{"case_id": "pair_controlled_conflict_heavy_000", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_000_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_000_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "pair_controlled_conflict_heavy_000_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_000_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_000_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "pair_controlled_conflict_heavy_000_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_001", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_001_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_001_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "pair_controlled_conflict_heavy_001_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "pair_controlled_conflict_heavy_001_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_001_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "pair_controlled_conflict_heavy_001_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_002", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_002_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_002_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "pair_controlled_conflict_heavy_002_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_002_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_002_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 5.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "pair_controlled_conflict_heavy_002_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_003", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_003_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_003_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "pair_controlled_conflict_heavy_003_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "pair_controlled_conflict_heavy_003_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_003_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 6.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "pair_controlled_conflict_heavy_003_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_004", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_004_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_004_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "pair_controlled_conflict_heavy_004_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_004_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_004_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "pair_controlled_conflict_heavy_004_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_005", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_005_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_005_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "pair_controlled_conflict_heavy_005_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_005_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_005_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "pair_controlled_conflict_heavy_005_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_006", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_006_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_006_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "pair_controlled_conflict_heavy_006_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "pair_controlled_conflict_heavy_006_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_006_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 5.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "pair_controlled_conflict_heavy_006_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_007", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_007_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_007_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "pair_controlled_conflict_heavy_007_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_007_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_007_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 6.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "pair_controlled_conflict_heavy_007_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_008", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_008_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_008_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "pair_controlled_conflict_heavy_008_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "pair_controlled_conflict_heavy_008_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_008_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "pair_controlled_conflict_heavy_008_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_009", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_009_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_009_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "pair_controlled_conflict_heavy_009_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_009_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_009_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "pair_controlled_conflict_heavy_009_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_010", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_010_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_010_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "pair_controlled_conflict_heavy_010_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_010_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_010_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 5.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "pair_controlled_conflict_heavy_010_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_011", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_011_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_011_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "pair_controlled_conflict_heavy_011_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "pair_controlled_conflict_heavy_011_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_011_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 6.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "pair_controlled_conflict_heavy_011_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_012", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_012_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_012_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "pair_controlled_conflict_heavy_012_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_012_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_012_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "pair_controlled_conflict_heavy_012_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_013", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_013_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_013_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "pair_controlled_conflict_heavy_013_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "pair_controlled_conflict_heavy_013_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_013_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "pair_controlled_conflict_heavy_013_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_014", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_014_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_014_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "pair_controlled_conflict_heavy_014_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_014_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_014_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 5.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "pair_controlled_conflict_heavy_014_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_015", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_015_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_015_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "pair_controlled_conflict_heavy_015_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_015_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_015_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 6.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "pair_controlled_conflict_heavy_015_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_016", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_016_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_016_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "pair_controlled_conflict_heavy_016_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "pair_controlled_conflict_heavy_016_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_016_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "pair_controlled_conflict_heavy_016_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_017", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_017_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_017_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "pair_controlled_conflict_heavy_017_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_017_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_017_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "pair_controlled_conflict_heavy_017_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_018", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_018_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_018_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "pair_controlled_conflict_heavy_018_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "pair_controlled_conflict_heavy_018_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_018_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 5.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "pair_controlled_conflict_heavy_018_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_019", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_019_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_019_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "pair_controlled_conflict_heavy_019_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_019_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_019_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 6.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "pair_controlled_conflict_heavy_019_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_020", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_020_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_020_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "pair_controlled_conflict_heavy_020_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_020_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_020_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "pair_controlled_conflict_heavy_020_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_021", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_021_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_021_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "pair_controlled_conflict_heavy_021_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "pair_controlled_conflict_heavy_021_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_021_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "pair_controlled_conflict_heavy_021_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_022", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_022_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_022_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "pair_controlled_conflict_heavy_022_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_022_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_022_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 5.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "pair_controlled_conflict_heavy_022_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_023", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_023_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_023_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "pair_controlled_conflict_heavy_023_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "pair_controlled_conflict_heavy_023_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_023_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 6.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "pair_controlled_conflict_heavy_023_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_024", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_024_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_024_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "pair_controlled_conflict_heavy_024_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_024_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_024_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "pair_controlled_conflict_heavy_024_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_025", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_025_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_025_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "pair_controlled_conflict_heavy_025_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_025_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_025_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "pair_controlled_conflict_heavy_025_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_026", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_026_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_026_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "pair_controlled_conflict_heavy_026_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "pair_controlled_conflict_heavy_026_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_026_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 5.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "pair_controlled_conflict_heavy_026_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_027", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_027_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_027_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "pair_controlled_conflict_heavy_027_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_027_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_027_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 6.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "pair_controlled_conflict_heavy_027_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_028", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_028_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_028_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "pair_controlled_conflict_heavy_028_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "pair_controlled_conflict_heavy_028_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_028_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "pair_controlled_conflict_heavy_028_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_029", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_029_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_029_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "pair_controlled_conflict_heavy_029_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_029_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_029_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "pair_controlled_conflict_heavy_029_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_030", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_030_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_030_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "pair_controlled_conflict_heavy_030_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_030_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_030_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 5.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "pair_controlled_conflict_heavy_030_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_031", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_031_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_031_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "pair_controlled_conflict_heavy_031_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "pair_controlled_conflict_heavy_031_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_031_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 6.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "pair_controlled_conflict_heavy_031_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_032", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_032_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_032_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "pair_controlled_conflict_heavy_032_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_032_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_032_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "pair_controlled_conflict_heavy_032_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_033", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_033_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_033_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "pair_controlled_conflict_heavy_033_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "pair_controlled_conflict_heavy_033_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_033_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "pair_controlled_conflict_heavy_033_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_034", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_034_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_034_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "pair_controlled_conflict_heavy_034_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_034_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_034_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 5.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "pair_controlled_conflict_heavy_034_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_035", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_035_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_035_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "pair_controlled_conflict_heavy_035_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_035_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_035_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 6.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "pair_controlled_conflict_heavy_035_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_036", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_036_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_036_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "pair_controlled_conflict_heavy_036_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "pair_controlled_conflict_heavy_036_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_036_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "pair_controlled_conflict_heavy_036_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_037", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_037_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_037_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "pair_controlled_conflict_heavy_037_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_037_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_037_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "pair_controlled_conflict_heavy_037_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_038", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_038_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_038_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "pair_controlled_conflict_heavy_038_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "pair_controlled_conflict_heavy_038_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_038_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 5.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "pair_controlled_conflict_heavy_038_trap_0"}]}
{"case_id": "pair_controlled_conflict_heavy_039", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_conflict_heavy_039_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_conflict_heavy_039_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "pair_controlled_conflict_heavy_039_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "pair_controlled_conflict_heavy_039_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_conflict_heavy_039_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 6.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "pair_controlled_conflict_heavy_039_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_000", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_000_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_000_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_000_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_000_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_000_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_001", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_001_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_001_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_001_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_001_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_001_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_002", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_002_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_002_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_002_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_002_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_002_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_003", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_003_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_003_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_003_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_003_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_003_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_004", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_004_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_004_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_004_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_004_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_004_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_005", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_005_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_005_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_005_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_005_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_005_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_006", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_006_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_006_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_006_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_006_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_006_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_007", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_007_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_007_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_007_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_007_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_007_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_008", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_008_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_008_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_008_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_008_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_008_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_009", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_009_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_009_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_009_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_009_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_009_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_010", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_010_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_010_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_010_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_010_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_010_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_011", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_011_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_011_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_011_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_011_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_011_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_012", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_012_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_012_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_012_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_012_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_012_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_013", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_013_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_013_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_013_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_013_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_013_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_014", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_014_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_014_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_014_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_014_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_014_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_015", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_015_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_015_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_015_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_015_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_015_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_016", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_016_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_016_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_016_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_016_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_016_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_017", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_017_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_017_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_017_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_017_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_017_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_018", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_018_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_018_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_018_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_018_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_018_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_019", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_019_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_019_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_019_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_019_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_019_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_020", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_020_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_020_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_020_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_020_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_020_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_021", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_021_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_021_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_021_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_021_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_021_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_022", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_022_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_022_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_022_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_022_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_022_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_023", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_023_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_023_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_023_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_023_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_023_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_024", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_024_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_024_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_024_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_024_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_024_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_025", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_025_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_025_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_025_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_025_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_025_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_026", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_026_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_026_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_026_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_026_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_026_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_027", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_027_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_027_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_027_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_027_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_027_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_028", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_028_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_028_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_028_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_028_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_028_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_029", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_029_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_029_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_029_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_029_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_029_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_030", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_030_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_030_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_030_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_030_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_030_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_031", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_031_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_031_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_031_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_031_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_031_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_032", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_032_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_032_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_032_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_032_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_032_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_033", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_033_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_033_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_033_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_033_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_033_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_034", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_034_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_034_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_034_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_034_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_034_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_035", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_035_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_035_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_035_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_035_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_035_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_036", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_036_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_036_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_036_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_036_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_036_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_037", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_037_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_037_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_037_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_037_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_037_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_038", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_038_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_038_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_038_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_038_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_038_trap_0"}]}
