New case

File a benchmark claim

Write the sentence you want judged. Attach evidence. Unequal retries and missing methodology will not survive evaluation.

Claim

Claimant (wallet or name)

Title

Exact claim statement

Models and benchmark

Model A

Model A version

Model B / baseline

Model B version

Benchmark

Benchmark version

Evaluation date

Metric

Reported result

Conditions — the part leaderboards skip

Retries A

Retries B

Temperature A

Temperature B

Tools A

Tools B

Prompt parity

System prompt parity

Sample size

Exclusions

Methodology and sources

Methodology

Source URLs (one per line)

Evidence

Hashes are computed over kind, URI, and note. The contract stores the hash, not the file.

Kind

URI

Note

Publishing writes the case to the public index immediately. Open the case file and choose Record on GenLayer to put hashes on Studionet — that step waits for validator finalization.