{
  "schema_version": 1,
  "columns": [
    "family",
    "baseline",
    "comparison",
    "metric",
    "method",
    "direction",
    "estimate",
    "ci_low",
    "ci_high",
    "valid_resamples",
    "interval_status"
  ],
  "rows": [
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "species_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.012658227848101278",
      "ci_low": "-0.00877192982456143",
      "ci_high": "0.03417721518987343",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "species_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.012658227848101278",
      "ci_low": "-0.00783987022119929",
      "ci_high": "0.03241408205015801",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "species_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.012658227848101278",
      "ci_low": "-0.007626582278481023",
      "ci_high": "0.03417721518987338",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "adjusted_tis_pct",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.3037974683544249",
      "ci_low": "-3.8792496269794223",
      "ci_high": "3.329113924050631",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "adjusted_tis_pct",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.3037974683544249",
      "ci_low": "-3.725911351306931",
      "ci_high": "3.010371801194593",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "adjusted_tis_pct",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.3037974683544249",
      "ci_low": "-3.9117088607594916",
      "ci_high": "3.278481012658233",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.004806800364395014",
      "ci_low": "-0.01378897597653663",
      "ci_high": "0.021662893761294676",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.004806800364395014",
      "ci_low": "-0.012404835891129447",
      "ci_high": "0.02122589716692596",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.004806800364395014",
      "ci_low": "-0.013126951880110737",
      "ci_high": "0.021937357924103965",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "reference_species_macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.011379084786033555",
      "ci_low": "-0.013880593385191285",
      "ci_high": "0.03392231001240996",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "reference_species_macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.011379084786033555",
      "ci_low": "-0.01194277716340649",
      "ci_high": "0.033364744391718694",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "reference_species_macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.011379084786033555",
      "ci_low": "-0.012894850696916193",
      "ci_high": "0.03435839099255477",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "species_answer_coverage",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.017721518987341756",
      "ci_low": "-0.008860759493670933",
      "ci_high": "0.04435994930291509",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "species_answer_coverage",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.017721518987341756",
      "ci_low": "-0.00791213708935232",
      "ci_high": "0.042435100061682364",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "species_answer_coverage",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.017721518987341756",
      "ci_low": "-0.007594936708860689",
      "ci_high": "0.04303797468354431",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "species_attempted_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.000803105340650534",
      "ci_low": "-0.03253330341619815",
      "ci_high": "0.030750184495774338",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "species_attempted_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.000803105340650534",
      "ci_low": "-0.030400361257877376",
      "ci_high": "0.028524339543464",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite LOW",
      "metric": "species_attempted_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.000803105340650534",
      "ci_low": "-0.0319847623664386",
      "ci_high": "0.029835713839310433",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.051898734177215244",
      "ci_low": "0.026480881199157588",
      "ci_high": "0.07769423558897243",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.051898734177215244",
      "ci_low": "0.028137200352390267",
      "ci_high": "0.07481589173836013",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.051898734177215244",
      "ci_low": "0.027848101265822767",
      "ci_high": "0.07721518987341769",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "adjusted_tis_pct",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "2.329113924050638",
      "ci_low": "-1.6666666666666714",
      "ci_high": "6.392441484023435",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "adjusted_tis_pct",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "2.329113924050638",
      "ci_low": "-1.3920515960212527",
      "ci_high": "5.739324134023497",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "adjusted_tis_pct",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "2.329113924050638",
      "ci_low": "-1.6202531645569636",
      "ci_high": "6.215189873417721",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.009482964976156527",
      "ci_low": "-0.010511864318224626",
      "ci_high": "0.02982738919826943",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.009482964976156527",
      "ci_low": "-0.008733982576903021",
      "ci_high": "0.02858790227129822",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.009482964976156527",
      "ci_low": "-0.009728150353715779",
      "ci_high": "0.029364932967896735",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "reference_species_macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.04265585356921331",
      "ci_low": "0.014482702267935355",
      "ci_high": "0.07005150518801322",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "reference_species_macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.04265585356921331",
