{ "version": "jev-only-research-results-1", "at": "2026-09-18T06:16:51.026Z", "task": "Jev-only baseline, prompt/input ablations and sparse source-only reasoning", "status": "research_complete_not_republished", "primary_measure": "Full family agreement with the retained model-adjudicated reference. Every original family remains in the denominator; every case in a family must match type, partition/identity, registry mapping, relationships, uncertainty and effective evidence tier.", "core_measure": "Type, pairwise identity and registry mapping agreement; not full entity-resolution agreement.", "candidate": { "policy": "policy-jev-only-v2.json", "freeze": { "sha256": "ca151ca289862e5bf123b78e2e00a9b9aac8cd7e436111ab955ecdea26c8ef5b" }, "wording": "Explicit sufficient evidence for identity and households; verbatim sources and four-case batches. Probability gates are diagnostic only in the Jev-only arm." }, "runs": [ { "label": "jev-only-panel1-v1", "panel": "development", "role": "baseline", "denominator": { "families": 200, "cases": 204, "unique_atoms": 275, "atom_incidences": 284, "unique_original_rows": 829, "original_row_incidences": 844 }, "full_equivalent_families": 162, "core_equivalent_families": 173, "valid_cases": 197, "known_model_cost_dollars": 0.027520710000000004, "unpriced_components": 0, "median_batch_service_seconds": 0.4024442089999998, "p90_batch_service_seconds": 0.7459652089999997, "observed_invocation_seconds": 13.416257583, "serial_component_seconds": 26.083073376999995, "model_usage": { "jev-1.13.0": { "calls": 58, "known_dollars": 0.027520710000000004, "unpriced_calls": 0, "serial_seconds": 26.083073376999998, "known_input_tokens": 655255, "known_output_tokens": 41545, "unmeasured_input_calls": 0, "unmeasured_output_calls": 0 } }, "autonomy": { "jev_attempted_cases": 199, "reasoning_assigned_cases": 0, "valid_without_reasoning": 197, "unresolved_cases": 7, "unsupported_cases": 5, "explicit_uncertainty_cases": 24 }, "types": { "correct": 263, "total": 284, "missing": 21 }, "identity": { "correct": 45, "total": 119, "false_merges": 0, "false_separations": 1, "unsupported_merges_over_unknown": 0, "abstentions_on_definite": 37, "missing": 36 }, "registry": { "correct": 260, "total": 284 }, "relationships": { "true_positives": 0, "false_positives": 0, "false_negatives": 15 }, "evidence": { "sha256": "13ea889c2792cda7f9da7d2d6cbb69638927b013399210ba811cbbca7169b199" } }, { "label": "jev-only-panel1-v2", "panel": "development", "role": "candidate", "denominator": { "families": 200, "cases": 204, "unique_atoms": 275, "atom_incidences": 284, "unique_original_rows": 829, "original_row_incidences": 844 }, "full_equivalent_families": 190, "core_equivalent_families": 193, "valid_cases": 199, "known_model_cost_dollars": 0.028261716000000006, "unpriced_components": 0, "median_batch_service_seconds": 0.43707625000000005, "p90_batch_service_seconds": 0.9505737079999999, "observed_invocation_seconds": 14.729534584000001, "serial_component_seconds": 28.69933657999999, "model_usage": { "jev-1.13.0": { "calls": 59, "known_dollars": 0.028261716000000006, "unpriced_calls": 0, "serial_seconds": 28.699336579999983, "known_input_tokens": 672898, "known_output_tokens": 41652, "unmeasured_input_calls": 0, "unmeasured_output_calls": 0 } }, "autonomy": { "jev_attempted_cases": 199, "reasoning_assigned_cases": 0, "valid_without_reasoning": 199, "unresolved_cases": 5, "unsupported_cases": 5, "explicit_uncertainty_cases": 3 }, "types": { "correct": 267, "total": 284, "missing": 16 }, "identity": { "correct": 87, "total": 119, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "abstentions_on_definite": 0, "missing": 32 }, "registry": { "correct": 265, "total": 284 }, "relationships": { "true_positives": 13, "false_positives": 0, "false_negatives": 2 }, "evidence": { "sha256": "02eac2c6bb27119150e1a671953864f2eaf54c83a95a41f779964059703dabe6" } }, { "label": "jev-only-panel1-v2-repeat2", "panel": "development", "role": "candidate-repeat", "denominator": { "families": 200, "cases": 204, "unique_atoms": 275, "atom_incidences": 284, "unique_original_rows": 829, "original_row_incidences": 844 }, "full_equivalent_families": 192, "core_equivalent_families": 193, "valid_cases": 198, "known_model_cost_dollars": 0.027636546000000005, "unpriced_components": 0, "median_batch_service_seconds": 0.4505591250000002, "p90_batch_service_seconds": 0.9035617500000007, "observed_invocation_seconds": 15.149321, "serial_component_seconds": 28.872019131, "model_usage": { "jev-1.13.0": { "calls": 58, "known_dollars": 0.027636546000000005, "unpriced_calls": 0, "serial_seconds": 28.872019131000002, "known_input_tokens": 658013, "known_output_tokens": 41594, "unmeasured_input_calls": 0, "unmeasured_output_calls": 0 } }, "autonomy": { "jev_attempted_cases": 199, "reasoning_assigned_cases": 0, "valid_without_reasoning": 198, "unresolved_cases": 6, "unsupported_cases": 5, "explicit_uncertainty_cases": 1 }, "types": { "correct": 263, "total": 284, "missing": 20 }, "identity": { "correct": 81, "total": 119, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "abstentions_on_definite": 0, "missing": 38 }, "registry": { "correct": 261, "total": 284 }, "relationships": { "true_positives": 13, "false_positives": 0, "false_negatives": 2 }, "evidence": { "sha256": "966ea7345286c4f9aaa83b65099ea11bdf91a7f95eee9e1c53b17ac74db32b69" } }, { "label": "jev-only-panel1-v2-repeat3", "panel": "development", "role": "candidate-repeat", "denominator": { "families": 200, "cases": 