Innermost Loop Kb

Benchmark

BenchBench

A benchmark that scores a model on whether it can author a benchmark package that strong solver models cannot simply clear.

open-source

Sources

Attribution and reporting linked by this record; links do not imply independent verification.

BenchBench inverts the usual arrangement: the model under test is the benchmark’s author, supplying public solver evidence, private gold answers, a generator, a verifier, a scorer and an account of likely failures, and it wins only if a panel of strong solvers cannot clear the package. It enters this corpus through the May 26 item, where GPT-5.2 led as top creator, and sits beside PostTrainBench as a benchmark built around models doing the measurement work themselves.

Visit resource ↗

Development timeline

Newest reporting dates first. Reporting dates may differ from event dates.

    Knowledge graphopen in browser →

    Published by

    Cite or correct this record

    Suggest a correction — opens a GitHub draft for you to review and submit.