      "ci_low": "0.017169947876953216",
      "ci_high": "0.06869691537347018",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "reference_species_macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.04265585356921331",
      "ci_low": "0.015664739381912672",
      "ci_high": "0.06969406420266008",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_answer_coverage",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.07974683544303796",
      "ci_low": "0.04911781839246544",
      "ci_high": "0.1106952289388294",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_answer_coverage",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.07974683544303796",
      "ci_low": "0.052454744622466125",
      "ci_high": "0.10789243544781503",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_answer_coverage",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.07974683544303796",
      "ci_low": "0.050632911392405",
      "ci_high": "0.11012658227848093",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_attempted_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.010628497468499032",
      "ci_low": "-0.04781228850576277",
      "ci_high": "0.02645755615870929",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_attempted_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.010628497468499032",
      "ci_low": "-0.045906501140475404",
      "ci_high": "0.0226457929411273",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_attempted_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.010628497468499032",
      "ci_low": "-0.047596600286473695",
      "ci_high": "0.02566760797031895",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.06075949367088607",
      "ci_low": "0.03608215366466634",
      "ci_high": "0.08646736115096904",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.06075949367088607",
      "ci_low": "0.03683144738524483",
      "ci_high": "0.08306253872709561",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.06075949367088607",
      "ci_low": "0.036708860759493644",
      "ci_high": "0.08607594936708857",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "adjusted_tis_pct",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.29113924050633244",
      "ci_low": "-4.109368192623395",
      "ci_high": "4.707421947294621",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "adjusted_tis_pct",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.29113924050633244",
      "ci_low": "-3.548897085685058",
      "ci_high": "4.049201668584569",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "adjusted_tis_pct",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.29113924050633244",
      "ci_low": "-4.063291139240505",
      "ci_high": "4.6835443037974684",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.007422781117155941",
      "ci_low": "-0.026861111224436497",
      "ci_high": "0.012892294383652787",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.007422781117155941",
      "ci_low": "-0.02605617401234015",
      "ci_high": "0.011622484417700276",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.007422781117155941",
      "ci_low": "-0.026667639307790165",
      "ci_high": "0.012886614637923692",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "reference_species_macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.05211075715748725",
      "ci_low": "0.022912816046097725",
      "ci_high": "0.07958342234899679",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "reference_species_macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.05211075715748725",
      "ci_low": "0.024810853272343728",
      "ci_high": "0.0781534049345496",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "reference_species_macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.05211075715748725",
      "ci_low": "0.023722889460872993",
      "ci_high": "0.07954903383581831",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_answer_coverage",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.12151898734177213",
      "ci_low": "0.09299291041193225",
      "ci_high": "0.15063291139240498",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_answer_coverage",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.12151898734177213",
      "ci_low": "0.09461979651853054",
      "ci_high": "0.14739258016947868",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_answer_coverage",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.12151898734177213",
      "ci_low": "0.09367088607594931",
      "ci_high": "0.14936708860759496",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_attempted_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.04031644388289857",
      "ci_low": "-0.0778384953375908",
      "ci_high": "-0.002785248499594991",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_attempted_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.04031644388289857",
      "ci_low": "-0.07623728748304254",
      "ci_high": "-0.005912774780249939",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MINIMAL corrected",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_attempted_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.04031644388289857",
      "ci_low": "-0.07777808613759372",
      "ci_high": "-0.002230292456585583",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.03924050632911397",
      "ci_low": "0.016310655874315617",
      "ci_high": "0.062418725617685245",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.03924050632911397",
      "ci_low": "0.018986885126125678",
      "ci_high": "0.05998623212230807",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.03924050632911397",
      "ci_low": "0.017721518987341756",
      "ci_high": "0.06075949367088612",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "adjusted_tis_pct",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "2.632911392405063",
      "ci_low": "-1.138038427748941",
      "ci_high": "6.400635503253165",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "adjusted_tis_pct",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "2.632911392405063",