204, "unique_atoms": 275, "atom_incidences": 284, "unique_original_rows": 829, "original_row_incidences": 844 }, "full_equivalent_families": 191, "core_equivalent_families": 193, "valid_cases": 199, "known_model_cost_dollars": 0.028261716000000006, "unpriced_components": 0, "median_batch_service_seconds": 0.4857413330000004, "p90_batch_service_seconds": 0.9408943330000001, "observed_invocation_seconds": 16.427603334, "serial_component_seconds": 31.954930334000004, "model_usage": { "jev-1.13.0": { "calls": 59, "known_dollars": 0.028261716000000006, "unpriced_calls": 0, "serial_seconds": 31.954930334, "known_input_tokens": 672898, "known_output_tokens": 41652, "unmeasured_input_calls": 0, "unmeasured_output_calls": 0 } }, "autonomy": { "jev_attempted_cases": 199, "reasoning_assigned_cases": 0, "valid_without_reasoning": 199, "unresolved_cases": 5, "unsupported_cases": 5, "explicit_uncertainty_cases": 2 }, "types": { "correct": 267, "total": 284, "missing": 16 }, "identity": { "correct": 87, "total": 119, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "abstentions_on_definite": 0, "missing": 32 }, "registry": { "correct": 265, "total": 284 }, "relationships": { "true_positives": 13, "false_positives": 0, "false_negatives": 2 }, "evidence": { "sha256": "e7643e439269b7190c4abce5cc53c2b0df024a399eb0d542aa55871655052b02" } }, { "label": "jev-only-panel1-v3", "panel": "development", "role": "positive-unit-wording", "denominator": { "families": 200, "cases": 204, "unique_atoms": 275, "atom_incidences": 284, "unique_original_rows": 829, "original_row_incidences": 844 }, "full_equivalent_families": 192, "core_equivalent_families": 193, "valid_cases": 199, "known_model_cost_dollars": 0.028306740000000004, "unpriced_components": 0, "median_batch_service_seconds": 0.48504720799999995, "p90_batch_service_seconds": 1.0547283750000005, "observed_invocation_seconds": 16.589544459, "serial_component_seconds": 32.091142411, "model_usage": { "jev-1.13.0": { "calls": 59, "known_dollars": 0.028306740000000004, "unpriced_calls": 0, "serial_seconds": 32.091142411, "known_input_tokens": 673970, "known_output_tokens": 41652, "unmeasured_input_calls": 0, "unmeasured_output_calls": 0 } }, "autonomy": { "jev_attempted_cases": 199, "reasoning_assigned_cases": 0, "valid_without_reasoning": 199, "unresolved_cases": 5, "unsupported_cases": 5, "explicit_uncertainty_cases": 1 }, "types": { "correct": 267, "total": 284, "missing": 16 }, "identity": { "correct": 87, "total": 119, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "abstentions_on_definite": 0, "missing": 32 }, "registry": { "correct": 265, "total": 284 }, "relationships": { "true_positives": 13, "false_positives": 0, "false_negatives": 2 }, "evidence": { "sha256": "94cab5977c58cc0c3ac98f273f38b457b9948608847f4a4c080d690c59ff437a" } }, { "label": "jev-only-panel1-v2-tables", "panel": "development", "role": "record-tables", "denominator": { "families": 200, "cases": 204, "unique_atoms": 275, "atom_incidences": 284, "unique_original_rows": 829, "original_row_incidences": 844 }, "full_equivalent_families": 188, "core_equivalent_families": 189, "valid_cases": 194, "known_model_cost_dollars": 0.025857132, "unpriced_components": 0, "median_batch_service_seconds": 0.5308284170000006, "p90_batch_service_seconds": 1.050555417, "observed_invocation_seconds": 17.675695499999996, "serial_component_seconds": 34.69140963000001, "model_usage": { "jev-1.13.0": { "calls": 60, "known_dollars": 0.025857132, "unpriced_calls": 0, "serial_seconds": 34.69140963, "known_input_tokens": 615646, "known_output_tokens": 42033, "unmeasured_input_calls": 0, "unmeasured_output_calls": 0 } }, "autonomy": { "jev_attempted_cases": 200, "reasoning_assigned_cases": 0, "valid_without_reasoning": 194, "unresolved_cases": 10, "unsupported_cases": 4, "explicit_uncertainty_cases": 1 }, "types": { "correct": 255, "total": 284, "missing": 29 }, "identity": { "correct": 76, "total": 119, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "abstentions_on_definite": 0, "missing": 43 }, "registry": { "correct": 252, "total": 284 }, "relationships": { "true_positives": 14, "false_positives": 0, "false_negatives": 1 }, "evidence": { "sha256": "6b418517501124fba99cafcc3bbbabc2e0fe6efc929fd126e3d9353aefeedab5" } }, { "label": "jev-only-panel1-v3-tables", "panel": "development", "role": "positive-unit-plus-tables", "denominator": { "families": 200, "cases": 204, "unique_atoms": 275, "atom_incidences": 284, "unique_original_rows": 829, "original_row_incidences": 844 }, "full_equivalent_families": 189, "core_equivalent_families": 189, "valid_cases": 194, "known_model_cost_dollars": 0.025909506000000002, "unpriced_components": 0, "median_batch_service_seconds": 0.5153231669999996, "p90_batch_service_seconds": 1.1874557490000006, "observed_invocation_seconds": 17.911778667, "serial_component_seconds": 35.06936941699999, "model_usage": { "jev-1.13.0": { "calls": 60, "known_dollars": 0.025909506000000002, "unpriced_calls": 0, "serial_seconds": 35.069369417, "known_input_tokens": 616893, "known_output_tokens": 42085, "unmeasured_input_calls": 0, "unmeasured_output_calls": 0 } }, "autonomy": { "jev_attempted_cases": 200, "reasoning_assigned_cases": 0, "valid_without_reasoning": 194, "unresolved_cases": 10, "unsupported_cases": 4, "explicit_uncertainty_cases": 1 }, "types": { "correct": 254, "total": 284, "missing": 30 }, "identity": { "correct": 71, "total": 119, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "abstentions_on_definite": 