197