      "ci_low": "-0.8658472153725308",
      "ci_high": "6.0815585961155545",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "adjusted_tis_pct",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "2.632911392405063",
      "ci_low": "-0.9620253164556942",
      "ci_high": "6.228164556962025",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.004676164611761513",
      "ci_low": "-0.012269020836779684",
      "ci_high": "0.023637578907668242",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.004676164611761513",
      "ci_low": "-0.011010482756206428",
      "ci_high": "0.022746712502446906",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.004676164611761513",
      "ci_low": "-0.012276795552650668",
      "ci_high": "0.023174205649078326",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "reference_species_macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.031276768783179754",
      "ci_low": "0.007239524385654836",
      "ci_high": "0.05706430323720421",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "reference_species_macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.031276768783179754",
      "ci_low": "0.009320117050686102",
      "ci_high": "0.05585732139946285",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "reference_species_macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.031276768783179754",
      "ci_low": "0.007428717757157653",
      "ci_high": "0.056488929704051824",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_answer_coverage",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0620253164556962",
      "ci_low": "0.03295310519645123",
      "ci_high": "0.09195402298850575",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_answer_coverage",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0620253164556962",
      "ci_low": "0.035572297340493664",
      "ci_high": "0.08953982287051922",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_answer_coverage",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0620253164556962",
      "ci_low": "0.032911392405063244",
      "ci_high": "0.09113924050632916",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_attempted_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.009825392127848498",
      "ci_low": "-0.04555494100873813",
      "ci_high": "0.025324214887484015",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_attempted_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.009825392127848498",
      "ci_low": "-0.04183707565233108",
      "ci_high": "0.02163918291583723",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite MEDIUM",
      "metric": "species_attempted_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.009825392127848498",
      "ci_low": "-0.04332877762206659",
      "ci_high": "0.02323761794112151",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.04810126582278479",
      "ci_low": "0.02252816020025028",
      "ci_high": "0.07393595596133196",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.04810126582278479",
      "ci_low": "0.02528369134381791",
      "ci_high": "0.07013729062779693",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.04810126582278479",
      "ci_low": "0.024050632911392422",
      "ci_high": "0.07215189873417721",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "adjusted_tis_pct",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.5949367088607573",
      "ci_low": "-3.500786382862035",
      "ci_high": "4.65060904321513",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "adjusted_tis_pct",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.5949367088607573",
      "ci_low": "-2.994439193015135",
      "ci_high": "4.2471472093307625",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "adjusted_tis_pct",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.5949367088607573",
      "ci_low": "-3.303797468354432",
      "ci_high": "4.481329113924052",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.012229581481550955",
      "ci_low": "-0.03067823376294436",
      "ci_high": "0.008104244225891255",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.012229581481550955",
      "ci_low": "-0.029752734805778724",
      "ci_high": "0.0066864013590835335",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.012229581481550955",
      "ci_low": "-0.030424040724747074",
      "ci_high": "0.007839632214102487",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "reference_species_macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0407316723714537",
      "ci_low": "0.012995491413717085",
      "ci_high": "0.06836746603793194",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "reference_species_macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0407316723714537",
      "ci_low": "0.014865444607304025",
      "ci_high": "0.0667996641256394",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "reference_species_macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0407316723714537",
      "ci_low": "0.013877623598792743",
      "ci_high": "0.06828218723192664",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_answer_coverage",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.10379746835443038",
      "ci_low": "0.0750620229007634",
      "ci_high": "0.13291139240506333",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_answer_coverage",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.10379746835443038",
      "ci_low": "0.07650148989547742",
      "ci_high": "0.1307392756680732",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_answer_coverage",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.10379746835443038",
      "ci_low": "0.07591772151898733",
      "ci_high": "0.13291139240506322",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_attempted_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.03951333854224803",
      "ci_low": "-0.07748511064789769",
