Developer PreviewAudit trail
Trials
A leaderboard number should eventually trace to a model identity, case version, Tutor response, rubric evidence, and sanitized operational metrics.
No public trials available yet.
No public model trials available yet.
Trial detail pages are reserved for public result artifacts; this website never calls a Judge from the browser.Traceability contract
Leaderboard→Model→Trial→Tutor response→Rubric evidence
Future trial fields
ModelModelVersionDatasetVersionGenerationSpecIdGenerationSpecVersionPromptVersionPromptSha256CaseVersionRunIndexTutorResponseCorrectnessDiagnosisGuidanceAdaptationActionabilityRubricResultsCriticalFailuresAnswerLeakageJudgeJudgePromptVersionTokensLatencyCost