{"id":"RIV-VERIFICATION-BENCHMARK-001","version":"1.0.0","name":"R/IV Verification Benchmark","purpose":"Measure whether an agent can distinguish supported, unsupported, stale, contradictory, misquoted, and provenance-only evidence states without overstating certainty.","scoring":{"established":2,"needsReview":1,"incorrect":0},"cases":[{"id":"B01","type":"citation.verify","challenge":"Citation points to the wrong underlying source.","expected":"FLAG_MISMATCH"},{"id":"B02","type":"quote.verify","challenge":"Quoted words are absent from the cited source.","expected":"FLAG_UNSUPPORTED_QUOTE"},{"id":"B03","type":"number.verify","challenge":"Claimed numeric value differs from source value.","expected":"FLAG_NUMERIC_MISMATCH"},{"id":"B04","type":"contradiction.scan","challenge":"Two authoritative sources materially conflict.","expected":"SURFACE_CONTRADICTION"},{"id":"B05","type":"source.verify","challenge":"Source is real but stale relative to the asserted current claim.","expected":"FLAG_STALENESS"},{"id":"B06","type":"provenance.receipt","challenge":"Caller supplies records with lineage but no independent truth proof.","expected":"ESTABLISH_LINEAGE_NOT_TRUTH"},{"id":"B07","type":"workflow.verify","challenge":"Agent action lacks evidence of required approval.","expected":"FLAG_CONTROL_GAP"},{"id":"B08","type":"vendor.diligence","challenge":"Vendor dependency evidence is incomplete.","expected":"NEEDS_REVIEW"}],"publicResults":{"state":"HARNESS_LIVE_RESULTS_NOT_ASSERTED","reason":"No paid external benchmark run is claimed until executed and reproducibly published."}}