{"case_id": "pair_controlled_provenance_heavy_039", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_provenance_heavy_039_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_controlled_provenance_heavy_039_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_controlled_provenance_heavy_039_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_provenance_heavy_039_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_controlled_provenance_heavy_039_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_000", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_000_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_000_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_000_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_000_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_000_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_001", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_001_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_001_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_001_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_001_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_001_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_002", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_002_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_002_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_002_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_002_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_002_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_003", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_003_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_003_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_003_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_003_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_003_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_004", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_004_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_004_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_004_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_004_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_004_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_005", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_005_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_005_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_005_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_005_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_005_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_006", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_006_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_006_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_006_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_006_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_006_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_007", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_007_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_007_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_007_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_007_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_007_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_008", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_008_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_008_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_008_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_008_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_008_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_009", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_009_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_009_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_009_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_009_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_009_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_010", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_010_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_010_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_010_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_010_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_010_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_011", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_011_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_011_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_011_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_011_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_011_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_012", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_012_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_012_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_012_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_012_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_012_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_013", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_013_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_013_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_013_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_013_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_013_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_014", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_014_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_014_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_014_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_014_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_014_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_015", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_015_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_015_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_015_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_015_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_015_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_016", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_016_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_016_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_016_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_016_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_016_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_017", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_017_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_017_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_017_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_017_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_017_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_018", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_018_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_018_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_018_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_018_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_018_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_019", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_019_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_019_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_019_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_019_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_019_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_020", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_020_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_020_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_020_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_020_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_020_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_021", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_021_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_021_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_021_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_021_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_021_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_022", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_022_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_022_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_022_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_022_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_022_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_023", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_023_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_023_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_023_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_023_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_023_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_024", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_024_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_024_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_024_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_024_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_024_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_025", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_025_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_025_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_025_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_025_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_025_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_026", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_026_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_026_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_026_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_026_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_026_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_027", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_027_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_027_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_027_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_027_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_027_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_028", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_028_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_028_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_028_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_028_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_028_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_029", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_029_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_029_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_029_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_029_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_029_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_030", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_030_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_030_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_030_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_030_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_030_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_031", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_031_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_031_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_031_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_031_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_031_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_032", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_032_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_032_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_032_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_032_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_032_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_033", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_033_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_033_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_033_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_033_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_033_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_034", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_034_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_034_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_034_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_034_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_034_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_035", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_035_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_035_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_035_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_035_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_035_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_036", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_036_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_036_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_036_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_036_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_036_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_037", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_037_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_037_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_037_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_037_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_037_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_038", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_038_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_038_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_038_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_038_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_038_trap_0"}]}
{"case_id": "pair_controlled_caveat_heavy_039", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_caveat_heavy_039_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_controlled_caveat_heavy_039_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_controlled_caveat_heavy_039_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_caveat_heavy_039_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_controlled_caveat_heavy_039_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_000", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_000_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_000_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_000_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_000_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_000_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_001", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_001_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_001_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_001_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_001_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_001_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_002", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_002_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_002_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_002_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_002_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_002_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_003", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_003_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_003_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_003_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_003_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_003_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_004", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_004_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_004_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_004_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_004_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_004_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_005", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_005_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_005_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_005_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_005_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_005_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_006", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_006_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_006_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_006_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_006_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_006_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_007", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_007_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_007_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_007_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_007_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_007_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_008", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_008_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_008_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_008_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_008_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_008_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_009", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_009_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_009_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_009_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_009_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_009_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_010", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_010_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_010_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_010_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_010_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_010_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_011", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_011_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_011_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_011_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_011_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_011_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_012", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_012_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_012_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_012_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_012_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_012_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_013", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_013_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_013_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_013_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_013_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_013_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_014", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_014_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_014_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_014_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_014_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_014_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_015", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_015_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_015_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_015_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_015_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_015_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_016", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_016_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_016_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_016_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_016_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_016_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_017", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_017_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_017_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_017_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_017_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_017_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_018", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_018_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_018_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_018_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_018_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_018_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_019", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_019_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_019_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_019_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_019_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_019_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_020", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_020_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_020_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_020_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_020_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_020_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_021", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_021_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_021_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_021_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_021_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_021_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_022", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_022_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_022_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_022_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_022_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_022_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_023", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_023_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_023_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_023_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_023_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_023_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_024", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_024_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_024_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_024_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_024_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_024_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_025", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_025_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_025_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_025_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_025_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_025_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_026", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_026_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_026_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_026_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_026_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_026_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_027", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_027_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_027_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_027_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_027_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_027_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_028", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_028_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_028_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_028_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_028_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_028_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_029", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_029_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_029_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_029_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_029_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_029_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_030", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_030_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_030_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_030_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_030_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_030_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_031", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_031_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_031_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_031_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_031_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_031_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_032", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_032_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_032_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_032_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_032_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_032_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_033", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_033_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_033_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_033_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_033_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_033_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_034", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_034_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_034_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_034_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_034_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_034_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_035", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_035_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_035_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_035_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_035_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_035_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_036", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_036_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_036_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_036_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_036_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_036_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_037", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_037_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_037_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_037_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_037_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_037_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_038", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_038_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_038_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_038_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_038_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_038_trap_0"}]}
{"case_id": "pair_controlled_unsupported_trap_039", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_unsupported_trap_039_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_controlled_unsupported_trap_039_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_controlled_unsupported_trap_039_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_controlled_unsupported_trap_039_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_controlled_unsupported_trap_039_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_000", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_000_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_000_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_000_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_000_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_000_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_001", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_001_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_001_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_001_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_001_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_001_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_002", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_002_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_002_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_002_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_002_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_002_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_003", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_003_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_003_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_003_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_003_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_003_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_004", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_004_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_004_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_004_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_004_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_004_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_005", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_005_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_005_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_005_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_005_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_005_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_006", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_006_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_006_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_006_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_006_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_006_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_007", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_007_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_007_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_007_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_007_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_007_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_008", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_008_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_008_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_008_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_008_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_008_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_009", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_009_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_009_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_009_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_009_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_009_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_010", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_010_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_010_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_010_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_010_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_010_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_011", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_011_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_011_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_011_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_011_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_011_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_012", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_012_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_012_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_012_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_012_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_012_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_013", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_013_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_013_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_013_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_013_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_013_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_014", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_014_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_014_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_014_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_014_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_014_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_015", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_015_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_015_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_015_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_015_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_015_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_016", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_016_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_016_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_016_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_016_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_016_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_017", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_017_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_017_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_017_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_017_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_017_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_018", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_018_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_018_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_018_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_018_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_018_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_019", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_019_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_019_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_019_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_019_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_019_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_020", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_020_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_020_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_020_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_020_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_020_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_021", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_021_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_021_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_021_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_021_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_021_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_022", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_022_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_022_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_022_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_022_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_022_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_023", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_023_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_023_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_023_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_023_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_023_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_024", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_024_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_024_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_024_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_024_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_024_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_025", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_025_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_025_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_025_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_025_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_025_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_026", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_026_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_026_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_026_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_026_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_026_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_027", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_027_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_027_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_027_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_027_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_027_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_028", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_028_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_028_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_028_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_028_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_028_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_029", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_029_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_029_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_029_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_029_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_029_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_030", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_030_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_030_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_030_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_030_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_030_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_031", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_031_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_031_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_031_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_031_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_031_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_032", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_032_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_032_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_032_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_032_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_032_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_033", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_033_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_033_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_033_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_033_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_033_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_034", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_034_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_034_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_034_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_034_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_034_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_035", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_035_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_035_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_035_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_035_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_035_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_036", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_036_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_036_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_036_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_036_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_036_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_037", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_037_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_037_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_037_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_037_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_037_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_038", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_038_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_038_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_038_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_038_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_038_trap_0"}]}