3, "missing": 45 }, "registry": { "correct": 247, "total": 284 }, "relationships": { "true_positives": 14, "false_positives": 0, "false_negatives": 1 }, "evidence": { "sha256": "961564b30e70a4b2891a68c7038d3b417a47a820ad507724061d4c098b55fa39" } }, { "label": "jev-only-panel1-v2-single", "panel": "development", "role": "single-case-context", "denominator": { "families": 200, "cases": 204, "unique_atoms": 275, "atom_incidences": 284, "unique_original_rows": 829, "original_row_incidences": 844 }, "full_equivalent_families": 134, "core_equivalent_families": 184, "valid_cases": 194, "known_model_cost_dollars": 0.03302136600000001, "unpriced_components": 0, "median_batch_service_seconds": 0.5321913329999989, "p90_batch_service_seconds": 0.7239599999999992, "observed_invocation_seconds": 59.667457833, "serial_component_seconds": 118.27177081999993, "model_usage": { "jev-1.13.0": { "calls": 212, "known_dollars": 0.03302136600000001, "unpriced_calls": 0, "serial_seconds": 118.27177082, "known_input_tokens": 786223, "known_output_tokens": 42007, "unmeasured_input_calls": 0, "unmeasured_output_calls": 0 } }, "autonomy": { "jev_attempted_cases": 199, "reasoning_assigned_cases": 0, "valid_without_reasoning": 194, "unresolved_cases": 10, "unsupported_cases": 5, "explicit_uncertainty_cases": 52 }, "types": { "correct": 253, "total": 284, "missing": 28 }, "identity": { "correct": 76, "total": 119, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "abstentions_on_definite": 2, "missing": 41 }, "registry": { "correct": 253, "total": 284 }, "relationships": { "true_positives": 11, "false_positives": 0, "false_negatives": 4 }, "evidence": { "sha256": "9638c136e58e1538583aee4e75cb8b351c283a714b4171a0410e2c9af56f75d8" } }, { "label": "jev-only-panel1-v3-single", "panel": "development", "role": "positive-unit-single-case", "denominator": { "families": 200, "cases": 204, "unique_atoms": 275, "atom_incidences": 284, "unique_original_rows": 829, "original_row_incidences": 844 }, "full_equivalent_families": 172, "core_equivalent_families": 185, "valid_cases": 197, "known_model_cost_dollars": 0.032722956000000004, "unpriced_components": 0, "median_batch_service_seconds": 0.629979041999999, "p90_batch_service_seconds": 1.0758217080000032, "observed_invocation_seconds": 72.166796417, "serial_component_seconds": 143.27477387899992, "model_usage": { "jev-1.13.0": { "calls": 213, "known_dollars": 0.032722956000000004, "unpriced_calls": 0, "serial_seconds": 143.274773879, "known_input_tokens": 779118, "known_output_tokens": 42057, "unmeasured_input_calls": 0, "unmeasured_output_calls": 0 } }, "autonomy": { "jev_attempted_cases": 199, "reasoning_assigned_cases": 0, "valid_without_reasoning": 197, "unresolved_cases": 7, "unsupported_cases": 5, "explicit_uncertainty_cases": 13 }, "types": { "correct": 254, "total": 284, "missing": 22 }, "identity": { "correct": 80, "total": 119, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "abstentions_on_definite": 0, "missing": 39 }, "registry": { "correct": 259, "total": 284 }, "relationships": { "true_positives": 13, "false_positives": 0, "false_negatives": 2 }, "evidence": { "sha256": "59b7e38bb71ef2857f39a8d6aafab8975dc8217441df061a8fda7756a19a315d" } }, { "label": "jev-only-panel2-v1", "panel": "confirmation", "role": "baseline", "denominator": { "families": 200, "cases": 205, "unique_atoms": 279, "atom_incidences": 293, "unique_original_rows": 903, "original_row_incidences": 951 }, "full_equivalent_families": 166, "core_equivalent_families": 171, "valid_cases": 202, "known_model_cost_dollars": 0.030720984000000003, "unpriced_components": 0, "median_batch_service_seconds": 0.7343378340000017, "p90_batch_service_seconds": 1.4136482079999997, "observed_invocation_seconds": 25.855904083000002, "serial_component_seconds": 50.638831582, "model_usage": { "jev-1.13.0": { "calls": 61, "known_dollars": 0.030720984000000003, "unpriced_calls": 0, "serial_seconds": 50.63883158200001, "known_input_tokens": 731452, "known_output_tokens": 46341, "unmeasured_input_calls": 0, "unmeasured_output_calls": 0 } }, "autonomy": { "jev_attempted_cases": 205, "reasoning_assigned_cases": 0, "valid_without_reasoning": 202, "unresolved_cases": 3, "unsupported_cases": 0, "explicit_uncertainty_cases": 30 }, "types": { "correct": 281, "total": 293, "missing": 10 }, "identity": { "correct": 53, "total": 118, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "abstentions_on_definite": 53, "missing": 12 }, "registry": { "correct": 283, "total": 293 }, "relationships": { "true_positives": 0, "false_positives": 0, "false_negatives": 12 }, "evidence": { "sha256": "8ece99387b3cda97c2cd1d6978e0784c172fd52b1f688e9734fba01655bd0043" } }, { "label": "jev-only-panel2-v2", "panel": "confirmation", "role": "candidate", "denominator": { "families": 200, "cases": 205, "unique_atoms": 279, "atom_incidences": 293, "unique_original_rows": 903, "original_row_incidences": 951 }, "full_equivalent_families": 195, "core_equivalent_families": 197, "valid_cases": 203, "known_model_cost_dollars": 0.03367018200000001, "unpriced_components": 0, "median_batch_service_seconds": 0.8053542079999988, "p90_batch_service_seconds": 1.7034368330000007, "observed_invocation_seconds": 28.643177124999998, "serial_component_seconds": 56.204118208000004, "model_usage": { "jev-1.13.0": { "calls": 66, "known_dollars": 0.03367018200000001, "unpriced_calls": 0, "serial_seconds": 56.204118208000004, "known_input_tokens": 801671, "known_output_tokens": 46723, "unmeasured_input_calls": 