      "ci_high": "-0.0016877103274893608",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_attempted_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.03951333854224803",
      "ci_low": "-0.07289171574724236",
      "ci_high": "-0.006669793083004685",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite LOW",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_attempted_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.03951333854224803",
      "ci_low": "-0.07447383412661505",
      "ci_high": "-0.004046244322110198",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.008860759493670822",
      "ci_low": "-0.012674271229404344",
      "ci_high": "0.030303030303030276",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.008860759493670822",
      "ci_low": "-0.011188284289550095",
      "ci_high": "0.027848603576451653",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.008860759493670822",
      "ci_low": "-0.012658227848101222",
      "ci_high": "0.029113924050632956",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "adjusted_tis_pct",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-2.0379746835443058",
      "ci_low": "-5.596484642409681",
      "ci_high": "1.4885630922400028",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "adjusted_tis_pct",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-2.0379746835443058",
      "ci_low": "-5.183277736679625",
      "ci_high": "1.351319566509437",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "adjusted_tis_pct",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-2.0379746835443058",
      "ci_low": "-5.481012658227847",
      "ci_high": "1.4556962025316482",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.016905746093312468",
      "ci_low": "-0.0342824961468063",
      "ci_high": "0.00020829189792970963",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.016905746093312468",
      "ci_low": "-0.03309914410972159",
      "ci_high": "-0.0009414022823956486",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.016905746093312468",
      "ci_low": "-0.033882567192727316",
      "ci_high": "-4.49920360291567e-05",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "reference_species_macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.009454903588273944",
      "ci_low": "-0.01657218360207228",
      "ci_high": "0.03391381004833618",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "reference_species_macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.009454903588273944",
      "ci_low": "-0.014546339191156459",
      "ci_high": "0.032112620317647216",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "reference_species_macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.009454903588273944",
      "ci_low": "-0.015990838776711314",
      "ci_high": "0.03340122021753884",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_answer_coverage",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.04177215189873418",
      "ci_low": "0.015423175795926482",
      "ci_high": "0.06786171574903965",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_answer_coverage",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.04177215189873418",
      "ci_low": "0.017113194470789378",
      "ci_high": "0.06556578706104034",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_answer_coverage",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.04177215189873418",
      "ci_low": "0.016455696202531622",
      "ci_high": "0.06708860759493673",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_attempted_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.029687946414399535",
      "ci_low": "-0.0593638146638876",
      "ci_high": "-0.00016064133675963035",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_attempted_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.029687946414399535",
      "ci_low": "-0.057487547818352296",
      "ci_high": "-0.0019079845773343161",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Flash-Lite",
      "baseline": "Flash-Lite MEDIUM",
      "comparison": "Flash-Lite HIGH",
      "metric": "species_attempted_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.029687946414399535",
      "ci_low": "-0.05788146668010432",
      "ci_high": "-0.000857012434534307",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "species_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0025316455696202667",
      "ci_low": "-0.01656103763161315",
      "ci_high": "0.02179487179487183",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "species_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0025316455696202667",
      "ci_low": "-0.01579738475308094",
      "ci_high": "0.021691679452755495",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "species_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0025316455696202667",
      "ci_low": "-0.016455696202531622",
      "ci_high": "0.021518987341772156",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "adjusted_tis_pct",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.21518987341771378",
      "ci_low": "-3.0489759858964827",
      "ci_high": "2.6267936762689192",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "adjusted_tis_pct",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.21518987341771378",
      "ci_low": "-3.063939119951779",
      "ci_high": "2.457864674003896",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "adjusted_tis_pct",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.21518987341771378",
      "ci_low": "-3.000316455696209",
      "ci_high": "2.5949367088607573",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.03427779926059271",
      "ci_low": "-0.05231082280412932",
      "ci_high": "-0.014715042764506452",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.03427779926059271",
      "ci_low": "-0.05110219516652369",
      "ci_high": "-0.01629407719762159",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.03427779926059271",
      "ci_low": "-0.05192369694396387",
      "ci_high": "-0.015499268351593199",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "reference_species_macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.002634057486977426",