{"case_id": "pair_controlled_no_conflict_complementary_039", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_no_conflict_complementary_039_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_controlled_no_conflict_complementary_039_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_controlled_no_conflict_complementary_039_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_controlled_no_conflict_complementary_039_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_controlled_no_conflict_complementary_039_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_000", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_000_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_000_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_000_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_000_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_000_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_000_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_001", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_001_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_001_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_001_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_001_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_001_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_001_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_002", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_002_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_002_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_002_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_002_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_002_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_002_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_003", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_003_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_003_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_003_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_003_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_003_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_003_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_004", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_004_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_004_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_004_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_004_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_004_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_004_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_005", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_005_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_005_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_005_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_005_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_005_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_005_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_006", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_006_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_006_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_006_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_006_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_006_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_006_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_007", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_007_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_007_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_007_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_007_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_007_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_007_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_008", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_008_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_008_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_008_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_008_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_008_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_008_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_009", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_009_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_009_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_009_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_009_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_009_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_009_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_010", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_010_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_010_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_010_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_010_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_010_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_010_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_011", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_011_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_011_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_011_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_011_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_011_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_011_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_012", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_012_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_012_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_012_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_012_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_012_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_012_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_013", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_013_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_013_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_013_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_013_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_013_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_013_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_014", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_014_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_014_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_014_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_014_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_014_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_014_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_015", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_015_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_015_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_015_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_015_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_015_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_015_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_016", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_016_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_016_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_016_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_016_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_016_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_016_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_017", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_017_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_017_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_017_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_017_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_017_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_017_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_018", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_018_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_018_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_018_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_018_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_018_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_018_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_019", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_019_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_019_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_019_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_019_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_019_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_019_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_020", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_020_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_020_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_020_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_020_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_020_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_020_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_021", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_021_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_021_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_021_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_021_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_021_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_021_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_022", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_022_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_022_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_022_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_022_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_022_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_022_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_023", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_023_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_023_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_023_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_023_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_023_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_023_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_024", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_024_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_024_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_024_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_024_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_024_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_024_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_025", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_025_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_025_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_025_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_025_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_025_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_025_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_026", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_026_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_026_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_026_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_026_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_026_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_026_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_027", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_027_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_027_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_027_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_027_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_027_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_027_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_028", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_028_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_028_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_028_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_028_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_028_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_028_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_029", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_029_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_029_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_029_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_029_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_029_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_029_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_030", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_030_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_030_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_030_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_030_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_030_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_030_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_031", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_031_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_031_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_031_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_031_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_031_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_031_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_032", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_032_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_032_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_032_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_032_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_032_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_032_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_033", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_033_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_033_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_033_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_033_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_033_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_033_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_034", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_034_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_034_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_034_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_034_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_034_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_034_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_035", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_035_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_035_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_035_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_035_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_035_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_035_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_036", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_036_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_036_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_036_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_036_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_036_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_036_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_037", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_037_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_037_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_037_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_037_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_037_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_037_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_038", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_038_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_038_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_038_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_038_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_038_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_038_trap_0"}]}
{"case_id": "pair_controlled_preference_ambiguous_039", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_preference_ambiguous_039_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_preference_ambiguous_039_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_controlled_preference_ambiguous_039_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_controlled_preference_ambiguous_039_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_preference_ambiguous_039_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_controlled_preference_ambiguous_039_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_000", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_000_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_000_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_000_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_000_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_000_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_001", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_001_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_001_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_001_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_001_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_001_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_002", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_002_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_002_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_002_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_002_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_002_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_003", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_003_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_003_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_003_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_003_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_003_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_004", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_004_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_004_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_004_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_004_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_004_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_005", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_005_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_005_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_005_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_005_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_005_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_006", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_006_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_006_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_006_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_006_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_006_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_007", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_007_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_007_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_007_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_007_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_007_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_008", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_008_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_008_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_008_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_008_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_008_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_009", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_009_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_009_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_009_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_009_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_009_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_010", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_010_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_010_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_010_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_010_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_010_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_011", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_011_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_011_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_011_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_011_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_011_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_012", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_012_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_012_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_012_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_012_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_012_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_013", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_013_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_013_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_013_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_013_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_013_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_014", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_014_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_014_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_014_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_014_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_014_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_015", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_015_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_015_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_015_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_015_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_015_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_016", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_016_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_016_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_016_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_016_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_016_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_017", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_017_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_017_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_017_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_017_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_017_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_018", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_018_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_018_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_018_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_018_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_018_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_019", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_019_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_019_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_019_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_019_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_019_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_020", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_020_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_020_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_020_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_020_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_020_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_021", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_021_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_021_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_021_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_021_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_021_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_022", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_022_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_022_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_022_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_022_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_022_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_023", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_023_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_023_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_023_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_023_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_023_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_024", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_024_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_024_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_024_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_024_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_024_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_025", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_025_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_025_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_025_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_025_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_025_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_026", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_026_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_026_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_026_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_026_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_026_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_027", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_027_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_027_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_027_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_027_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_027_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_028", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_028_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_028_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_028_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_028_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_028_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_029", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_029_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_029_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_029_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_029_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_029_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_030", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_030_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_030_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_030_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_030_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_030_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_031", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_031_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_031_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_031_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_031_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_031_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_032", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_032_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_032_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_032_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_032_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_032_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_033", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_033_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_033_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_033_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_033_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_033_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_034", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_034_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_034_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_034_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_034_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_034_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_035", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_035_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_035_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_035_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_035_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_035_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_036", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_036_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_036_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_036_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_036_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_036_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_037", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_037_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_037_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_037_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_037_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_037_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_038", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_038_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_038_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_038_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_038_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_038_trap_0"}]}
{"case_id": "pair_controlled_weak_shared_assumption_039", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_weak_shared_assumption_039_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_controlled_weak_shared_assumption_039_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_controlled_weak_shared_assumption_039_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_controlled_weak_shared_assumption_039_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_controlled_weak_shared_assumption_039_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_000", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_000_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_000_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_000_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "pair_controlled_minority_critical_caveat_000_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_000_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_000_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_001", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_001_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_001_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_001_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "pair_controlled_minority_critical_caveat_001_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_001_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_001_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_002", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_002_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_002_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_002_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "pair_controlled_minority_critical_caveat_002_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_002_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_002_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_003", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_003_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_003_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_003_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "pair_controlled_minority_critical_caveat_003_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_003_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_003_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_004", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_004_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_004_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_004_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "pair_controlled_minority_critical_caveat_004_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_004_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_004_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_005", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_005_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_005_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_005_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "pair_controlled_minority_critical_caveat_005_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_005_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_005_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_006", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_006_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_006_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_006_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "pair_controlled_minority_critical_caveat_006_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_006_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_006_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_007", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_007_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_007_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_007_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "pair_controlled_minority_critical_caveat_007_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_007_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_007_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_008", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_008_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_008_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_008_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "pair_controlled_minority_critical_caveat_008_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_008_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_008_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_009", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_009_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_009_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_009_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "pair_controlled_minority_critical_caveat_009_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_009_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_009_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_010", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_010_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_010_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_010_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "pair_controlled_minority_critical_caveat_010_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_010_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_010_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_011", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_011_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_011_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_011_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "pair_controlled_minority_critical_caveat_011_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_011_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_011_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_012", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_012_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_012_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_012_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "pair_controlled_minority_critical_caveat_012_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_012_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_012_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_013", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_013_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_013_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_013_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "pair_controlled_minority_critical_caveat_013_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_013_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_013_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_014", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_014_