0, "unmeasured_output_calls": 0 } }, "autonomy": { "jev_attempted_cases": 205, "reasoning_assigned_cases": 0, "valid_without_reasoning": 203, "unresolved_cases": 2, "unsupported_cases": 0, "explicit_uncertainty_cases": 2 }, "types": { "correct": 288, "total": 293, "missing": 4 }, "identity": { "correct": 116, "total": 118, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "abstentions_on_definite": 0, "missing": 2 }, "registry": { "correct": 289, "total": 293 }, "relationships": { "true_positives": 12, "false_positives": 0, "false_negatives": 0 }, "evidence": { "sha256": "62b5c65c011142a440b57070ccee81848f1cdf7b2ce7c6e08b559ff31ed92732" } }, { "label": "jev-only-panel2-v2-repeat2", "panel": "confirmation", "role": "candidate-repeat", "denominator": { "families": 200, "cases": 205, "unique_atoms": 279, "atom_incidences": 293, "unique_original_rows": 903, "original_row_incidences": 951 }, "full_equivalent_families": 194, "core_equivalent_families": 196, "valid_cases": 203, "known_model_cost_dollars": 0.034643406000000015, "unpriced_components": 0, "median_batch_service_seconds": 0.9308250000000007, "p90_batch_service_seconds": 1.849932041, "observed_invocation_seconds": 33.137635125, "serial_component_seconds": 65.267250958, "model_usage": { "jev-1.13.0": { "calls": 67, "known_dollars": 0.034643406000000015, "unpriced_calls": 0, "serial_seconds": 65.26725095799999, "known_input_tokens": 824843, "known_output_tokens": 46777, "unmeasured_input_calls": 0, "unmeasured_output_calls": 0 } }, "autonomy": { "jev_attempted_cases": 205, "reasoning_assigned_cases": 0, "valid_without_reasoning": 203, "unresolved_cases": 2, "unsupported_cases": 0, "explicit_uncertainty_cases": 2 }, "types": { "correct": 287, "total": 293, "missing": 4 }, "identity": { "correct": 116, "total": 118, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "abstentions_on_definite": 0, "missing": 2 }, "registry": { "correct": 289, "total": 293 }, "relationships": { "true_positives": 11, "false_positives": 0, "false_negatives": 1 }, "evidence": { "sha256": "1eac586e82844ab619a27f04bea30ce9797dfa29c4b0a8371a52bf267b43fd04" } }, { "label": "jev-only-panel2-v2-repeat3", "panel": "confirmation", "role": "candidate-repeat", "denominator": { "families": 200, "cases": 205, "unique_atoms": 279, "atom_incidences": 293, "unique_original_rows": 903, "original_row_incidences": 951 }, "full_equivalent_families": 195, "core_equivalent_families": 196, "valid_cases": 203, "known_model_cost_dollars": 0.03367018200000001, "unpriced_components": 0, "median_batch_service_seconds": 0.8775607499999988, "p90_batch_service_seconds": 1.6888915829999969, "observed_invocation_seconds": 31.144035457999998, "serial_component_seconds": 61.24914204399999, "model_usage": { "jev-1.13.0": { "calls": 66, "known_dollars": 0.03367018200000001, "unpriced_calls": 0, "serial_seconds": 61.249142044000024, "known_input_tokens": 801671, "known_output_tokens": 46723, "unmeasured_input_calls": 0, "unmeasured_output_calls": 0 } }, "autonomy": { "jev_attempted_cases": 205, "reasoning_assigned_cases": 0, "valid_without_reasoning": 203, "unresolved_cases": 2, "unsupported_cases": 0, "explicit_uncertainty_cases": 2 }, "types": { "correct": 287, "total": 293, "missing": 4 }, "identity": { "correct": 115, "total": 118, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "abstentions_on_definite": 1, "missing": 2 }, "registry": { "correct": 289, "total": 293 }, "relationships": { "true_positives": 12, "false_positives": 0, "false_negatives": 0 }, "evidence": { "sha256": "52723534bb0c1557f28c074c72567d44d6d03be6fc57f454a4eb1e304bd074aa" } } ], "sparse_development": { "version": "article-sparse-prefix-report-1", "at": "2026-09-18T05:52:00.824Z", "base": "jev-only-panel1-v2", "sparse_label": "jev-sparse-panel1-v2", "receipt_sha256": "5300909c88feb4f1503a5495120a549de077ce0d5292990a31c257d2366eba35", "reporter_sha256": "7227c6839fc3148ddd17210993332a269951216dd6ec30eaf269b57505327944", "status": "completed", "curve": [ { "selected_families": 0, "selected_fraction": 0, "attempted_families": 0, "unattempted_families": 0, "family_denominator": 200, "case_denominator": 204, "full_equivalent_families": 190, "core_equivalent_families": 193, "valid_families": 195, "valid_cases": 199, "repaired_families": 0, "regressed_families": 0, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "missing_identity_pairs": 32, "relationship_false_positives": 0, "relationship_false_negatives": 2, "known_model_cost_dollars": 0.028261716000000006, "unpriced_components": 0, "serial_component_seconds": 28.69933657999999, "valid_cases_without_reasoning_assignment": 199 }, { "selected_families": 1, "selected_fraction": 0.005, "attempted_families": 1, "unattempted_families": 0, "family_denominator": 200, "case_denominator": 204, "full_equivalent_families": 191, "core_equivalent_families": 194, "valid_families": 196, "valid_cases": 200, "repaired_families": 1, "regressed_families": 0, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "missing_identity_pairs": 4, "relationship_false_positives": 0, "relationship_false_negatives": 2, "known_model_cost_dollars": 0.09486046600000002, "unpriced_components": 0, "serial_component_seconds": 44.302005329999986, "valid_cases_without_reasoning_assignment": 199 }, { "selected_families": 2, "selected_fraction": 0.01, "attempted_families": 2, "unattempted_families": 0, "family_denominator": 200, "case_denominator": 204, "full_equivalent_families": 192, "core_equivalent_families": 