SAFE NEAR / benchmark
FORMAL VERIFICATION · SUBMISSION REGISTRY

Proofs, with provenance.

Every submission. Every outcome. Independently checked against a frozen benchmark, with a signed record of the result.

Loading submission records…Published from fv01 · refreshes automatically
Total submissions—All attempts, including rechecks
Accepted proofs —Independent verification passed
Tasks accepted—Across recorded benchmark revisions
Benchmark revisions—Compare results within a revision

Submission registry 0

Explore verification results and the metadata behind each run.

↓   Download JSON
Benchmark submission results. Select a task to inspect its receipt and metadata.
Task / submission Outcome Model / harness Reported cost ⓘ Verification Benchmark Submitted UTC Details
UNDERSTANDING THE RECORD

Verification is evidence.
Metadata is context.

A successful proof and a low reported cost answer different questions. We keep that distinction visible.

01

Independent, frozen evaluation

Proofs are checked against an operator-approved benchmark commit and judge image. Acceptance is specific to that revision. Rechecks are separate, permanent records.

02

Signed results, reported costs

Ed25519 receipts record outcomes, provenance, and verification duration. Model settings, generation time, cost, and campaign reports are declared metadata, not attestations of model usage.

03

Public metadata, private proofs

Proof sources and compiler logs remain private. Queued and running entries are provisional snapshots; completed signed receipts establish the final outcome. Infrastructure failures are never counted as proof rejections.

SUBMISSION RECORD