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_014_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_014_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "pair_controlled_minority_critical_caveat_014_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_014_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_014_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_015", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_015_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_015_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_015_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "pair_controlled_minority_critical_caveat_015_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_015_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_015_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_016", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_016_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_016_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_016_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "pair_controlled_minority_critical_caveat_016_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_016_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_016_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_017", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_017_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_017_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_017_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "pair_controlled_minority_critical_caveat_017_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_017_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_017_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_018", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_018_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_018_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_018_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "pair_controlled_minority_critical_caveat_018_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_018_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_018_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_019", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_019_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_019_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_019_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "pair_controlled_minority_critical_caveat_019_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_019_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_019_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_020", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_020_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_020_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_020_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "pair_controlled_minority_critical_caveat_020_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_020_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_020_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_021", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_021_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_021_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_021_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "pair_controlled_minority_critical_caveat_021_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_021_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_021_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_022", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_022_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_022_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_022_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "pair_controlled_minority_critical_caveat_022_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_022_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_022_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_023", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_023_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_023_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_023_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "pair_controlled_minority_critical_caveat_023_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_023_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_023_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_024", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_024_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_024_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_024_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "pair_controlled_minority_critical_caveat_024_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_024_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_024_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_025", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_025_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_025_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_025_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "pair_controlled_minority_critical_caveat_025_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_025_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_025_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_026", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_026_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_026_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_026_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "pair_controlled_minority_critical_caveat_026_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_026_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_026_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_027", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_027_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_027_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_027_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "pair_controlled_minority_critical_caveat_027_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_027_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_027_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_028", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_028_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_028_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_028_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "pair_controlled_minority_critical_caveat_028_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_028_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_028_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_029", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_029_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_029_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_029_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "pair_controlled_minority_critical_caveat_029_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_029_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_029_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_030", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_030_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_030_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_030_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "pair_controlled_minority_critical_caveat_030_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_030_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_030_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_031", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_031_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_031_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_031_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "pair_controlled_minority_critical_caveat_031_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_031_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_031_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_032", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_032_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_032_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_032_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "pair_controlled_minority_critical_caveat_032_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_032_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_032_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_033", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_033_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_033_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_033_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "pair_controlled_minority_critical_caveat_033_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_033_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_033_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_034", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_034_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_034_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_034_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "pair_controlled_minority_critical_caveat_034_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_034_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_034_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_035", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_035_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_035_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_035_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "pair_controlled_minority_critical_caveat_035_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_035_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_035_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_036", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_036_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_036_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_036_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "pair_controlled_minority_critical_caveat_036_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_036_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_036_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_037", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_037_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_037_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_037_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "pair_controlled_minority_critical_caveat_037_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_037_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_037_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_038", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_038_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_038_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_038_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "pair_controlled_minority_critical_caveat_038_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_038_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_038_trap_0"}]}
{"case_id": "pair_controlled_minority_critical_caveat_039", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_controlled_minority_critical_caveat_039_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_controlled_minority_critical_caveat_039_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_controlled_minority_critical_caveat_039_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "pair_controlled_minority_critical_caveat_039_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_controlled_minority_critical_caveat_039_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_controlled_minority_critical_caveat_039_trap_0"}]}
{"case_id": "pair_natural_000", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_000_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_000_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_000_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_000_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_000_trap_0"}]}
{"case_id": "pair_natural_001", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_001_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_001_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_001_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_001_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_001_trap_0"}]}
{"case_id": "pair_natural_002", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_002_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_002_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_002_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "pair_natural_002_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_002_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_002_trap_0"}]}
{"case_id": "pair_natural_003", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_003_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_003_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_003_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_003_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_003_trap_0"}]}
{"case_id": "pair_natural_004", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_004_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_004_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_004_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_004_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_004_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_004_trap_0"}]}
{"case_id": "pair_natural_005", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_005_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_005_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "pair_natural_005_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "pair_natural_005_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_005_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "pair_natural_005_trap_0"}]}
{"case_id": "pair_natural_006", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_006_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_006_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_006_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_006_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_006_trap_0"}]}
{"case_id": "pair_natural_007", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_007_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_007_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_007_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_007_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_007_trap_0"}]}
{"case_id": "pair_natural_008", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_008_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_008_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_008_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_008_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_008_trap_0"}]}
{"case_id": "pair_natural_009", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_009_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_009_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_009_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_009_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_009_trap_0"}]}
{"case_id": "pair_natural_010", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_010_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_010_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_010_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "pair_natural_010_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_010_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_010_trap_0"}]}
{"case_id": "pair_natural_011", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_011_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_011_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_011_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_011_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_011_trap_0"}]}
{"case_id": "pair_natural_012", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_012_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_012_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_012_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_012_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_012_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_012_trap_0"}]}
{"case_id": "pair_natural_013", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_013_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_013_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "pair_natural_013_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "pair_natural_013_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_013_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "pair_natural_013_trap_0"}]}
{"case_id": "pair_natural_014", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_014_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_014_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_014_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_014_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_014_trap_0"}]}
{"case_id": "pair_natural_015", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_015_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_015_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_015_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_015_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_015_trap_0"}]}
{"case_id": "pair_natural_016", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_016_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_016_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_016_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_016_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_016_trap_0"}]}
{"case_id": "pair_natural_017", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_017_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_017_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_017_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_017_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_017_trap_0"}]}
{"case_id": "pair_natural_018", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_018_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_018_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_018_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "pair_natural_018_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_018_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_018_trap_0"}]}
{"case_id": "pair_natural_019", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_019_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_019_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_019_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_019_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_019_trap_0"}]}
{"case_id": "pair_natural_020", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_020_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_020_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_020_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_020_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_020_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_020_trap_0"}]}
{"case_id": "pair_natural_021", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_021_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_021_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "pair_natural_021_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "pair_natural_021_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_021_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "pair_natural_021_trap_0"}]}
{"case_id": "pair_natural_022", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_022_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_022_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_022_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_022_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_022_trap_0"}]}
{"case_id": "pair_natural_023", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_023_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_023_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_023_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_023_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_023_trap_0"}]}
{"case_id": "pair_natural_024", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_024_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_024_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_024_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_024_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_024_trap_0"}]}
{"case_id": "pair_natural_025", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_025_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_025_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_025_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_025_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_025_trap_0"}]}
{"case_id": "pair_natural_026", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_026_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_026_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_026_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "pair_natural_026_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_026_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_026_trap_0"}]}
{"case_id": "pair_natural_027", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_027_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_027_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_027_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_027_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_027_trap_0"}]}
{"case_id": "pair_natural_028", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_028_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_028_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_028_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_028_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_028_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_028_trap_0"}]}
{"case_id": "pair_natural_029", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_029_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_029_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "pair_natural_029_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "pair_natural_029_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_029_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "pair_natural_029_trap_0"}]}
{"case_id": "pair_natural_030", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_030_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_030_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_030_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_030_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_030_trap_0"}]}
{"case_id": "pair_natural_031", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_031_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_031_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_031_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_031_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_031_trap_0"}]}
{"case_id": "pair_natural_032", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_032_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_032_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_032_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_032_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_032_trap_0"}]}
{"case_id": "pair_natural_033", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_033_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_033_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_033_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_033_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_033_trap_0"}]}
{"case_id": "pair_natural_034", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_034_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_034_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_034_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "pair_natural_034_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_034_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_034_trap_0"}]}
{"case_id": "pair_natural_035", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_035_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_035_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_035_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_035_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_035_trap_0"}]}
{"case_id": "pair_natural_036", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_036_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_036_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_036_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_036_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_036_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_036_trap_0"}]}
{"case_id": "pair_natural_037", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_037_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_037_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "pair_natural_037_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "pair_natural_037_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_037_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "pair_natural_037_trap_0"}]}
{"case_id": "pair_natural_038", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_038_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_038_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_038_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_038_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_038_trap_0"}]}
{"case_id": "pair_natural_039", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_039_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_039_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_039_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_039_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_039_trap_0"}]}
{"case_id": "pair_natural_040", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_040_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_040_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_040_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_040_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_040_trap_0"}]}
{"case_id": "pair_natural_041", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_041_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_041_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_041_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_041_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_041_trap_0"}]}
{"case_id": "pair_natural_042", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_042_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_042_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_042_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "pair_natural_042_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_042_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_042_trap_0"}]}
{"case_id": "pair_natural_043", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_043_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_043_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_043_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_043_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_043_trap_0"}]}
{"case_id": "pair_natural_044", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_044_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_044_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_044_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_044_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_044_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_044_trap_0"}]}
{"case_id": "pair_natural_045", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_045_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_045_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "pair_natural_045_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "pair_natural_045_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_045_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "pair_natural_045_trap_0"}]}
{"case_id": "pair_natural_046", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_046_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_046_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_046_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_046_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_046_trap_0"}]}
{"case_id": "pair_natural_047", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_047_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_047_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_047_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_047_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_047_trap_0"}]}
{"case_id": "pair_natural_048", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_048_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_048_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_048_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_048_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_048_trap_0"}]}
{"case_id": "pair_natural_049", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_049_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_049_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_049_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_049_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_049_trap_0"}]}
{"case_id": "pair_natural_050", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_050_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_050_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_050_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "pair_natural_050_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_050_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_050_trap_0"}]}
{"case_id": "pair_natural_051", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_051_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_051_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_051_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_051_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_051_trap_0"}]}
{"case_id": "pair_natural_052", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_052_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_052_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_052_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_052_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_052_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_052_trap_0"}]}
{"case_id": "pair_natural_053", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_053_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_053_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "pair_natural_053_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "pair_natural_053_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_053_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "pair_natural_053_trap_0"}]}
{"case_id": "pair_natural_054", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_054_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_054_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_054_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_054_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_054_trap_0"}]}
{"case_id": "pair_natural_055", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_055_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_055_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_055_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_055_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_055_trap_0"}]}
{"case_id": "pair_natural_056", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_056_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_056_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_056_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_056_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_056_trap_0"}]}
{"case_id": "pair_natural_057", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_057_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_057_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_057_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_057_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_057_trap_0"}]}
{"case_id": "pair_natural_058", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_058_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_058_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_058_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "pair_natural_058_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_058_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_058_trap_0"}]}
{"case_id": "pair_natural_059", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_059_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_059_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_059_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_059_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_059_trap_0"}]}
{"case_id": "pair_natural_060", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_060_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_060_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_060_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_060_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_060_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_060_trap_0"}]}
{"case_id": "pair_natural_061", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_061_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_061_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "pair_natural_061_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "pair_natural_061_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_061_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "pair_natural_061_trap_0"}]}
{"case_id": "pair_natural_062", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_062_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_062_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_062_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_062_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_062_trap_0"}]}
{"case_id": "pair_natural_063", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_063_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_063_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_063_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_063_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_063_trap_0"}]}
{"case_id": "pair_natural_064", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_064_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_064_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_064_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_064_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_064_trap_0"}]}
{"case_id": "pair_natural_065", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_065_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_065_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_065_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_065_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_065_trap_0"}]}
{"case_id": "pair_natural_066", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_066_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_066_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_066_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "pair_natural_066_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_066_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_066_trap_0"}]}
{"case_id": "pair_natural_067", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_067_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_067_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_067_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_067_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_067_trap_0"}]}
{"case_id": "pair_natural_068", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_068_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_068_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_068_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_068_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_068_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_068_trap_0"}]}
{"case_id": "pair_natural_069", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_069_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_069_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "pair_natural_069_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "pair_natural_069_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_069_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "pair_natural_069_trap_0"}]}
{"case_id": "pair_natural_070", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_070_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_070_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_070_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_070_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_070_trap_0"}]}
{"case_id": "pair_natural_071", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_071_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_071_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_071_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_071_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_071_trap_0"}]}
{"case_id": "pair_natural_072", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_072_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_072_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_072_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_072_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_072_trap_0"}]}