      "ci_low": "-0.01878921197005225",
      "ci_high": "0.024278757546586885",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "reference_species_macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.002634057486977426",
      "ci_low": "-0.01728738615127038",
      "ci_high": "0.02297245213203342",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "reference_species_macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.002634057486977426",
      "ci_low": "-0.018170751107971662",
      "ci_high": "0.023519307679167326",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "species_answer_coverage",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.006329113924050667",
      "ci_low": "-0.012579409455376047",
      "ci_high": "0.02518891687657432",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "species_answer_coverage",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.006329113924050667",
      "ci_low": "-0.011433708696366987",
      "ci_high": "0.02480644706910528",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "species_answer_coverage",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.006329113924050667",
      "ci_low": "-0.012658227848101222",
      "ci_high": "0.024050632911392422",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "species_attempted_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.004347604818716544",
      "ci_low": "-0.02885656856200422",
      "ci_high": "0.020007774786961172",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "species_attempted_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.004347604818716544",
      "ci_low": "-0.02885858587850038",
      "ci_high": "0.019349207879821487",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / LOW thinking",
      "metric": "species_attempted_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.004347604818716544",
      "ci_low": "-0.029076754690896798",
      "ci_high": "0.019522394635578037",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.07341772151898729",
      "ci_low": "0.05067867816181875",
      "ci_high": "0.09673570879877096",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.07341772151898729",
      "ci_low": "0.05325963066469408",
      "ci_high": "0.09409395720313438",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.07341772151898729",
      "ci_low": "0.051898734177215244",
      "ci_high": "0.09493670886075944",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "adjusted_tis_pct",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "1.3417721518987449",
      "ci_low": "-1.8067754077791705",
      "ci_high": "4.677669859638886",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "adjusted_tis_pct",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "1.3417721518987449",
      "ci_low": "-1.5550630628162267",
      "ci_high": "4.398491127367709",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "adjusted_tis_pct",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "1.3417721518987449",
      "ci_low": "-1.696202531645568",
      "ci_high": "4.481012658227854",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.0016832959375816525",
      "ci_low": "-0.02058400907949872",
      "ci_high": "0.01945174158295954",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.0016832959375816525",
      "ci_low": "-0.018606560814070028",
      "ci_high": "0.017647976526510525",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.0016832959375816525",
      "ci_low": "-0.019499334366407176",
      "ci_high": "0.01792767688620506",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "reference_species_macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.06306361774689806",
      "ci_low": "0.04001353545581016",
      "ci_high": "0.08656357168942977",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "reference_species_macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.06306361774689806",
      "ci_low": "0.042485927424560456",
      "ci_high": "0.08465598244618668",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "reference_species_macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.06306361774689806",
      "ci_low": "0.04154181666456895",
      "ci_high": "0.08489368213994926",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_answer_coverage",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.11265822784810131",
      "ci_low": "0.08905637184564462",
      "ci_high": "0.13665503557744935",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_answer_coverage",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.11265822784810131",
      "ci_low": "0.09017715494772446",
      "ci_high": "0.13368496921503256",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_answer_coverage",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.11265822784810131",
      "ci_low": "0.08987341772151902",
      "ci_high": "0.1354430379746836",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_attempted_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.030086975061296428",
      "ci_low": "-0.056220950644351474",
      "ci_high": "-0.003061092193470638",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_attempted_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.030086975061296428",
      "ci_low": "-0.0532936538621701",
      "ci_high": "-0.004536867627219506",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / MEDIUM thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_attempted_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.030086975061296428",
      "ci_low": "-0.05532217119135423",
      "ci_high": "-0.004907690472107125",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.07088607594936702",
      "ci_low": "0.04797979797979801",
      "ci_high": "0.09398496240601495",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.07088607594936702",
      "ci_low": "0.04980061692878163",
      "ci_high": "0.092170150385923",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.07088607594936702",