195, "valid_families": 197, "valid_cases": 201, "repaired_families": 2, "regressed_families": 0, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "missing_identity_pairs": 3, "relationship_false_positives": 0, "relationship_false_negatives": 1, "known_model_cost_dollars": 0.11160191600000002, "unpriced_components": 0, "serial_component_seconds": 63.45557324699999, "valid_cases_without_reasoning_assignment": 199 }, { "selected_families": 5, "selected_fraction": 0.025, "attempted_families": 5, "unattempted_families": 0, "family_denominator": 200, "case_denominator": 204, "full_equivalent_families": 195, "core_equivalent_families": 198, "valid_families": 200, "valid_cases": 204, "repaired_families": 5, "regressed_families": 0, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "missing_identity_pairs": 0, "relationship_false_positives": 0, "relationship_false_negatives": 0, "known_model_cost_dollars": 0.15247181600000004, "unpriced_components": 0, "serial_component_seconds": 96.330919705, "valid_cases_without_reasoning_assignment": 199 }, { "selected_families": 10, "selected_fraction": 0.05, "attempted_families": 10, "unattempted_families": 0, "family_denominator": 200, "case_denominator": 204, "full_equivalent_families": 199, "core_equivalent_families": 199, "valid_families": 200, "valid_cases": 204, "repaired_families": 9, "regressed_families": 0, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "missing_identity_pairs": 0, "relationship_false_positives": 0, "relationship_false_negatives": 0, "known_model_cost_dollars": 0.16483716600000003, "unpriced_components": 0, "serial_component_seconds": 141.45414737100003, "valid_cases_without_reasoning_assignment": 193 } ], "measured_stage_times": { "base_invocation_seconds": 14.729534584000001, "entire_adjudication_invocation_seconds": 58.082118 }, "interpretation": "Derived prefixes reuse one real source-only call per selected family. Costs include the complete Jev base plus whole included family calls; no call is prorated. Prefixes are not separately timed deployments. Failed or invalid review answers do not erase a retained Jev answer. Every original family stays in the denominator.", "limitations": [ "Model-reference agreement, not human ground truth.", "Luna participated in reference construction; this review is source-only but not model-independent of the reference.", "Selected probabilities are ranking signals, not calibrated correctness guarantees.", "Development and failed-probe costs are accounted separately from deployment-shaped curve points." ] }, "sparse_confirmation_first_pass": { "version": "article-sparse-prefix-report-1", "at": "2026-09-18T06:01:45.445Z", "base": "jev-only-panel2-v2", "sparse_label": "jev-sparse-panel2-v2", "receipt_sha256": "50fe704d93d97ebad0c59d3f66e1a9bffb68b35bac505c31ab9cfa9b465ceb1f", "reporter_sha256": "7227c6839fc3148ddd17210993332a269951216dd6ec30eaf269b57505327944", "status": "halted", "curve": [ { "selected_families": 0, "selected_fraction": 0, "attempted_families": 0, "unattempted_families": 0, "family_denominator": 200, "case_denominator": 205, "full_equivalent_families": 195, "core_equivalent_families": 197, "valid_families": 198, "valid_cases": 203, "repaired_families": 0, "regressed_families": 0, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "missing_identity_pairs": 2, "relationship_false_positives": 0, "relationship_false_negatives": 0, "known_model_cost_dollars": 0.03367018200000001, "unpriced_components": 0, "serial_component_seconds": 56.204118208000004, "valid_cases_without_reasoning_assignment": 203 }, { "selected_families": 1, "selected_fraction": 0.005, "attempted_families": 1, "unattempted_families": 0, "family_denominator": 200, "case_denominator": 205, "full_equivalent_families": 196, "core_equivalent_families": 198, "valid_families": 199, "valid_cases": 204, "repaired_families": 1, "regressed_families": 0, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "missing_identity_pairs": 1, "relationship_false_positives": 0, "relationship_false_negatives": 0, "known_model_cost_dollars": 0.03617133200000001, "unpriced_components": 0, "serial_component_seconds": 69.311639458, "valid_cases_without_reasoning_assignment": 203 }, { "selected_families": 2, "selected_fraction": 0.01, "attempted_families": 2, "unattempted_families": 0, "family_denominator": 200, "case_denominator": 205, "full_equivalent_families": 197, "core_equivalent_families": 199, "valid_families": 200, "valid_cases": 205, "repaired_families": 2, "regressed_families": 0, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "missing_identity_pairs": 0, "relationship_false_positives": 0, "relationship_false_negatives": 0, "known_model_cost_dollars": 0.03792738200000001, "unpriced_components": 0, "serial_component_seconds": 80.02735737400002, "valid_cases_without_reasoning_assignment": 203 }, { "selected_families": 5, "selected_fraction": 0.025, "attempted_families": 5, "unattempted_families": 0, "family_denominator": 200, "case_denominator": 205, "full_equivalent_families": 199, "core_equivalent_families": 199, "valid_families": 200, "valid_cases": 205, "repaired_families": 4, "regressed_families": 0, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "missing_identity_pairs": 0, "relationship_false_positives": 0, "relationship_false_negatives": 0, "known_model_cost_dollars": 0.043460382000000006, "unpriced_components": 0, "serial_component_seconds": 114.89278808200001, "valid_cases_without_reasoning_assignment": 200 }, { "selected_families": 