{"case_id": "pair_natural_073", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_073_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_073_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_073_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_073_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_073_trap_0"}]}
{"case_id": "pair_natural_074", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_074_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_074_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_074_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "pair_natural_074_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_074_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_074_trap_0"}]}
{"case_id": "pair_natural_075", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_075_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_075_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_075_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_075_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_075_trap_0"}]}
{"case_id": "pair_natural_076", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_076_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_076_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_076_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_076_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_076_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_076_trap_0"}]}
{"case_id": "pair_natural_077", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_077_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_077_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "pair_natural_077_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "pair_natural_077_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_077_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "pair_natural_077_trap_0"}]}
{"case_id": "pair_natural_078", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_078_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_078_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_078_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_078_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_078_trap_0"}]}
{"case_id": "pair_natural_079", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_079_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_079_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_079_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_079_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_079_trap_0"}]}
{"case_id": "pair_natural_080", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_080_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_080_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_080_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_080_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_080_trap_0"}]}
{"case_id": "pair_natural_081", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_081_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_081_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_081_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_081_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_081_trap_0"}]}
{"case_id": "pair_natural_082", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_082_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_082_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_082_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "pair_natural_082_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_082_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_082_trap_0"}]}
{"case_id": "pair_natural_083", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_083_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_083_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_083_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_083_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_083_trap_0"}]}
{"case_id": "pair_natural_084", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_084_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_084_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_084_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_084_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_084_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_084_trap_0"}]}
{"case_id": "pair_natural_085", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_085_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_085_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "pair_natural_085_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "pair_natural_085_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_085_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "pair_natural_085_trap_0"}]}
{"case_id": "pair_natural_086", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_086_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_086_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_086_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_086_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_086_trap_0"}]}
{"case_id": "pair_natural_087", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_087_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_087_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_087_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_087_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_087_trap_0"}]}
{"case_id": "pair_natural_088", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_088_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_088_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_088_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_088_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_088_trap_0"}]}
{"case_id": "pair_natural_089", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_089_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_089_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_089_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_089_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_089_trap_0"}]}
{"case_id": "pair_natural_090", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_090_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_090_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_090_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "pair_natural_090_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_090_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_090_trap_0"}]}
{"case_id": "pair_natural_091", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_091_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_091_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_091_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_091_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_091_trap_0"}]}
{"case_id": "pair_natural_092", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_092_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_092_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_092_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_092_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_092_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_092_trap_0"}]}
{"case_id": "pair_natural_093", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_093_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_093_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "pair_natural_093_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "pair_natural_093_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_093_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "pair_natural_093_trap_0"}]}
{"case_id": "pair_natural_094", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_094_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_094_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_094_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_094_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_094_trap_0"}]}
{"case_id": "pair_natural_095", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_095_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_095_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_095_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_095_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_095_trap_0"}]}
{"case_id": "pair_natural_096", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_096_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_096_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_096_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_096_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_096_trap_0"}]}
{"case_id": "pair_natural_097", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_097_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_097_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_097_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_097_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_097_trap_0"}]}
{"case_id": "pair_natural_098", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_098_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_098_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_098_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "pair_natural_098_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_098_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_098_trap_0"}]}
{"case_id": "pair_natural_099", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_099_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_099_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_099_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_099_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_099_trap_0"}]}
{"case_id": "pair_natural_100", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_100_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_100_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_100_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_100_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_100_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_100_trap_0"}]}
{"case_id": "pair_natural_101", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_101_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_101_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "pair_natural_101_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "pair_natural_101_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_101_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "pair_natural_101_trap_0"}]}
{"case_id": "pair_natural_102", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_102_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_102_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_102_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_102_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_102_trap_0"}]}
{"case_id": "pair_natural_103", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_103_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_103_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_103_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_103_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_103_trap_0"}]}
{"case_id": "pair_natural_104", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_104_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_104_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_104_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_104_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_104_trap_0"}]}
{"case_id": "pair_natural_105", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_105_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_105_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_105_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_105_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_105_trap_0"}]}
{"case_id": "pair_natural_106", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_106_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_106_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_106_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "pair_natural_106_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_106_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_106_trap_0"}]}
{"case_id": "pair_natural_107", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_107_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_107_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_107_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_107_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_107_trap_0"}]}
{"case_id": "pair_natural_108", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_108_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_108_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_108_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_108_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_108_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_108_trap_0"}]}
{"case_id": "pair_natural_109", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_109_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_109_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "pair_natural_109_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "pair_natural_109_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_109_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "pair_natural_109_trap_0"}]}
{"case_id": "pair_natural_110", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_110_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_110_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_110_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_110_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_110_trap_0"}]}
{"case_id": "pair_natural_111", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_111_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_111_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_111_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_111_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_111_trap_0"}]}
{"case_id": "pair_natural_112", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_112_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_112_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_112_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_112_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_112_trap_0"}]}
{"case_id": "pair_natural_113", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_113_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_113_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_113_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_113_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_113_trap_0"}]}
{"case_id": "pair_natural_114", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_114_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_114_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_114_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "pair_natural_114_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_114_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_114_trap_0"}]}
{"case_id": "pair_natural_115", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_115_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_115_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_115_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_115_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_115_trap_0"}]}
{"case_id": "pair_natural_116", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_116_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_116_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_116_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_116_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_116_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_116_trap_0"}]}
{"case_id": "pair_natural_117", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_117_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_117_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "pair_natural_117_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "pair_natural_117_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_117_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "pair_natural_117_trap_0"}]}
{"case_id": "pair_natural_118", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_118_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_118_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_118_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_118_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_118_trap_0"}]}
{"case_id": "pair_natural_119", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_119_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_119_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_119_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_119_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_119_trap_0"}]}
{"case_id": "pair_natural_120", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_120_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_120_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_120_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_120_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_120_trap_0"}]}
{"case_id": "pair_natural_121", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_121_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_121_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_121_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_121_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_121_trap_0"}]}
{"case_id": "pair_natural_122", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_122_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_122_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_122_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "pair_natural_122_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_122_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_122_trap_0"}]}
{"case_id": "pair_natural_123", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_123_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_123_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_123_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_123_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_123_trap_0"}]}
{"case_id": "pair_natural_124", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_124_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_124_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_124_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_124_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_124_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_124_trap_0"}]}
{"case_id": "pair_natural_125", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_125_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_125_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "pair_natural_125_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "pair_natural_125_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_125_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "pair_natural_125_trap_0"}]}
{"case_id": "pair_natural_126", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_126_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_126_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_126_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_126_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_126_trap_0"}]}
{"case_id": "pair_natural_127", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_127_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_127_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_127_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_127_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_127_trap_0"}]}
{"case_id": "pair_natural_128", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_128_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_128_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_128_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_128_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_128_trap_0"}]}
{"case_id": "pair_natural_129", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_129_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_129_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_129_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_129_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_129_trap_0"}]}
{"case_id": "pair_natural_130", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_130_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_130_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_130_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "pair_natural_130_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_130_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_130_trap_0"}]}
{"case_id": "pair_natural_131", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_131_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_131_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_131_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_131_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_131_trap_0"}]}
{"case_id": "pair_natural_132", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_132_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_132_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_132_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_132_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_132_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_132_trap_0"}]}
{"case_id": "pair_natural_133", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_133_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_133_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "pair_natural_133_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "pair_natural_133_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_133_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "pair_natural_133_trap_0"}]}
{"case_id": "pair_natural_134", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_134_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_134_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_134_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_134_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_134_trap_0"}]}
{"case_id": "pair_natural_135", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_135_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_135_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_135_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_135_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_135_trap_0"}]}
{"case_id": "pair_natural_136", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_136_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_136_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_136_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_136_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_136_trap_0"}]}
{"case_id": "pair_natural_137", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_137_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_137_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_137_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_137_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_137_trap_0"}]}
{"case_id": "pair_natural_138", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_138_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_138_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_138_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "pair_natural_138_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_138_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_138_trap_0"}]}
{"case_id": "pair_natural_139", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_139_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_139_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_139_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_139_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_139_trap_0"}]}
{"case_id": "pair_natural_140", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_140_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_140_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_140_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_140_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_140_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_140_trap_0"}]}
{"case_id": "pair_natural_141", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_141_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_141_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "pair_natural_141_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "pair_natural_141_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_141_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "pair_natural_141_trap_0"}]}
{"case_id": "pair_natural_142", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_142_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_142_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_142_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_142_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_142_trap_0"}]}
{"case_id": "pair_natural_143", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_143_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_143_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_143_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_143_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_143_trap_0"}]}
{"case_id": "pair_natural_144", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_144_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_144_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_144_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_144_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_144_trap_0"}]}
{"case_id": "pair_natural_145", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_145_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_145_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_145_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_145_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_145_trap_0"}]}
{"case_id": "pair_natural_146", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_146_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_146_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_146_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "pair_natural_146_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_146_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_146_trap_0"}]}
{"case_id": "pair_natural_147", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_147_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_147_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_147_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_147_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_147_trap_0"}]}
{"case_id": "pair_natural_148", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_148_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_148_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_148_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_148_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_148_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_148_trap_0"}]}
{"case_id": "pair_natural_149", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_149_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_149_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "pair_natural_149_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "pair_natural_149_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_149_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "pair_natural_149_trap_0"}]}
{"case_id": "pair_natural_150", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_150_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_150_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_150_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_150_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_150_trap_0"}]}
{"case_id": "pair_natural_151", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_151_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_151_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_151_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_151_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_151_trap_0"}]}
{"case_id": "pair_natural_152", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_152_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_152_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_152_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_152_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_152_trap_0"}]}
{"case_id": "pair_natural_153", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_153_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_153_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_153_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_153_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_153_trap_0"}]}
{"case_id": "pair_natural_154", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_154_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_154_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_154_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "pair_natural_154_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_154_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_154_trap_0"}]}
{"case_id": "pair_natural_155", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_155_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_155_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_155_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_155_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_155_trap_0"}]}
{"case_id": "pair_natural_156", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_156_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_156_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_156_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_156_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_156_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_156_trap_0"}]}
{"case_id": "pair_natural_157", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_157_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_157_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "pair_natural_157_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "pair_natural_157_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_157_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "pair_natural_157_trap_0"}]}
{"case_id": "pair_natural_158", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_158_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_158_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_158_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_158_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_158_trap_0"}]}
{"case_id": "pair_natural_159", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_159_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_159_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_159_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_159_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_159_trap_0"}]}
{"case_id": "pair_natural_160", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_160_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_160_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_160_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_160_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_160_trap_0"}]}
{"case_id": "pair_natural_161", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_161_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_161_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_161_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_161_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_161_trap_0"}]}
{"case_id": "pair_natural_162", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_162_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_162_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_162_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "pair_natural_162_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_162_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_162_trap_0"}]}
{"case_id": "pair_natural_163", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_163_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_163_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_163_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_163_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_163_trap_0"}]}
{"case_id": "pair_natural_164", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_164_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_164_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_164_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_164_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_164_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_164_trap_0"}]}
{"case_id": "pair_natural_165", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_165_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_165_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "pair_natural_165_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "pair_natural_165_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_165_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "pair_natural_165_trap_0"}]}
{"case_id": "pair_natural_166", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_166_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_166_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_166_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_166_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_166_trap_0"}]}
{"case_id": "pair_natural_167", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_167_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_167_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_167_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_167_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_167_trap_0"}]}
{"case_id": "pair_natural_168", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_168_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_168_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_168_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_168_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_168_trap_0"}]}
{"case_id": "pair_natural_169", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_169_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_169_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_169_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_169_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_169_trap_0"}]}
{"case_id": "pair_natural_170", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_170_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_170_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_170_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "pair_natural_170_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_170_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_170_trap_0"}]}
{"case_id": "pair_natural_171", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_171_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_171_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_171_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_171_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_171_trap_0"}]}
{"case_id": "pair_natural_172", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_172_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_172_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_172_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_172_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_172_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_172_trap_0"}]}
{"case_id": "pair_natural_173", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_173_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_173_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "pair_natural_173_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "pair_natural_173_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_173_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "pair_natural_173_trap_0"}]}
{"case_id": "pair_natural_174", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_174_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_174_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_174_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_174_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_174_trap_0"}]}
{"case_id": "pair_natural_175", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_175_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_175_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_175_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_175_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_175_trap_0"}]}
{"case_id": "pair_natural_176", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_176_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_176_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_176_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_176_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_176_trap_0"}]}
{"case_id": "pair_natural_177", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_177_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_177_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_177_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_177_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_177_trap_0"}]}
