All benchmarks
Public evaluations
All recorded runs
Open a run to see its cases, results, and measurement limits.
Python detectionPortable defaults and explicit POSIX sensitivity on the OWASP suiteDevelopment rerunModel-assisted detectionDid the model add coverage on the same 28 cases?Recorded pilotPaired model evaluationBaseline and candidate outputs on matched cases, plus a fresh sampleDevelopment evaluationFull paired model runCase outcomes, verification replay and budget historyDevelopment evaluationModel gap evaluationEarlier development cases and a fresh 28-case evaluationDevelopment evaluationModel follow-upBefore and after results, including untouched casesDevelopment evaluationRepair behaviorHow two generated patches behaved in controlled testsAuthored regressionRepository detection regressionsPreviously missed cases on the original repository revisionsDevelopment rerunRepository outcome testsFirst-pass repository results and the earlier external rerunRecorded resultsOWASP development evaluationSupported CWE results, original baseline and exclusionsDevelopment rerunMaintainer-authored rule fixturesPer-rule scores, family totals and exact inputsFixture benchmark