{
	"method": "Paired case-cluster percentile bootstrap",
	"replicates": 1000000,
	"seed": 20260514,
	"rng": "NumPy PCG64 multinomial case multiplicities",
	"numpy_version": "2.5.3",
	"family_model_count": 4,
	"available_models": [
		"gpt-5-6-luna-summaries-high",
		"gpt-6-luna-summaries-none",
		"jev-grok-short-summaries",
		"jev-luna-summaries"
	],
	"case_count": 91,
	"unit_count": 387,
	"familywise_alpha": 0.05,
	"interval_confidence": 0.9972222222222222,
	"correction": "Bonferroni across 6 model pairs and three metrics",
	"metrics": {
		"gpt-5-6-luna-summaries-high": {
			"micro_brier": 0.1548945736434109,
			"equal_case_brier": 0.1611943673786531,
			"accuracy": 0.7855297157622739
		},
		"gpt-6-luna-summaries-none": {
			"micro_brier": 0.17152093023255813,
			"equal_case_brier": 0.186942624439053,
			"accuracy": 0.7416020671834626
		},
		"jev-grok-short-summaries": {
			"micro_brier": 0.32725193798449614,
			"equal_case_brier": 0.30504730523444806,
			"accuracy": 0.38242894056847543
		},
		"jev-luna-summaries": {
			"micro_brier": 0.3675413436692507,
			"equal_case_brier": 0.3441620877931591,
			"accuracy": 0.32558139534883723
		}
	},
	"pairs": [
		{
			"model_a": "gpt-5-6-luna-summaries-high",
			"model_b": "gpt-6-luna-summaries-none",
			"metric": "micro_brier",
			"delta": -0.01662635658914724,
			"ci_low": -0.048416712907243656,
			"ci_high": 0.013932617431045118,
			"significant": false
		},
		{
			"model_a": "gpt-5-6-luna-summaries-high",
			"model_b": "gpt-6-luna-summaries-none",
			"metric": "equal_case_brier",
			"delta": -0.025748257060399904,
			"ci_low": -0.06994084137737264,
			"ci_high": 0.008839401833286135,
			"significant": false
		},
		{
			"model_a": "gpt-5-6-luna-summaries-high",
			"model_b": "gpt-6-luna-summaries-none",
			"metric": "accuracy",
			"delta": 0.043927648578811374,
			"ci_low": -0.017587939698492483,
			"ci_high": 0.11331444759206799,
			"significant": false
		},
		{
			"model_a": "gpt-5-6-luna-summaries-high",
			"model_b": "jev-grok-short-summaries",
			"metric": "micro_brier",
			"delta": -0.17235736434108526,
			"ci_low": -0.28375696379009213,
			"ci_high": -0.027008478596244282,
			"significant": true
		},
		{
			"model_a": "gpt-5-6-luna-summaries-high",
			"model_b": "jev-grok-short-summaries",
			"metric": "equal_case_brier",
			"delta": -0.14385293785579498,
			"ci_low": -0.22247598810366218,
			"ci_high": -0.05793509924153209,
			"significant": true
		},
		{
			"model_a": "gpt-5-6-luna-summaries-high",
			"model_b": "jev-grok-short-summaries",
			"metric": "accuracy",
			"delta": 0.4031007751937985,
			"ci_low": 0.15088352481319917,
			"ci_high": 0.6111111111111112,
			"significant": true
		},
		{
			"model_a": "gpt-5-6-luna-summaries-high",
			"model_b": "jev-luna-summaries",
			"metric": "micro_brier",
			"delta": -0.2126467700258398,
			"ci_low": -0.33361214037061654,
			"ci_high": -0.051356669052162754,
			"significant": true
		},
		{
			"model_a": "gpt-5-6-luna-summaries-high",
			"model_b": "jev-luna-summaries",
			"metric": "equal_case_brier",
			"delta": -0.182967720414506,
			"ci_low": -0.27147164665344575,
			"ci_high": -0.08512657099448162,
			"significant": true
		},
		{
			"model_a": "gpt-5-6-luna-summaries-high",
			"model_b": "jev-luna-summaries",
			"metric": "accuracy",
			"delta": 0.4599483204134367,
			"ci_low": 0.19047619047619047,
			"ci_high": 0.6761363636363636,
			"significant": true
		},
		{
			"model_a": "gpt-6-luna-summaries-none",
			"model_b": "jev-grok-short-summaries",
			"metric": "micro_brier",
			"delta": -0.15573100775193802,
			"ci_low": -0.2626595636256951,
			"ci_high": -0.024311956314953982,
			"significant": true
		},
		{
			"model_a": "gpt-6-luna-summaries-none",
			"model_b": "jev-grok-short-summaries",
			"metric": "equal_case_brier",
			"delta": -0.11810468079539507,
			"ci_low": -0.20174294296179998,
			"ci_high": -0.025781551936158543,
			"significant": true
		},
		{
			"model_a": "gpt-6-luna-summaries-none",
			"model_b": "jev-grok-short-summaries",
			"metric": "accuracy",
			"delta": 0.3591731266149871,
			"ci_low": 0.10511363636363645,
			"ci_high": 0.5745257452574526,
			"significant": true
		},
		{
			"model_a": "gpt-6-luna-summaries-none",
			"model_b": "jev-luna-summaries",
			"metric": "micro_brier",
			"delta": -0.19602041343669255,
			"ci_low": -0.31382497236551216,
			"ci_high": -0.048226181372723644,