      "ci_low": "0.048101265822784844",
      "ci_high": "0.09240506329113929",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "adjusted_tis_pct",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "1.5569620253164587",
      "ci_low": "-1.7446671568970056",
      "ci_high": "4.938214228680939",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "adjusted_tis_pct",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "1.5569620253164587",
      "ci_low": "-1.4679363709427005",
      "ci_high": "4.7802143531573975",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "adjusted_tis_pct",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "1.5569620253164587",
      "ci_low": "-1.7215189873417673",
      "ci_high": "4.823101265822776",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.03259450332301106",
      "ci_low": "0.013386030619603596",
      "ci_high": "0.05209707835650405",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.03259450332301106",
      "ci_low": "0.01481766070800805",
      "ci_high": "0.05124625711341411",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.03259450332301106",
      "ci_low": "0.014101932974432102",
      "ci_high": "0.051457673444269054",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "reference_species_macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.060429560259920634",
      "ci_low": "0.03690560128527441",
      "ci_high": "0.0835704339623341",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "reference_species_macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.060429560259920634",
      "ci_low": "0.03889438660130272",
      "ci_high": "0.08269856854439989",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "reference_species_macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.060429560259920634",
      "ci_low": "0.037867036745375074",
      "ci_high": "0.08287377700682827",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_answer_coverage",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.10632911392405064",
      "ci_low": "0.08322663252240725",
      "ci_high": "0.13016400874038178",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_answer_coverage",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.10632911392405064",
      "ci_low": "0.084427481361817",
      "ci_high": "0.12649930589804007",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_answer_coverage",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.10632911392405064",
      "ci_low": "0.08354430379746836",
      "ci_high": "0.12911392405063293",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_attempted_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.025739370242579884",
      "ci_low": "-0.05296472293481144",
      "ci_high": "0.0010925964821968523",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_attempted_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.025739370242579884",
      "ci_low": "-0.05041133791139144",
      "ci_high": "0.0007397197287607623",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Gemini 3.8",
      "baseline": "Gemini 3.8 HIGH media / LOW thinking",
      "comparison": "Gemini 3.8 HIGH media / HIGH thinking",
      "metric": "species_attempted_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.025739370242579884",
      "ci_low": "-0.052212130726124614",
      "ci_high": "0.001058481824756596",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "species_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0012658227848101333",
      "ci_low": "0.0",
      "ci_high": "0.003870967741935516",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "species_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0012658227848101333",
      "ci_low": "0.0",
      "ci_high": "0.004219409282700426",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "species_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0012658227848101333",
      "ci_low": "0.0",
      "ci_high": "0.0037974683544304",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "adjusted_tis_pct",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.2151898734177209",
      "ci_low": "-0.6988564167725517",
      "ci_high": "0.2544529262086517",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "adjusted_tis_pct",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.2151898734177209",
      "ci_low": "-0.6444188722669724",
      "ci_high": "0.2067510548523206",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "adjusted_tis_pct",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.2151898734177209",
      "ci_low": "-0.6962025316455716",
      "ci_high": "0.2531645569620249",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.0009736616611710491",
      "ci_low": "-0.0012257909236593207",
      "ci_high": "-0.0003817345784623284",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.0009736616611710491",
      "ci_low": "-0.001232209230379999",
      "ci_high": "-0.00038402296033580196",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.0009736616611710491",
      "ci_low": "-0.0012264474088379592",
      "ci_high": "-0.0004167391976629943",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "reference_species_macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0004851585586380103",
      "ci_low": "0.0",
      "ci_high": "0.0016769531147627066",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "reference_species_macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0004851585586380103",
      "ci_low": "0.0",
      "ci_high": "0.001699516330069596",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "reference_species_macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0004851585586380103",
      "ci_low": "0.0",
      "ci_high": "0.0016034664073885686",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "species_answer_coverage",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.005063291139240533",
      "ci_low": "0.0012437810945273853",
      "ci_high": "0.010296010296010242",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "species_answer_coverage",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.005063291139240533",