10, "selected_fraction": 0.05, "attempted_families": 10, "unattempted_families": 0, "family_denominator": 200, "case_denominator": 205, "full_equivalent_families": 199, "core_equivalent_families": 199, "valid_families": 200, "valid_cases": 205, "repaired_families": 5, "regressed_families": 1, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "missing_identity_pairs": 0, "relationship_false_positives": 0, "relationship_false_negatives": 0, "known_model_cost_dollars": 0.05733363200000001, "unpriced_components": 0, "serial_component_seconds": 177.18173991499998, "valid_cases_without_reasoning_assignment": 194 } ], "measured_stage_times": { "base_invocation_seconds": 28.643177124999998, "entire_adjudication_invocation_seconds": 63.855605708000006 }, "interpretation": "Derived prefixes reuse one real source-only call per selected family. Costs include the complete Jev base plus whole included family calls; no call is prorated. Prefixes are not separately timed deployments. Failed or invalid review answers do not erase a retained Jev answer. Every original family stays in the denominator.", "limitations": [ "Model-reference agreement, not human ground truth.", "Luna participated in reference construction; this review is source-only but not model-independent of the reference.", "Selected probabilities are ranking signals, not calibrated correctness guarantees.", "Development and failed-probe costs are accounted separately from deployment-shaped curve points." ] }, "sparse_confirmation_after_one_format_recovery": { "selected_families": 10, "selected_fraction": 0.05, "reasoning_calls": 11, "family_denominator": 200, "case_denominator": 205, "full_equivalent_families": 199, "core_equivalent_families": 199, "valid_families": 200, "valid_cases": 205, "repaired_families": 5, "regressed_families": 1, "known_model_cost_dollars": 0.06105213200000001, "unpriced_components": 0, "false_merges": 0, "false_separations": 0, "abstentions_on_definite": 2, "missing_identity_pairs": 0, "relationship_false_positives": 0, "relationship_false_negatives": 0, "recovery_invocation_seconds": 11.388211833000002, "sum_of_measured_stage_wall_seconds": 103.886994666, "recovery_note": "One explicit same-source/same-prompt attempt for the rank-9 malformed case identifier. The original failed call remains charged. No reference outcome selected the recovery, and no model answer was hand-repaired." }, "confirmation_controls": { "evidence": { "sha256": "5c5f517815b2dd6031b1b7ef3612f27564524af745b663e5994536cee43afc99" }, "denominator": { "families": 200, "cases": 205, "unique_atoms": 279, "atom_incidences": 293, "unique_original_rows": 903, "original_row_incidences": 951 }, "arms": { "fable": { "cases": 205, "finished_batches": 52, "planned_batches": 52, "valid_cases": 205, "consumer_equivalent_cases": 205, "core_equivalent_cases": 205, "families": 200, "consumer_equivalent_families": 200, "core_equivalent_families": 200, "types": { "correct": 293, "total": 293, "missing": 0 }, "identity": { "correct": 118, "total": 118, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "abstentions_on_definite": 0, "missing": 0 }, "registry": { "correct": 293, "total": 293 }, "relationships": { "true_positives": 12, "false_positives": 0, "false_negatives": 0 }, "known_model_cost_dollars": 9.831786999999998, "unpriced_components": 0, "model_usage": { "anthropic/claude-fable-5-1": { "calls": 52, "known_dollars": 9.831786999999998, "unpriced_calls": 0, "serial_seconds": 844.6439144199999, "known_input_tokens": 858617, "known_output_tokens": 49729, "unmeasured_input_calls": 0, "unmeasured_output_calls": 0 } }, "component_serial_seconds": 844.6439144200001, "batch_component_seconds": { "median": 14.608669083999994, "p90": 22.203791208000002 }, "observed_invocation_wall_seconds": null, "jev_only_cases": 0, "strata": { "name_form_conflict": { "families": 25, "consumer_equivalent": 25, "core_equivalent": 25 }, "namespace:recipient_registry": { "families": 70, "consumer_equivalent": 70, "core_equivalent": 70 }, "ordinary": { "families": 173, "consumer_equivalent": 173, "core_equivalent": 173 }, "person_match": { "families": 48, "consumer_equivalent": 48, "core_equivalent": 48 }, "person_separation": { "families": 12, "consumer_equivalent": 12, "core_equivalent": 12 }, "relation:same_household": { "families": 12, "consumer_equivalent": 12, "core_equivalent": 12 }, "rows:1": { "families": 112, "consumer_equivalent": 112, "core_equivalent": 112 }, "rows:2-5": { "families": 60, "consumer_equivalent": 60, "core_equivalent": 60 }, "rows:21-100": { "families": 13, "consumer_equivalent": 13, "core_equivalent": 13 }, "rows:6-20": { "families": 15, "consumer_equivalent": 15, "core_equivalent": 15 }, "suffix_conflict": { "families": 6, "consumer_equivalent": 6, "core_equivalent": 6 }, "tier:0": { "families": 200, "consumer_equivalent": 200, "core_equivalent": 200 }, "type:organization": { "families": 6, "consumer_equivalent": 6, "core_equivalent": 6 }, "type:person": { "families": 194, "consumer_equivalent": 194, "core_equivalent": 194 } }, "autonomy": { "jev_attempted_cases": 0, "reasoning_assigned_cases": 205, "valid_without_reasoning": 0, "unresolved_cases": 0, "unsupported_cases": 0, "explicit_uncertainty_cases": 0 }, "attempted_batch_component_seconds": { "median": 14.608669083999994, "p90": 22.203791208000002 } }, "cheap": { "cases": 205, "finished_batches": 52, "planned_batches": 52, "valid_cases": 205, "consumer_equivalent_cases": 204, "core_equivalent_cases": 205, "families": 