{"schemaVersion":"ready-scan.discovery-surface.v1","id":"urn:readyscan:corpus:buyer-question:appen-q60741122-answer-bc2bce56f63e404203f7","key":"buyer_question:appen-q60741122-answer-bc2bce56f63e404203f7","kind":"buyer_question","slug":"appen-q60741122-answer-bc2bce56f63e404203f7","canonicalUrl":"https://directory.readyscan.ai/corpus/buyer-question/appen-q60741122-answer-bc2bce56f63e404203f7","machineUrl":"https://directory.readyscan.ai/discovery/corpus/buyer-question/appen-q60741122-answer-bc2bce56f63e404203f7.json","datePublished":"2026-10-19","dateModified":"2026-10-19","materialFingerprint":"fnv1a64:314b2366129c1248","membershipFingerprint":"fnv1a64:97d6947e754d54ab","relatedEntitySlugs":["appen-q60741122"],"relatedSurfaceKeys":["buyer_question:appen-q60741122-answer-873b4812edb2bbb37331","buyer_question:appen-q60741122-answer-8cae7a262f2e9bcdd3c7","buyer_question:appen-q60741122-answer-a20f702e31f0b8d39a9b","evidence:appen-q60741122"],"sourceIds":["ready-scan-subjects"],"provenance":[{"snapshotId":"appen-q60741122-content-7701ea131d19ac5f-snapshot","sourceId":"ready-scan-subjects","sourceRecordId":"Q60741122","sourceRevisionId":"Q60741122:content:4d8151698aa6e6c6","sourceContentHash":"fnv1a64:e48fc5e1ca301da6","recordRevisionId":"Q60741122:content:4d8151698aa6e6c6","recordContentHash":"fnv1a64:e65b383053c987aa","sourceCanonicalUrl":"https://www.appen.com/","sourceModifiedAt":"2026-10-19T12:16:11.655Z","observedOn":"2026-10-19","permission":{"basis":"public_information","evidenceUrl":"https://www.appen.com/"},"attributionUrl":"https://www.appen.com/"}],"quality":{"method":"same_kind_unique_trigram_floor","minimumUniqueTrigramRatio":0.3,"status":"passed"},"title":"How do we validate LLM-as-a-judge accuracy?","summary":"Cited source answers across 1 active corpus record.","question":"How do we validate LLM-as-a-judge accuracy?","answers":[{"entitySlug":"appen-q60741122","entityName":"Appen","answer":"“We calibrate and validate every LLM judge against human-rater ground truth before it is deployed at scale, measuring inter-annotator agreement with Krippendorff's alpha and Cohen's kappa. Agreement scores and audit trails are documented for every deployment, and human validation is retained wherever the stakes are highest.”","sourceUrls":["https://www.appen.com/ai-research-services"]}]}