{"case_id": "pair_natural_178", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_178_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_178_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_178_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "pair_natural_178_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_178_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_178_trap_0"}]}
{"case_id": "pair_natural_179", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_179_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_179_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_179_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_179_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_179_trap_0"}]}
{"case_id": "pair_natural_180", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_180_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_180_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_180_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_180_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_180_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_180_trap_0"}]}
{"case_id": "pair_natural_181", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_181_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_181_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "pair_natural_181_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "pair_natural_181_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_181_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "pair_natural_181_trap_0"}]}
{"case_id": "pair_natural_182", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_182_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_182_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_182_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_182_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_182_trap_0"}]}
{"case_id": "pair_natural_183", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_183_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_183_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_183_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_183_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_183_trap_0"}]}
{"case_id": "pair_natural_184", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_184_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_184_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_184_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_184_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_184_trap_0"}]}
{"case_id": "pair_natural_185", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_185_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_185_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_185_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_185_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_185_trap_0"}]}
{"case_id": "pair_natural_186", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_186_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_186_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_186_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "pair_natural_186_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_186_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_186_trap_0"}]}
{"case_id": "pair_natural_187", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_187_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_187_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_187_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_187_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_187_trap_0"}]}
{"case_id": "pair_natural_188", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_188_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_188_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_188_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_188_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_188_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_188_trap_0"}]}
{"case_id": "pair_natural_189", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_189_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_189_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "pair_natural_189_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "pair_natural_189_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_189_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "pair_natural_189_trap_0"}]}
{"case_id": "pair_natural_190", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_190_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_190_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_190_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_190_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_190_trap_0"}]}
{"case_id": "pair_natural_191", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_191_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_191_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_191_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_191_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_191_trap_0"}]}
{"case_id": "pair_natural_192", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_192_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "pair_natural_192_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "pair_natural_192_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_192_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "pair_natural_192_trap_0"}]}
{"case_id": "pair_natural_193", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_193_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "pair_natural_193_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "pair_natural_193_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "pair_natural_193_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "pair_natural_193_trap_0"}]}
{"case_id": "pair_natural_194", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_194_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_194_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "pair_natural_194_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "pair_natural_194_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_194_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "pair_natural_194_trap_0"}]}
{"case_id": "pair_natural_195", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_195_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "pair_natural_195_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "pair_natural_195_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_195_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "pair_natural_195_trap_0"}]}
{"case_id": "pair_natural_196", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_196_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_196_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "pair_natural_196_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "pair_natural_196_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_196_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "pair_natural_196_trap_0"}]}
{"case_id": "pair_natural_197", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_197_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "pair_natural_197_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "pair_natural_197_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "pair_natural_197_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "pair_natural_197_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 4.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "pair_natural_197_trap_0"}]}
{"case_id": "pair_natural_198", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_198_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "pair_natural_198_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "pair_natural_198_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "pair_natural_198_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "pair_natural_198_trap_0"}]}
{"case_id": "pair_natural_199", "caveats": [{"attaches_to": "final_decision", "caveat_id": "pair_natural_199_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "pair_natural_199_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "pair_natural_199_prov_1", "expected_sources": ["A2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "pair_natural_199_rel_0", "units": ["A2", "B1"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "pair_natural_199_trap_0"}]}
{"case_id": "k3_000", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_000_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_000_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "k3_000_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "k3_000_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_000_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_000_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_000_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "k3_000_trap_0"}]}
{"case_id": "k3_001", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_001_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_001_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_001_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_001_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_001_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_001_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_001_trap_0"}]}
{"case_id": "k3_002", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_002_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_002_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_002_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_002_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_002_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_002_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_002_trap_0"}]}
{"case_id": "k3_003", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_003_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_003_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_003_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_003_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_003_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_003_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_003_trap_0"}]}
{"case_id": "k3_004", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_004_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_004_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_004_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_004_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_004_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_004_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_004_trap_0"}]}
{"case_id": "k3_005", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_005_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_005_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_005_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_005_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_005_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_005_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_005_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_005_trap_0"}]}
{"case_id": "k3_006", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_006_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_006_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_006_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_006_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_006_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_006_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_006_trap_0"}]}
{"case_id": "k3_007", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_007_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_007_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_007_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "k3_007_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_007_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_007_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_007_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_007_trap_0"}]}
{"case_id": "k3_008", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_008_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_008_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "k3_008_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "k3_008_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_008_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_008_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_008_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "k3_008_trap_0"}]}
{"case_id": "k3_009", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_009_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_009_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_009_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_009_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_009_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_009_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_009_trap_0"}]}
{"case_id": "k3_010", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_010_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_010_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_010_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_010_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_010_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_010_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_010_trap_0"}]}
{"case_id": "k3_011", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_011_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_011_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_011_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_011_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_011_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_011_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_011_trap_0"}]}
{"case_id": "k3_012", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_012_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_012_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_012_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_012_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_012_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_012_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_012_trap_0"}]}
{"case_id": "k3_013", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_013_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_013_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_013_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_013_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_013_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_013_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_013_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_013_trap_0"}]}
{"case_id": "k3_014", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_014_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_014_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_014_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_014_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_014_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_014_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_014_trap_0"}]}
{"case_id": "k3_015", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_015_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_015_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_015_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "k3_015_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_015_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_015_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_015_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_015_trap_0"}]}
{"case_id": "k3_016", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_016_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_016_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "k3_016_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "k3_016_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_016_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_016_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_016_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "k3_016_trap_0"}]}
{"case_id": "k3_017", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_017_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_017_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_017_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_017_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_017_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_017_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_017_trap_0"}]}
{"case_id": "k3_018", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_018_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_018_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_018_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_018_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_018_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_018_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_018_trap_0"}]}
{"case_id": "k3_019", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_019_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_019_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_019_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_019_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_019_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_019_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_019_trap_0"}]}
{"case_id": "k3_020", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_020_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_020_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_020_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_020_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_020_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_020_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_020_trap_0"}]}
{"case_id": "k3_021", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_021_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_021_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_021_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_021_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_021_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_021_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_021_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_021_trap_0"}]}
{"case_id": "k3_022", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_022_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_022_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_022_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_022_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_022_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_022_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_022_trap_0"}]}
{"case_id": "k3_023", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_023_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_023_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_023_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "k3_023_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_023_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_023_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_023_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_023_trap_0"}]}
{"case_id": "k3_024", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_024_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_024_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "k3_024_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "k3_024_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_024_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_024_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_024_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "k3_024_trap_0"}]}
{"case_id": "k3_025", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_025_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_025_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_025_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_025_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_025_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_025_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_025_trap_0"}]}
{"case_id": "k3_026", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_026_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_026_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_026_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_026_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_026_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_026_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_026_trap_0"}]}
{"case_id": "k3_027", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_027_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_027_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_027_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_027_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_027_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_027_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_027_trap_0"}]}
{"case_id": "k3_028", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_028_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_028_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_028_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_028_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_028_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_028_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_028_trap_0"}]}
{"case_id": "k3_029", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_029_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_029_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_029_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_029_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_029_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_029_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_029_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_029_trap_0"}]}
{"case_id": "k3_030", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_030_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_030_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_030_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_030_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_030_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_030_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_030_trap_0"}]}
{"case_id": "k3_031", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_031_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_031_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_031_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "k3_031_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_031_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_031_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_031_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_031_trap_0"}]}
{"case_id": "k3_032", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_032_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_032_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "k3_032_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "k3_032_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_032_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_032_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_032_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "k3_032_trap_0"}]}
{"case_id": "k3_033", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_033_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_033_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_033_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_033_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_033_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_033_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_033_trap_0"}]}
{"case_id": "k3_034", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_034_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_034_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_034_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_034_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_034_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_034_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_034_trap_0"}]}
{"case_id": "k3_035", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_035_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_035_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_035_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_035_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_035_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_035_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_035_trap_0"}]}
{"case_id": "k3_036", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_036_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_036_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_036_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_036_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_036_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_036_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_036_trap_0"}]}
{"case_id": "k3_037", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_037_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_037_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_037_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_037_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_037_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_037_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_037_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_037_trap_0"}]}
{"case_id": "k3_038", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_038_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_038_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_038_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_038_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_038_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_038_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_038_trap_0"}]}
{"case_id": "k3_039", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_039_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_039_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_039_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "k3_039_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_039_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_039_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_039_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_039_trap_0"}]}
{"case_id": "k3_040", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_040_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_040_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "k3_040_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "k3_040_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_040_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_040_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_040_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "k3_040_trap_0"}]}
{"case_id": "k3_041", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_041_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_041_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_041_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_041_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_041_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_041_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_041_trap_0"}]}
{"case_id": "k3_042", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_042_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_042_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_042_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_042_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_042_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_042_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_042_trap_0"}]}
{"case_id": "k3_043", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_043_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_043_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_043_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_043_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_043_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_043_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_043_trap_0"}]}
{"case_id": "k3_044", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_044_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_044_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_044_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_044_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_044_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_044_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_044_trap_0"}]}
{"case_id": "k3_045", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_045_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_045_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_045_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_045_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_045_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_045_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_045_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_045_trap_0"}]}
{"case_id": "k3_046", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_046_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_046_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_046_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_046_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_046_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_046_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_046_trap_0"}]}
{"case_id": "k3_047", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_047_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_047_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_047_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "k3_047_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_047_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_047_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_047_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_047_trap_0"}]}
{"case_id": "k3_048", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_048_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_048_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "k3_048_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "k3_048_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_048_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_048_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_048_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "k3_048_trap_0"}]}
{"case_id": "k3_049", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_049_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_049_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_049_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_049_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_049_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_049_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_049_trap_0"}]}
{"case_id": "k3_050", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_050_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_050_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_050_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_050_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_050_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_050_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_050_trap_0"}]}
{"case_id": "k3_051", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_051_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_051_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_051_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_051_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_051_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_051_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_051_trap_0"}]}
{"case_id": "k3_052", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_052_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_052_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_052_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_052_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_052_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_052_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_052_trap_0"}]}
{"case_id": "k3_053", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_053_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_053_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_053_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_053_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_053_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_053_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_053_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_053_trap_0"}]}
{"case_id": "k3_054", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_054_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_054_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_054_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_054_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_054_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_054_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_054_trap_0"}]}
{"case_id": "k3_055", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_055_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_055_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_055_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "k3_055_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_055_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_055_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_055_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_055_trap_0"}]}
{"case_id": "k3_056", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_056_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_056_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "k3_056_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "k3_056_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_056_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_056_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_056_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "k3_056_trap_0"}]}
{"case_id": "k3_057", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_057_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_057_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_057_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_057_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_057_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_057_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_057_trap_0"}]}
{"case_id": "k3_058", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_058_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_058_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_058_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_058_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_058_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_058_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_058_trap_0"}]}
{"case_id": "k3_059", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_059_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_059_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_059_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_059_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_059_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_059_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_059_trap_0"}]}
{"case_id": "k3_060", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_060_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_060_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_060_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_060_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_060_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_060_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_060_trap_0"}]}
{"case_id": "k3_061", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_061_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_061_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_061_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_061_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_061_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_061_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_061_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_061_trap_0"}]}
{"case_id": "k3_062", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_062_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_062_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_062_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_062_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_062_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_062_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_062_trap_0"}]}
{"case_id": "k3_063", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_063_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_063_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_063_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "k3_063_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_063_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_063_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_063_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_063_trap_0"}]}