200, "consumer_equivalent_families": 199, "core_equivalent_families": 200, "types": { "correct": 293, "total": 293, "missing": 0 }, "identity": { "correct": 118, "total": 118, "false_merges": 0, "false_separations": 0, "unsupported_merges_over_unknown": 0, "abstentions_on_definite": 0, "missing": 0 }, "registry": { "correct": 293, "total": 293 }, "relationships": { "true_positives": 11, "false_positives": 0, "false_negatives": 1 }, "known_model_cost_dollars": 0.30124655, "unpriced_components": 8, "model_usage": { "pi/openai/gpt-5.6-luna": { "calls": 52, "known_dollars": 0.16846504999999995, "unpriced_calls": 8, "serial_seconds": 608.4847014639998, "known_input_tokens": 496709, "known_output_tokens": 36912, "unmeasured_input_calls": 8, "unmeasured_output_calls": 8 }, "pi/google/gemini-3.7-flash": { "calls": 10, "known_dollars": 0.1327815, "unpriced_calls": 0, "serial_seconds": 106.59963216599994, "known_input_tokens": 92507, "known_output_tokens": 16907, "unmeasured_input_calls": 0, "unmeasured_output_calls": 0 } }, "component_serial_seconds": 715.0843336300001, "batch_component_seconds": { "median": 13.267582458999998, "p90": 16.756536583 }, "observed_invocation_wall_seconds": null, "jev_only_cases": 0, "strata": { "name_form_conflict": { "families": 25, "consumer_equivalent": 24, "core_equivalent": 25 }, "namespace:recipient_registry": { "families": 70, "consumer_equivalent": 70, "core_equivalent": 70 }, "ordinary": { "families": 173, "consumer_equivalent": 173, "core_equivalent": 173 }, "person_match": { "families": 48, "consumer_equivalent": 47, "core_equivalent": 48 }, "person_separation": { "families": 12, "consumer_equivalent": 11, "core_equivalent": 12 }, "relation:same_household": { "families": 12, "consumer_equivalent": 11, "core_equivalent": 12 }, "rows:1": { "families": 112, "consumer_equivalent": 112, "core_equivalent": 112 }, "rows:2-5": { "families": 60, "consumer_equivalent": 59, "core_equivalent": 60 }, "rows:21-100": { "families": 13, "consumer_equivalent": 13, "core_equivalent": 13 }, "rows:6-20": { "families": 15, "consumer_equivalent": 15, "core_equivalent": 15 }, "suffix_conflict": { "families": 6, "consumer_equivalent": 6, "core_equivalent": 6 }, "tier:0": { "families": 200, "consumer_equivalent": 199, "core_equivalent": 200 }, "type:organization": { "families": 6, "consumer_equivalent": 6, "core_equivalent": 6 }, "type:person": { "families": 194, "consumer_equivalent": 193, "core_equivalent": 194 } }, "autonomy": { "jev_attempted_cases": 0, "reasoning_assigned_cases": 205, "valid_without_reasoning": 0, "unresolved_cases": 0, "unsupported_cases": 0, "explicit_uncertainty_cases": 0 }, "attempted_batch_component_seconds": { "median": 13.267582458999998, "p90": 16.756536583 } }, "jev-first": { "cases": 205, "finished_batches": 52, "planned_batches": 52, "valid_cases": 205, "consumer_equivalent_cases": 203, "core_equivalent_cases": 203, "families": 200, "consumer_equivalent_families": 198, "core_equivalent_families": 198, "types": { "correct": 292, "total": 293, "missing": 0 }, "identity": { "correct": 117, "total": 118, "false_merges": 1, "false_separations": 0, "unsupported_merges_over_unknown": 0, "abstentions_on_definite": 0, "missing": 0 }, "registry": { "correct": 293, "total": 293 }, "relationships": { "true_positives": 11, "false_positives": 0, "false_negatives": 1 }, "known_model_cost_dollars": 0.11237322800000002, "unpriced_components": 4, "model_usage": { "jev-1.13.0": { "calls": 65, "known_dollars": 0.03204112800000001, "unpriced_calls": 1, "serial_seconds": 47.351881168000006, "known_input_tokens": 762884, "known_output_tokens": 46560, "unmeasured_input_calls": 1, "unmeasured_output_calls": 1 }, "pi/openai/gpt-5.6-luna": { "calls": 31, "known_dollars": 0.0479531, "unpriced_calls": 3, "serial_seconds": 209.35622162500007, "known_input_tokens": 179906, "known_output_tokens": 2484, "unmeasured_input_calls": 3, "unmeasured_output_calls": 3 }, "pi/google/gemini-3.7-flash": { "calls": 3, "known_dollars": 0.032379000000000005, "unpriced_calls": 0, "serial_seconds": 33.509490666, "known_input_tokens": 20837, "known_output_tokens": 4467, "unmeasured_input_calls": 0, "unmeasured_output_calls": 0 } }, "component_serial_seconds": 290.21759345900006, "batch_component_seconds": { "median": 6.360088374000013, "p90": 10.756573166000017 }, "observed_invocation_wall_seconds": null, "jev_only_cases": 165, "strata": { "name_form_conflict": { "families": 25, "consumer_equivalent": 24, "core_equivalent": 24 }, "namespace:recipient_registry": { "families": 70, "consumer_equivalent": 69, "core_equivalent": 69 }, "ordinary": { "families": 173, "consumer_equivalent": 173, "core_equivalent": 173 }, "person_match": { "families": 48, "consumer_equivalent": 48, "core_equivalent": 48 }, "person_separation": { "families": 12, "consumer_equivalent": 11, "core_equivalent": 11 }, "relation:same_household": { "families": 12, "consumer_equivalent": 11, "core_equivalent": 11 }, "rows:1": { "families": 112, "consumer_equivalent": 111, "core_equivalent": 111 }, "rows:2-5": { "families": 60, "consumer_equivalent": 59, "core_equivalent": 59 }, "rows:21-100": { "families": 13, "consumer_equivalent": 13, "core_equivalent": 13 }, "rows:6-20": { "families": 15, "consumer_equivalent": 15, "core_equivalent": 15 }, "suffix_conflict": { "families": 6, "consumer_equivalent": 6, "core_equivalent": 6 }, "tier:0": { "families": 200, "consumer_equivalent": 198, "core_equivalent": 198 }, "type:organization": { "families": 6, "consumer_equivalent": 