      "ci_low": "0.0012658227848101333",
      "ci_high": "0.010126582278480956",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "species_answer_coverage",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.005063291139240533",
      "ci_low": "0.0012658227848101333",
      "ci_high": "0.010126582278481067",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "species_attempted_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.0013052558301427375",
      "ci_low": "-0.004221075823580803",
      "ci_high": "0.0015470001335905659",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "species_attempted_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.0013052558301427375",
      "ci_low": "-0.0038979186608903275",
      "ci_high": "0.0012703688998047765",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "Qwen repair descriptive",
      "baseline": "Qwen Instruct first_pass",
      "comparison": "Qwen Instruct primary_repaired",
      "metric": "species_attempted_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.0013052558301427375",
      "ci_low": "-0.004293547236998841",
      "ci_high": "0.001466380015382969",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "BioCLIP geography exploratory",
      "baseline": "BioCLIP 2.5 / primary_blind",
      "comparison": "BioCLIP 2.5 / operational_country_ablation",
      "metric": "species_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.03164556962025317",
      "ci_low": "0.02020202020202022",
      "ci_high": "0.04441765459013536",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "BioCLIP geography exploratory",
      "baseline": "BioCLIP 2.5 / primary_blind",
      "comparison": "BioCLIP 2.5 / operational_country_ablation",
      "metric": "species_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.03164556962025317",
      "ci_low": "0.021356653192096182",
      "ci_high": "0.04307313642756688",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "BioCLIP geography exploratory",
      "baseline": "BioCLIP 2.5 / primary_blind",
      "comparison": "BioCLIP 2.5 / operational_country_ablation",
      "metric": "species_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.03164556962025317",
      "ci_low": "0.020253164556962022",
      "ci_high": "0.044303797468354444",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "BioCLIP geography exploratory",
      "baseline": "BioCLIP 2.5 / primary_blind",
      "comparison": "BioCLIP 2.5 / operational_country_ablation",
      "metric": "macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0830444554753087",
      "ci_low": "0.07360477090970713",
      "ci_high": "0.08875945713328645",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "BioCLIP geography exploratory",
      "baseline": "BioCLIP 2.5 / primary_blind",
      "comparison": "BioCLIP 2.5 / operational_country_ablation",
      "metric": "macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0830444554753087",
      "ci_low": "0.07504592524809253",
      "ci_high": "0.08901679832130246",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "BioCLIP geography exploratory",
      "baseline": "BioCLIP 2.5 / primary_blind",
      "comparison": "BioCLIP 2.5 / operational_country_ablation",
      "metric": "macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0830444554753087",
      "ci_low": "0.07408161061689587",
      "ci_high": "0.08850468613662489",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "BioCLIP geography exploratory",
      "baseline": "BioCLIP 2.5 / primary_blind",
      "comparison": "BioCLIP 2.5 / operational_country_ablation",
      "metric": "reference_species_macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.01926799987790967",
      "ci_low": "0.008942956964169285",
      "ci_high": "0.03142061371253507",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "BioCLIP geography exploratory",
      "baseline": "BioCLIP 2.5 / primary_blind",
      "comparison": "BioCLIP 2.5 / operational_country_ablation",
      "metric": "reference_species_macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.01926799987790967",
      "ci_low": "0.010012519144793739",
      "ci_high": "0.03092999073881233",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "BioCLIP geography exploratory",
      "baseline": "BioCLIP 2.5 / primary_blind",
      "comparison": "BioCLIP 2.5 / operational_country_ablation",
      "metric": "reference_species_macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.01926799987790967",
      "ci_low": "0.008955459644005308",
      "ci_high": "0.030882184463797806",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "BioCLIP geography exploratory",
      "baseline": "BioCLIP 2.5 / primary_blind",
      "comparison": "BioCLIP 2.5 / operational_country_ablation",
      "metric": "species_answer_coverage",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0",
      "ci_low": "0.0",
      "ci_high": "0.0",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "BioCLIP geography exploratory",
      "baseline": "BioCLIP 2.5 / primary_blind",
      "comparison": "BioCLIP 2.5 / operational_country_ablation",
      "metric": "species_answer_coverage",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0",
      "ci_low": "0.0",
      "ci_high": "0.0",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "BioCLIP geography exploratory",
      "baseline": "BioCLIP 2.5 / primary_blind",
      "comparison": "BioCLIP 2.5 / operational_country_ablation",
      "metric": "species_answer_coverage",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0",
      "ci_low": "0.0",
      "ci_high": "0.0",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "BioCLIP geography exploratory",
      "baseline": "BioCLIP 2.5 / primary_blind",
      "comparison": "BioCLIP 2.5 / operational_country_ablation",
      "metric": "species_attempted_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.03164556962025317",
      "ci_low": "0.02020202020202022",