{"case_id": "k3_064", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_064_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_064_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "k3_064_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "k3_064_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_064_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_064_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_064_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "k3_064_trap_0"}]}
{"case_id": "k3_065", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_065_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_065_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_065_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_065_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_065_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_065_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_065_trap_0"}]}
{"case_id": "k3_066", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_066_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_066_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_066_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_066_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_066_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_066_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_066_trap_0"}]}
{"case_id": "k3_067", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_067_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_067_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_067_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_067_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_067_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_067_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_067_trap_0"}]}
{"case_id": "k3_068", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_068_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_068_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_068_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_068_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_068_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_068_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_068_trap_0"}]}
{"case_id": "k3_069", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_069_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_069_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_069_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_069_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_069_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_069_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_069_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_069_trap_0"}]}
{"case_id": "k3_070", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_070_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_070_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_070_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_070_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_070_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_070_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_070_trap_0"}]}
{"case_id": "k3_071", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_071_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_071_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_071_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "k3_071_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_071_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_071_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_071_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_071_trap_0"}]}
{"case_id": "k3_072", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_072_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_072_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "k3_072_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "k3_072_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_072_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_072_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_072_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "k3_072_trap_0"}]}
{"case_id": "k3_073", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_073_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_073_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_073_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_073_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_073_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_073_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_073_trap_0"}]}
{"case_id": "k3_074", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_074_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_074_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_074_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_074_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_074_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_074_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_074_trap_0"}]}
{"case_id": "k3_075", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_075_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_075_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_075_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_075_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_075_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_075_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_075_trap_0"}]}
{"case_id": "k3_076", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_076_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_076_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_076_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_076_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_076_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_076_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_076_trap_0"}]}
{"case_id": "k3_077", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_077_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_077_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_077_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_077_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_077_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_077_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_077_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_077_trap_0"}]}
{"case_id": "k3_078", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_078_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_078_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_078_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_078_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_078_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_078_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_078_trap_0"}]}
{"case_id": "k3_079", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_079_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_079_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_079_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "k3_079_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_079_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_079_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_079_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_079_trap_0"}]}
{"case_id": "k3_080", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_080_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_080_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "k3_080_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "k3_080_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_080_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_080_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_080_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "k3_080_trap_0"}]}
{"case_id": "k3_081", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_081_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_081_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_081_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_081_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_081_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_081_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_081_trap_0"}]}
{"case_id": "k3_082", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_082_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_082_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_082_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_082_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_082_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_082_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_082_trap_0"}]}
{"case_id": "k3_083", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_083_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_083_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_083_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_083_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_083_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_083_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_083_trap_0"}]}
{"case_id": "k3_084", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_084_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_084_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_084_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_084_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_084_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_084_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_084_trap_0"}]}
{"case_id": "k3_085", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_085_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_085_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_085_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_085_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_085_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_085_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_085_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_085_trap_0"}]}
{"case_id": "k3_086", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_086_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_086_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_086_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_086_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_086_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_086_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_086_trap_0"}]}
{"case_id": "k3_087", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_087_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_087_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_087_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "k3_087_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_087_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_087_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_087_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_087_trap_0"}]}
{"case_id": "k3_088", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_088_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_088_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "k3_088_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "k3_088_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_088_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_088_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_088_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "k3_088_trap_0"}]}
{"case_id": "k3_089", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_089_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_089_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_089_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_089_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_089_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_089_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_089_trap_0"}]}
{"case_id": "k3_090", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_090_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_090_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_090_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_090_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_090_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_090_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_090_trap_0"}]}
{"case_id": "k3_091", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_091_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_091_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_091_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_091_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_091_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_091_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_091_trap_0"}]}
{"case_id": "k3_092", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_092_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_092_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_092_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_092_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_092_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_092_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_092_trap_0"}]}
{"case_id": "k3_093", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_093_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_093_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_093_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_093_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_093_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_093_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_093_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_093_trap_0"}]}
{"case_id": "k3_094", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_094_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_094_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_094_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_094_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_094_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_094_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_094_trap_0"}]}
{"case_id": "k3_095", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_095_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_095_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_095_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "k3_095_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_095_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_095_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_095_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_095_trap_0"}]}
{"case_id": "k3_096", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_096_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_096_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "k3_096_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "k3_096_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_096_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_096_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_096_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "k3_096_trap_0"}]}
{"case_id": "k3_097", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_097_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_097_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_097_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_097_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_097_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_097_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_097_trap_0"}]}
{"case_id": "k3_098", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_098_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_098_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_098_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_098_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_098_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_098_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_098_trap_0"}]}
{"case_id": "k3_099", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_099_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_099_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_099_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_099_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_099_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_099_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_099_trap_0"}]}
{"case_id": "k3_100", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_100_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_100_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_100_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_100_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_100_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_100_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_100_trap_0"}]}
{"case_id": "k3_101", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_101_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_101_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_101_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_101_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_101_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_101_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_101_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_101_trap_0"}]}
{"case_id": "k3_102", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_102_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_102_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_102_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_102_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_102_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_102_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_102_trap_0"}]}
{"case_id": "k3_103", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_103_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_103_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_103_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "k3_103_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_103_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_103_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_103_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_103_trap_0"}]}
{"case_id": "k3_104", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_104_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_104_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "k3_104_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "k3_104_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_104_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_104_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_104_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "k3_104_trap_0"}]}
{"case_id": "k3_105", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_105_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_105_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_105_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_105_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_105_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_105_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_105_trap_0"}]}
{"case_id": "k3_106", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_106_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_106_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_106_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_106_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_106_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_106_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_106_trap_0"}]}
{"case_id": "k3_107", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_107_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_107_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_107_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_107_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_107_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_107_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_107_trap_0"}]}
{"case_id": "k3_108", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_108_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_108_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_108_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_108_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_108_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_108_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_108_trap_0"}]}
{"case_id": "k3_109", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_109_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_109_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_109_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_109_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_109_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_109_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_109_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_109_trap_0"}]}
{"case_id": "k3_110", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_110_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_110_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_110_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_110_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_110_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_110_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_110_trap_0"}]}
{"case_id": "k3_111", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_111_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_111_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_111_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "k3_111_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_111_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_111_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_111_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_111_trap_0"}]}
{"case_id": "k3_112", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_112_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_112_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "k3_112_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "k3_112_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_112_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_112_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_112_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "k3_112_trap_0"}]}
{"case_id": "k3_113", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_113_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_113_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_113_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_113_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_113_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_113_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_113_trap_0"}]}
{"case_id": "k3_114", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_114_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_114_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_114_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_114_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_114_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_114_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_114_trap_0"}]}
{"case_id": "k3_115", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_115_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_115_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_115_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_115_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_115_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_115_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_115_trap_0"}]}
{"case_id": "k3_116", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_116_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_116_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_116_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_116_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_116_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_116_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_116_trap_0"}]}
{"case_id": "k3_117", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_117_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_117_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_117_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_117_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_117_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_117_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_117_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_117_trap_0"}]}
{"case_id": "k3_118", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_118_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_118_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_118_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_118_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_118_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_118_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_118_trap_0"}]}
{"case_id": "k3_119", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_119_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_119_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_119_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "k3_119_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_119_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_119_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_119_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_119_trap_0"}]}
{"case_id": "k3_120", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_120_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_120_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "k3_120_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "k3_120_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_120_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_120_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_120_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "k3_120_trap_0"}]}
{"case_id": "k3_121", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_121_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_121_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_121_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_121_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_121_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_121_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_121_trap_0"}]}
{"case_id": "k3_122", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_122_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_122_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_122_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_122_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_122_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_122_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_122_trap_0"}]}
{"case_id": "k3_123", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_123_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_123_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_123_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_123_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_123_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_123_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_123_trap_0"}]}
{"case_id": "k3_124", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_124_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_124_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_124_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_124_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_124_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_124_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_124_trap_0"}]}
{"case_id": "k3_125", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_125_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_125_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_125_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_125_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_125_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_125_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_125_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_125_trap_0"}]}
{"case_id": "k3_126", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_126_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_126_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_126_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_126_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_126_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_126_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_126_trap_0"}]}
{"case_id": "k3_127", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_127_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_127_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_127_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "k3_127_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_127_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_127_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_127_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_127_trap_0"}]}
{"case_id": "k3_128", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_128_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_128_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "k3_128_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "k3_128_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_128_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_128_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_128_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "k3_128_trap_0"}]}
{"case_id": "k3_129", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_129_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_129_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_129_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_129_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_129_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_129_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_129_trap_0"}]}
{"case_id": "k3_130", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_130_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_130_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_130_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_130_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_130_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_130_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_130_trap_0"}]}
{"case_id": "k3_131", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_131_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_131_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_131_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_131_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_131_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_131_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_131_trap_0"}]}
{"case_id": "k3_132", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_132_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_132_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_132_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_132_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_132_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_132_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_132_trap_0"}]}
{"case_id": "k3_133", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_133_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_133_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_133_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_133_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_133_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_133_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_133_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_133_trap_0"}]}
{"case_id": "k3_134", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_134_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_134_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_134_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_134_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_134_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_134_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_134_trap_0"}]}
{"case_id": "k3_135", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_135_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_135_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_135_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "k3_135_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_135_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_135_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_135_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_135_trap_0"}]}
{"case_id": "k3_136", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_136_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_136_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "k3_136_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "k3_136_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_136_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_136_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_136_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "k3_136_trap_0"}]}
{"case_id": "k3_137", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_137_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_137_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_137_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_137_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_137_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_137_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_137_trap_0"}]}
{"case_id": "k3_138", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_138_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_138_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_138_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_138_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_138_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_138_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_138_trap_0"}]}
{"case_id": "k3_139", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_139_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_139_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_139_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_139_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_139_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_139_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_139_trap_0"}]}
{"case_id": "k3_140", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_140_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_140_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_140_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_140_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_140_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_140_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_140_trap_0"}]}
{"case_id": "k3_141", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_141_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_141_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_141_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_141_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_141_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_141_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_141_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_141_trap_0"}]}
{"case_id": "k3_142", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_142_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_142_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_142_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_142_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_142_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_142_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_142_trap_0"}]}
{"case_id": "k3_143", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_143_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_143_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_143_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "k3_143_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_143_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_143_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_143_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_143_trap_0"}]}
{"case_id": "k3_144", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_144_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_144_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "k3_144_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "k3_144_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_144_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_144_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_144_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "k3_144_trap_0"}]}
{"case_id": "k3_145", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_145_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_145_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_145_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_145_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_145_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_145_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_145_trap_0"}]}