5, "core_equivalent": 5 }, "type:person": { "families": 194, "consumer_equivalent": 193, "core_equivalent": 193 } }, "autonomy": { "jev_attempted_cases": 205, "reasoning_assigned_cases": 40, "valid_without_reasoning": 165, "unresolved_cases": 0, "unsupported_cases": 0, "explicit_uncertainty_cases": 0 }, "attempted_batch_component_seconds": { "median": 6.360088374000013, "p90": 10.756573166000017 } } }, "note": "Retained completed controls; no new Fable call. Controls ran in a different execution wave. Cheap and earlier hybrid costs have unpriced components and are not complete bills." }, "scale_scenarios": { "basis": "Linear token/call-cost arithmetic for one million preassembled families with exactly this panel's task mix. Not raw rows, unique people, petabytes, a load test or a provider-capacity claim. Pure Jev retains its observed unresolved fraction. Infrastructure, retrieval, candidate generation, reference construction and authoring are excluded.", "points": [ { "name": "corrected_jev_only", "observed_full_agreement": 0.975, "observed_unresolved_cases": 2, "known_dollars_per_million_preassembled_families": 168.35091000000003, "fable_known_cost_ratio": 292.0027875109198 }, { "name": "jev_plus_five_of_200_reviews", "observed_full_agreement": 0.995, "observed_unresolved_cases": 0, "known_dollars_per_million_preassembled_families": 217.30191000000005, "fable_known_cost_ratio": 226.22412752837738 }, { "name": "jev_plus_ten_of_200_reviews_with_recovery", "observed_full_agreement": 0.995, "observed_unresolved_cases": 0, "known_dollars_per_million_preassembled_families": 305.26066000000003, "fable_known_cost_ratio": 161.03920826221102 }, { "name": "fable_control", "observed_full_agreement": 1, "observed_unresolved_cases": 0, "known_dollars_per_million_preassembled_families": 49158.93499999999, "fable_known_cost_ratio": 1 } ] }, "accounting": { "research": { "phase_id": "er-jev-article-jev-only-1", "ceiling_dollars": 10, "recorded_model_estimates_dollars": 0.5644491140000005, "unpriced_commitments_dollars": 0.426862, "protected_reserves_dollars": 0.5, "hold": { "reason": "Article comparison completed; no further paid inference authorized by this phase", "at": "2026-09-18T06:05:44.970Z" }, "config": { "sha256": "926d86292988a2e6165055b9c4f78258a02cc3666a67b7a7d1ab2034ea0ce637" }, "state": { "sha256": "1e2263a01a3b0fd0cd9915585f8ca90a0fba199de82fad33eada32b4c8ea2bff" } }, "total_authorized_dollars": 85, "all_study_recorded_model_estimates_dollars": 31.985710472, "all_study_unpriced_commitments_dollars": 9.583791, "all_phases_sealed": true, "note": "The $10 research ceiling was an internal allocation, not new authorization. Four failed Jev transport/capacity probes retain unknown charges at their reservations. The priced malformed Luna attempt and its explicit recovery are included." }, "verification": { "version": "jev-research-input-verification-1", "at": "2026-09-18T06:13:53.810Z", "pass": true, "full_runs": 13, "jev_requests_reconstructed": 1098, "questions_reconstructed": 8607, "own_jev_proposals_reconstructed": 171, "source_only_native_requests": 21, "native_artifact_files_checked": 1134, "contract_and_scorer_unchanged": true, "candidate_unchanged_since_freeze": true, "recovery_changed_only_attempt_identity": true, "all_phases_sealed": true, "verifier_sha256": "2083f2fdd3f24c58b6ddc024447f0419d88f4fc50befad525facbf3dd0ad58a7", "scope": "Exact source/question reconstruction and retained artifact checks, not human correctness or provider-weight independence.", "evidence": "runs/jev-only-study-1/input-verification.json", "manifest": "runs/jev-only-study-1/report-inputs.json", "assembler_sha256": "450d2a9e0d22535c4888352ad8308b2152c1071235b9d23d95463d6b949f2adf" }, "findings": [ "Corrected batched Jev generalized from the exposed panel to the previously uninspected confirmation outcomes.", "Additional wording was within the batched repeat range; reversible tables and single-case context regressed and were not selected.", "The larger intact source exceeded the provider token limit both in a combined batch and alone; no source was truncated.", "More reasoning was not monotonic: the ten-family confirmation prefix repaired all five initial family misses but also regressed a previously matching family before recovery.", "The budgeted review selector used only output validity, explicit uncertainty and selected probabilities; source-only reviewers received neither Jev proposals nor reference answers." ], "limitations": [ "Agreement is not human accuracy. Luna participated in reference construction, so its source-only reviews are not model-independent of that reference.", "Each panel has 200 families. Panel 2 has no fresh committee, joint, placeholder, registry-mapping or filer strata; it mainly confirms typing, identity and households.", "Three corrected repetitions quantify observed variation, not a population confidence guarantee. Sparse prefixes share real whole-family calls and are not separately timed deployments.", "Jev-only runs retain unresolved outputs, including local capacity exclusions; unsupported cases are not silently dropped.", "Probabilities were used as ranking signals, not calibrated probabilities of correctness.", "Service clocks include harness/accounting overhead; different execution waves are not a controlled provider-latency comparison.", "The experiment demonstrates direct Jev calls plus Hankweave-operated reasoning jobs, not an exclusive requirement for Hankweave or demonstrated petabyte throughput." ] }