      "ci_high": "0.04441765459013536",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "BioCLIP geography exploratory",
      "baseline": "BioCLIP 2.5 / primary_blind",
      "comparison": "BioCLIP 2.5 / operational_country_ablation",
      "metric": "species_attempted_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.03164556962025317",
      "ci_low": "0.021356653192096182",
      "ci_high": "0.04307313642756688",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "BioCLIP geography exploratory",
      "baseline": "BioCLIP 2.5 / primary_blind",
      "comparison": "BioCLIP 2.5 / operational_country_ablation",
      "metric": "species_attempted_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.03164556962025317",
      "ci_low": "0.020253164556962022",
      "ci_high": "0.044303797468354444",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "species_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0075949367088607445",
      "ci_low": "0.0025000000000000022",
      "ci_high": "0.014066946435647144",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "species_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0075949367088607445",
      "ci_low": "0.002531645569620211",
      "ci_high": "0.014064697609001366",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "species_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0075949367088607445",
      "ci_low": "0.002531645569620211",
      "ci_high": "0.013924050632911411",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "adjusted_tis_pct",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "10.962025316455698",
      "ci_low": "8.600452378450163",
      "ci_high": "13.41648243829267",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "adjusted_tis_pct",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "10.962025316455698",
      "ci_low": "9.14157876230661",
      "ci_high": "12.847599183516902",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "adjusted_tis_pct",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "10.962025316455698",
      "ci_low": "8.91139240506329",
      "ci_high": "13.164556962025316",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0888890253980846",
      "ci_low": "0.07957373295398308",
      "ci_high": "0.09540879025774401",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0888890253980846",
      "ci_low": "0.08079485728142047",
      "ci_high": "0.09517275655389702",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.0888890253980846",
      "ci_low": "0.08037164030336456",
      "ci_high": "0.09491899199858472",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "reference_species_macro_f1",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.009161285709509803",
      "ci_low": "0.002909305117222938",
      "ci_high": "0.016806847937287792",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "reference_species_macro_f1",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.009161285709509803",
      "ci_low": "0.003254906384915386",
      "ci_high": "0.015982290542564444",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "reference_species_macro_f1",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.009161285709509803",
      "ci_low": "0.003313031953742024",
      "ci_high": "0.0162672278110755",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "species_answer_coverage",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.13670886075949362",
      "ci_low": "-0.16223150871655065",
      "ci_high": "-0.11209040562625672",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "species_answer_coverage",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.13670886075949362",
      "ci_low": "-0.1571378682850202",
      "ci_high": "-0.11736249685616774",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "species_answer_coverage",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "-0.13670886075949362",
      "ci_low": "-0.1594936708860759",
      "ci_high": "-0.11518987341772152",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "species_attempted_accuracy",
      "method": "source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.2453548540232614",
      "ci_low": "0.2023717077797373",
      "ci_high": "0.2900872021886512",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "species_attempted_accuracy",
      "method": "species_balanced_source_cluster",
      "direction": "comparison_minus_baseline",
      "estimate": "0.2453548540232614",
      "ci_low": "0.21439689438867954",
      "ci_high": "0.2778816564002792",
      "valid_resamples": "10000",
      "interval_status": "ok"
    },
    {
      "family": "SpeciesNet geography exploratory",
      "baseline": "SpeciesNet v4.0.3b / primary_blind",
      "comparison": "SpeciesNet v4.0.3b / operational_country_ablation",
      "metric": "species_attempted_accuracy",
      "method": "legacy_event_stratified",
      "direction": "comparison_minus_baseline",
      "estimate": "0.2453548540232614",
      "ci_low": "0.20897114539875783",
      "ci_high": "0.28289530480058506",
      "valid_resamples": "10000",
      "interval_status": "ok"
    }
  ],
  "unit_policy": {
    "table": "paired_intervals.csv",
    "metric_units": {
      "adjusted_tis_pct": "percentage_points_difference_of_five_point_maximum",
      "macro_f1": "f1_difference",
      "reference_species_macro_f1": "f1_difference",
      "species_accuracy": "proportion_point_difference",
      "species_answer_coverage": "proportion_point_difference",
      "species_attempted_accuracy": "proportion_point_difference"
    },
    "counts": {
      "valid_resamples": "bootstrap_resamples"
    },
    "interval": "ci_low and ci_high use the same units as estimate; pointwise 95% when interval_status is ok",
    "estimate": "unit selected by each row's metric_units entry; signed difference, not an absolute score",
    "adjustment": "intervals are pointwise; Holm-adjusted p-values are only in paired_tests.csv"
  },
  "note": "Saved aggregate candidate analysis; intervals are pointwise and exploratory tests have separate Holm-adjusted p-values."
}