			"significant": true
		},
		{
			"model_a": "gpt-6-luna-summaries-none",
			"model_b": "jev-luna-summaries",
			"metric": "equal_case_brier",
			"delta": -0.1572194633541061,
			"ci_low": -0.2508599714100385,
			"ci_high": -0.05372736262695618,
			"significant": true
		},
		{
			"model_a": "gpt-6-luna-summaries-none",
			"model_b": "jev-luna-summaries",
			"metric": "accuracy",
			"delta": 0.4160206718346253,
			"ci_low": 0.14562806732102843,
			"ci_high": 0.6412213740458015,
			"significant": true
		},
		{
			"model_a": "jev-grok-short-summaries",
			"model_b": "jev-luna-summaries",
			"metric": "micro_brier",
			"delta": -0.04028940568475453,
			"ci_low": -0.07056180669490858,
			"ci_high": -0.009809152712377481,
			"significant": true
		},
		{
			"model_a": "jev-grok-short-summaries",
			"model_b": "jev-luna-summaries",
			"metric": "equal_case_brier",
			"delta": -0.039114782558711036,
			"ci_low": -0.06586134701211464,
			"ci_high": -0.013505275200989401,
			"significant": true
		},
		{
			"model_a": "jev-grok-short-summaries",
			"model_b": "jev-luna-summaries",
			"metric": "accuracy",
			"delta": 0.056847545219638196,
			"ci_low": 0.0023094688221709236,
			"ci_high": 0.12806539509536785,
			"significant": true
		}
	],
	"significant_pairs": [
		{
			"better": "gpt-5-6-luna-summaries-high",
			"worse": "jev-grok-short-summaries",
			"metric": "micro_brier"
		},
		{
			"better": "gpt-5-6-luna-summaries-high",
			"worse": "jev-grok-short-summaries",
			"metric": "equal_case_brier"
		},
		{
			"better": "gpt-5-6-luna-summaries-high",
			"worse": "jev-grok-short-summaries",
			"metric": "accuracy"
		},
		{
			"better": "gpt-5-6-luna-summaries-high",
			"worse": "jev-luna-summaries",
			"metric": "micro_brier"
		},
		{
			"better": "gpt-5-6-luna-summaries-high",
			"worse": "jev-luna-summaries",
			"metric": "equal_case_brier"
		},
		{
			"better": "gpt-5-6-luna-summaries-high",
			"worse": "jev-luna-summaries",
			"metric": "accuracy"
		},
		{
			"better": "gpt-6-luna-summaries-none",
			"worse": "jev-grok-short-summaries",
			"metric": "micro_brier"
		},
		{
			"better": "gpt-6-luna-summaries-none",
			"worse": "jev-grok-short-summaries",
			"metric": "equal_case_brier"
		},
		{
			"better": "gpt-6-luna-summaries-none",
			"worse": "jev-grok-short-summaries",
			"metric": "accuracy"
		},
		{
			"better": "gpt-6-luna-summaries-none",
			"worse": "jev-luna-summaries",
			"metric": "micro_brier"
		},
		{
			"better": "gpt-6-luna-summaries-none",
			"worse": "jev-luna-summaries",
			"metric": "equal_case_brier"
		},
		{
			"better": "gpt-6-luna-summaries-none",
			"worse": "jev-luna-summaries",
			"metric": "accuracy"
		},
		{
			"better": "jev-grok-short-summaries",
			"worse": "jev-luna-summaries",
			"metric": "micro_brier"
		},
		{
			"better": "jev-grok-short-summaries",
			"worse": "jev-luna-summaries",
			"metric": "equal_case_brier"
		},
		{
			"better": "jev-grok-short-summaries",
			"worse": "jev-luna-summaries",
			"metric": "accuracy"
		}
	],
	"caveat": "Exploratory. Cross-case dependence has not been verified. Unavailable pairs are not tested.",
	"sources": {
		"jev-luna-summaries": "jev-luna-summaries.jsonl",
		"gpt-6-luna-summaries-none": "gpt-6-luna-summaries-none.jsonl",
		"gpt-5-6-luna-summaries-high": "gpt-5-6-luna-summaries-high.jsonl",
		"jev-grok-short-summaries": "jev-grok-short-summaries.jsonl"
	},
	"provenance": {
		"release": "cycle-1-91-2026-09-08-luna-r5",
		"release_digest": "4311a6ee59c3ae1fcb392f1dd768ab00d1d7a7c7ea135271f3e117746233ed8a",
		"models": {
			"jev-luna-summaries": {
				"forecast_run": "35459485218",
				"source": "summary_export_units"
			},
			"gpt-6-luna-summaries-none": {
				"forecast_run": "35799226577",
				"source": "summary_export_units"
			},
			"gpt-5-6-luna-summaries-high": {
				"forecast_run": "35773842374",
				"source": "summary_export_units"
			},
			"jev-grok-short-summaries": {
				"forecast_run": "37167631221",
				"scoring_run": "37174741889",
				"source": "protected_summary_export_units",
				"summary_packet_group": "grok-short"
			}
		},
		"comparison_scope": "Four summary conditions on the same outcomes; three controls share Luna packets, while Grok shorter summaries use a different representation. Cross-pipeline differences cannot isolate model ability or summarization effects."
	},
	"missing_models": []
}