{"case_id": "k3_146", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_146_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_146_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_146_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_146_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_146_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_146_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_146_trap_0"}]}
{"case_id": "k3_147", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_147_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_147_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_147_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_147_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_147_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_147_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_147_trap_0"}]}
{"case_id": "k3_148", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_148_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_148_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_148_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_148_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_148_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_148_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_148_trap_0"}]}
{"case_id": "k3_149", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_149_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_149_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_149_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_149_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_149_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_149_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_149_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_149_trap_0"}]}
{"case_id": "k3_150", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_150_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_150_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_150_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_150_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_150_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_150_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_150_trap_0"}]}
{"case_id": "k3_151", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_151_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_151_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_151_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "k3_151_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_151_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_151_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_151_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_151_trap_0"}]}
{"case_id": "k3_152", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_152_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_152_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "k3_152_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "k3_152_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_152_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_152_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_152_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "k3_152_trap_0"}]}
{"case_id": "k3_153", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_153_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_153_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_153_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_153_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_153_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_153_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_153_trap_0"}]}
{"case_id": "k3_154", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_154_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_154_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_154_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_154_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_154_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_154_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_154_trap_0"}]}
{"case_id": "k3_155", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_155_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_155_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_155_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_155_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_155_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_155_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_155_trap_0"}]}
{"case_id": "k3_156", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_156_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_156_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_156_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_156_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_156_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_156_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_156_trap_0"}]}
{"case_id": "k3_157", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_157_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_157_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_157_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_157_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_157_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_157_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_157_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_157_trap_0"}]}
{"case_id": "k3_158", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_158_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_158_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_158_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_158_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_158_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_158_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_158_trap_0"}]}
{"case_id": "k3_159", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_159_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_159_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_159_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "k3_159_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_159_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_159_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_159_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_159_trap_0"}]}
{"case_id": "k3_160", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_160_cav_0", "text": "do not ship if refund rate rises by more than 2.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_160_conf_0", "description": "do not ship if refund rate rises by more than 2.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the refund rate guardrail.", "claim_id": "k3_160_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if refund rate rises by more than 2.0 percentage points", "claim_id": "k3_160_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_160_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if refund rate rises by more than 2.0 percentage points", "final_claim": "Shipping should remain conditioned on the refund rate guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_160_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_160_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If refund rate rises by more than 2.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary latency metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The refund rate increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no refund rate risk.", "trap_id": "k3_160_trap_0"}]}
{"case_id": "k3_161", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_161_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_161_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_161_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_161_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_161_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_161_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the completion gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_161_trap_0"}]}
{"case_id": "k3_162", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_162_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_162_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_162_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_162_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_162_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_162_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves accuracy for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_162_trap_0"}]}
{"case_id": "k3_163", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_163_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_163_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_163_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_163_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_163_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_163_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate retention gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_163_trap_0"}]}
{"case_id": "k3_164", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_164_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_164_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_164_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_164_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_164_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_164_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_164_trap_0"}]}
{"case_id": "k3_165", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_165_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_165_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_165_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_165_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_165_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_165_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_165_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured latency gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_165_trap_0"}]}
{"case_id": "k3_166", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_166_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_166_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_166_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_166_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_166_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_166_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the completion gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_166_trap_0"}]}
{"case_id": "k3_167", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_167_cav_0", "text": "critical cohort shows higher complaint volume", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_167_conf_0", "description": "critical cohort shows higher complaint volume", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_167_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher complaint volume", "claim_id": "k3_167_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_167_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher complaint volume", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_167_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_167_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher complaint volume.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_167_trap_0"}]}
{"case_id": "k3_168", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_168_cav_0", "text": "do not ship if manual review time rises by more than 4.5 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_168_conf_0", "description": "do not ship if manual review time rises by more than 4.5 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the manual review time guardrail.", "claim_id": "k3_168_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if manual review time rises by more than 4.5 percentage points", "claim_id": "k3_168_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_168_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if manual review time rises by more than 4.5 percentage points", "final_claim": "Shipping should remain conditioned on the manual review time guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_168_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_168_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If manual review time rises by more than 4.5 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary retention metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The manual review time increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no manual review time risk.", "trap_id": "k3_168_trap_0"}]}
{"case_id": "k3_169", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_169_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_169_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_169_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_169_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_169_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_169_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the coverage gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_169_trap_0"}]}
{"case_id": "k3_170", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_170_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_170_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_170_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_170_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_170_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_170_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves latency for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_170_trap_0"}]}
{"case_id": "k3_171", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_171_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_171_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_171_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_171_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_171_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_171_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate completion gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_171_trap_0"}]}
{"case_id": "k3_172", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_172_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_172_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_172_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_172_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_172_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_172_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_172_trap_0"}]}
{"case_id": "k3_173", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_173_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_173_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_173_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_173_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_173_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_173_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_173_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured retention gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_173_trap_0"}]}
{"case_id": "k3_174", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_174_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_174_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_174_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_174_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_174_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_174_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the coverage gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_174_trap_0"}]}
{"case_id": "k3_175", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_175_cav_0", "text": "critical cohort shows higher refund rate", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_175_conf_0", "description": "critical cohort shows higher refund rate", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_175_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher refund rate", "claim_id": "k3_175_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_175_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher refund rate", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_175_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_175_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher refund rate.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_175_trap_0"}]}
{"case_id": "k3_176", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_176_cav_0", "text": "do not ship if on-call load rises by more than 2.4 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_176_conf_0", "description": "do not ship if on-call load rises by more than 2.4 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the on-call load guardrail.", "claim_id": "k3_176_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if on-call load rises by more than 2.4 percentage points", "claim_id": "k3_176_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_176_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if on-call load rises by more than 2.4 percentage points", "final_claim": "Shipping should remain conditioned on the on-call load guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_176_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_176_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If on-call load rises by more than 2.4 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary completion metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The on-call load increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no on-call load risk.", "trap_id": "k3_176_trap_0"}]}
{"case_id": "k3_177", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_177_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_177_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_177_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_177_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_177_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_177_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the accuracy gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_177_trap_0"}]}
{"case_id": "k3_178", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_178_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_178_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_178_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_178_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_178_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_178_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves retention for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_178_trap_0"}]}
{"case_id": "k3_179", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_179_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_179_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_179_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_179_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_179_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_179_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate coverage gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_179_trap_0"}]}
{"case_id": "k3_180", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_180_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_180_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_180_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_180_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_180_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_180_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary latency metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_180_trap_0"}]}
{"case_id": "k3_181", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_181_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_181_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_181_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_181_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_181_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_181_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_181_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured completion gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_181_trap_0"}]}
{"case_id": "k3_182", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_182_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_182_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_182_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_182_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_182_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_182_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the accuracy gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_182_trap_0"}]}
{"case_id": "k3_183", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_183_cav_0", "text": "critical cohort shows higher manual review time", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_183_conf_0", "description": "critical cohort shows higher manual review time", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_183_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher manual review time", "claim_id": "k3_183_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_183_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher manual review time", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_183_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_183_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher manual review time.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_183_trap_0"}]}
{"case_id": "k3_184", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_184_cav_0", "text": "do not ship if dropout risk rises by more than 5.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_184_conf_0", "description": "do not ship if dropout risk rises by more than 5.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the dropout risk guardrail.", "claim_id": "k3_184_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if dropout risk rises by more than 5.0 percentage points", "claim_id": "k3_184_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_184_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if dropout risk rises by more than 5.0 percentage points", "final_claim": "Shipping should remain conditioned on the dropout risk guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_184_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_184_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary coverage metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If dropout risk rises by more than 5.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary coverage metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The dropout risk increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no dropout risk risk.", "trap_id": "k3_184_trap_0"}]}
{"case_id": "k3_185", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_185_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_185_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_185_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_185_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_185_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_185_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the latency gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_185_trap_0"}]}
{"case_id": "k3_186", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_186_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_186_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_186_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_186_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_186_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_186_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves completion for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_186_trap_0"}]}
{"case_id": "k3_187", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_187_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_187_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_187_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_187_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_187_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_187_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate accuracy gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_187_trap_0"}]}
{"case_id": "k3_188", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_188_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_188_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_188_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_188_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_188_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_188_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary retention metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_188_trap_0"}]}
{"case_id": "k3_189", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_189_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_189_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_189_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_189_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_189_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_189_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_189_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured coverage gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_189_trap_0"}]}
{"case_id": "k3_190", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_190_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_190_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_190_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_190_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_190_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_190_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the latency gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_190_trap_0"}]}
{"case_id": "k3_191", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_191_cav_0", "text": "critical cohort shows higher on-call load", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_191_conf_0", "description": "critical cohort shows higher on-call load", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_191_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher on-call load", "claim_id": "k3_191_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_191_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher on-call load", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_191_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_191_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher on-call load.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_191_trap_0"}]}
{"case_id": "k3_192", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_192_cav_0", "text": "do not ship if complaint volume rises by more than 3.0 percentage points", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_192_conf_0", "description": "do not ship if complaint volume rises by more than 3.0 percentage points", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Shipping should remain conditioned on the complaint volume guardrail.", "claim_id": "k3_192_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "do not ship if complaint volume rises by more than 3.0 percentage points", "claim_id": "k3_192_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_192_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "do not ship if complaint volume rises by more than 3.0 percentage points", "final_claim": "Shipping should remain conditioned on the complaint volume guardrail."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_192_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_192_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary accuracy metric improves by 3.1 percentage points.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If complaint volume rises by more than 3.0 percentage points, do not ship.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Ship because the primary accuracy metric improved significantly.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The complaint volume increase may be seasonal noise.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollout has no complaint volume risk.", "trap_id": "k3_192_trap_0"}]}
{"case_id": "k3_193", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_193_cav_0", "text": "small-team usage was not measured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The evidence supports an enterprise-only rollout.", "claim_id": "k3_193_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "small-team usage was not measured", "claim_id": "k3_193_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_193_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "small-team usage was not measured", "final_claim": "The evidence supports an enterprise-only rollout."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_193_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_193_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "Source A reports the retention gain on enterprise users.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A also notes that small-team usage was not measured.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Source B reports stable support tickets in the pilot.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends rollout only for enterprise users.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Small teams were verified to benefit equally.", "trap_id": "k3_193_trap_0"}]}
{"case_id": "k3_194", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_194_cav_0", "text": "keep the rollback path enabled", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The rollout should keep the rollback path enabled.", "claim_id": "k3_194_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "keep the rollback path enabled", "claim_id": "k3_194_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_194_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "keep the rollback path enabled", "final_claim": "The rollout should keep the rollback path enabled."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_194_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_194_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot improves coverage for the target cohort.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The result depends on keeping the rollback path enabled.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The new design reduces manual work.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "The rollback path caused no severe failures during the pilot.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The rollback path can be removed immediately.", "trap_id": "k3_194_trap_0"}]}
{"case_id": "k3_195", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_195_cav_0", "text": "retention remains unmeasured", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "A limited launch is supported, but retention remains unmeasured.", "claim_id": "k3_195_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "retention remains unmeasured", "claim_id": "k3_195_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_195_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "retention remains unmeasured", "final_claim": "A limited launch is supported, but retention remains unmeasured."}, "relations": [{"operation": "condition", "relation": "additive", "relation_id": "k3_195_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_195_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The pilot shows a moderate latency gain.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The pilot did not measure long-term retention.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Candidate B recommends a limited launch.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B says retention should be checked after launch.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Long-term retention improved by 12 percent.", "trap_id": "k3_195_trap_0"}]}
{"case_id": "k3_196", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_196_cav_0", "text": "normal support checklist", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "The sources jointly support shipping with the normal checklist.", "claim_id": "k3_196_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "normal support checklist", "claim_id": "k3_196_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_196_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "normal support checklist", "final_claim": "The sources jointly support shipping with the normal checklist."}, "relations": [{"operation": "merge", "relation": "additive", "relation_id": "k3_196_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_196_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The primary completion metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "The rollout needs the normal support checklist.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "Support load remained stable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping with the normal checklist.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The sources disagree about the support load.", "trap_id": "k3_196_trap_0"}]}
{"case_id": "k3_197", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_197_cav_0", "text": "risk tolerance determines whether to wait", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_197_conf_0", "description": "risk tolerance determines whether to wait", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance.", "claim_id": "k3_197_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "risk tolerance determines whether to wait", "claim_id": "k3_197_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_197_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "risk tolerance determines whether to wait", "final_claim": "Either shipping or waiting is defensible; the decision depends on risk tolerance."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_197_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_197_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The measured accuracy gain is positive but modest.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A recommends waiting for another week of data.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "The user-facing impact is likely noticeable.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B recommends shipping to learn faster.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "Both sources unanimously recommend immediate full launch.", "trap_id": "k3_197_trap_0"}]}
{"case_id": "k3_198", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_198_cav_0", "text": "traffic mix should remain similar", "unit_id": "A2"}], "conflicts": [], "expected_provenance": [{"claim": "Shipping is supported only under a similar traffic mix.", "claim_id": "k3_198_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "traffic mix should remain similar", "claim_id": "k3_198_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_198_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "traffic mix should remain similar", "final_claim": "Shipping is supported only under a similar traffic mix."}, "relations": [{"operation": "condition", "relation": "subsumes", "relation_id": "k3_198_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_198_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The result assumes traffic remains close to the pilot mix.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "If traffic shifts, the retention gain may not transfer.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because pilot traffic matched the target segment.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not test holiday traffic.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "The result is guaranteed under all traffic mixes.", "trap_id": "k3_198_trap_0"}]}
{"case_id": "k3_199", "caveats": [{"attaches_to": "final_decision", "caveat_id": "k3_199_cav_0", "text": "critical cohort shows higher dropout risk", "unit_id": "A2"}], "conflicts": [{"conflict_id": "k3_199_conf_0", "description": "critical cohort shows higher dropout risk", "units": ["A2", "B1"]}], "expected_provenance": [{"claim": "The aggregate gain should not override the critical-cohort caveat.", "claim_id": "k3_199_prov_0", "expected_sources": ["A1", "B1"]}, {"claim": "critical cohort shows higher dropout risk", "claim_id": "k3_199_prov_1", "expected_sources": ["A2"]}, {"claim": "monitor before full expansion", "claim_id": "k3_199_prov_2", "expected_sources": ["C2"]}], "reference": {"caveat": "critical cohort shows higher dropout risk", "final_claim": "The aggregate gain should not override the critical-cohort caveat."}, "relations": [{"operation": "condition", "relation": "conflict", "relation_id": "k3_199_rel_0", "units": ["A2", "B1"]}, {"operation": "merge", "relation": "additive", "relation_id": "k3_199_rel_1", "units": ["A2", "C2"]}], "units": [{"flags": ["metric"], "source": "A", "text": "The average coverage metric improves.", "type": "claim", "unit_id": "A1"}, {"flags": ["caveat", "condition"], "source": "A", "text": "A small but critical cohort shows higher dropout risk.", "type": "caveat", "unit_id": "A2"}, {"flags": ["recommendation"], "source": "B", "text": "B recommends shipping because aggregate metrics improved.", "type": "recommendation", "unit_id": "B1"}, {"flags": ["evidence"], "source": "B", "text": "B does not discuss the critical cohort.", "type": "evidence", "unit_id": "B2"}, {"flags": ["recommendation"], "source": "C", "text": "Candidate C supports a staged rollout.", "type": "recommendation", "unit_id": "C1"}, {"flags": ["caveat"], "source": "C", "text": "Monitoring must happen before full expansion.", "type": "caveat", "unit_id": "C2"}], "unsupported_traps": [{"text": "No cohort-specific risk was observed.", "trap_id": "k3_199_trap_0"}]}