{
  "inputs": {
    "summary": {
      "path": "benchmark/expanded/summary.json",
      "sha256": "2e373db135eb1c0625252524b5aa5874f8e33b193d0669f28c9a3c70f4ea3a18"
    },
    "quality": {
      "path": "benchmark/expanded/runs/quality-20261002T093341Z/responses.jsonl",
      "sha256": "c508f6173a27bed9834a294329414cf499bdc078381eb3f41e4d66c08f9f3f8b"
    },
    "timing": {
      "path": "benchmark/expanded/runs/timing-20261002T095011Z/responses.jsonl",
      "sha256": "6625f9f8375f3552e8311044155230b7b8d9efac550d8338b2213d5d38f50cc9"
    },
    "cache": {
      "path": "benchmark/expanded/runs/cache-20261002T095722Z/responses.jsonl",
      "sha256": "a837d11e4ad92e01c258a5d5e1defe82f6b64b7e4440ec6f383defc30d650e3f"
    }
  },
  "generator": {
    "path": "benchmark/expanded/paper_figures.py",
    "sha256": "4ffa97c02d9afea34359992934d9899e0baca0ef81c697d60c476fff4c6f01c7",
    "shared_style_sha256": "00b037750fbb47947ebdeeab2ed708f196d19165bf430460edd4c4455a0e0bf3",
    "matplotlib": "3.11.2",
    "numpy": "2.5.3"
  },
  "figures": {
    "fig4-quality": {
      "svg": "fig4-quality.svg",
      "png": "fig4-quality.png",
      "size_inches": [
        12.4,
        5.3
      ],
      "png_pixels": [
        3224,
        1378
      ],
      "title": "Binary-task quality",
      "caption": "Accuracy on four binary tasks, counting every invalid response as incorrect. Points show observed accuracy and horizontal intervals are nominal 95% Wilson intervals. The table gives correct answers over all attempted cases. BoolQ, RTE, and WiC measure agreement with published reference labels; synthetic policy targets come from a deterministic oracle. The intervals are exploratory, do not account for every shared passage or template, and should not be used to infer model differences from overlap alone.",
      "svg_sha256": "32a81d0c89a933f89643d9aff202057fc81aacc09bc9f79b7f4baa1adaed6b3c",
      "png_sha256": "7a743ae976c2cab9e330863b55cfc4ea242cf3abce13e7594f0219fd8dba68f8"
    },
    "fig5-probability": {
      "svg": "fig5-probability.svg",
      "png": "fig5-probability.png",
      "size_inches": [
        11.2,
        5.7
      ],
      "png_pixels": [
        2912,
        1482
      ],
      "title": "Probability estimates against exact targets",
      "caption": "Model-generated probabilities versus exact event probabilities on 300 prespecified finite probability problems. Each panel reports its own valid-response denominator and mean absolute error in percentage points. Invalid responses have no plottable probability and are omitted here but retained in the study's failure accounting. The dashed diagonal is exact agreement. No jitter is added to the probability values, and overlapping marks remain present.",
      "svg_sha256": "f3b1c4391bab15a1261c750668612f4045c364f6c888db08e7114dff8e8ac3e0",
      "png_sha256": "3cd3dc24ae6dd0c475e9c01d084bb4428bd5a5c9cd55c7c6736b4167afdcbb0e"
    },
    "fig6-timing": {
      "svg": "fig6-timing.svg",
      "png": "fig6-timing.png",
      "size_inches": [
        12.4,
        5.0
      ],
      "png_pixels": [
        3224,
        1300
      ],
      "title": "Isolated repeated latency across five tasks",
      "caption": "The isolated serial timing protocol selected 12 cases per task before outcomes were inspected and repeated them three times per model. Each plotted point is one case/model's median over those three repeats. Grey connectors pair the same case across models, and heavy horizontal bars mark medians across case medians. Cases missing a valid positive timing in either arm in any repeat are excluded with counts retained in the analysis. Common logarithmic time axes are used. No concurrent quality-run latency is included.",
      "svg_sha256": "47058569b68f081b59f4133ce71c1808fff1ba772a0895c27421b3833f205116",
      "png_sha256": "79bcda9edf5608c8c52e1906b8dbb150fd41a7850433f5a43c81ab4d8097b5f2"
    },
    "fig7-cache": {
      "svg": "fig7-cache.svg",
      "png": "fig7-cache.png",
      "size_inches": [
        12.4,
        5.8
      ],
      "png_pixels": [
        3224,
        1508
      ],
      "title": "Four-block cache cost and latency",
      "caption": "Four separately primed long-prefix blocks, each containing 24 distinct deterministic policy cases and all three model arms. Panel A scales each block's actual measured mean cost to 1,000 decisions. The white hatched extension on cached Luna is that block's distinct priming charge divided across its 24 measured cases. Panel B shows individual complete-response latencies as faint dots and block medians as connected marks, excluding all priming requests. The four blocks share one session and are not independent days or deployments. Returned cache counters establish treatment status; no cache-effect interpretation is warranted for a failed control.",
      "svg_sha256": "e688af46487173e3f521837c60376b4310cd8af2b50d1748073298110f77c272",
      "png_sha256": "674f91b0e39dd1021ffb8b5aa979bfb885571501e4c43eda15401232949b5349"
    }
  },
  "metrics": {
    "binary_quality": {
      "boolq": {
        "Jev": {
          "accuracy": 0.917737003058104,
          "wilson95": [
            0.9078221114479283,
            0.9266715665031486
          ],
          "correct": 3001,
          "attempts": 3270,
          "invalid": 0
        },
        "Luna": {
          "accuracy": 0.863914373088685,
          "wilson95": [
            0.8517343947848243,
            0.8752403319807552
          ],
          "correct": 2825,
          "attempts": 3270,
          "invalid": 0
        }
      },
      "rte": {
        "Jev": {
          "accuracy": 0.9061371841155235,
          "wilson95": [
            0.8660240489486702,
            0.9351397142108616
          ],
          "correct": 251,
          "attempts": 277,
          "invalid": 0
        },
        "Luna": {
          "accuracy": 0.8700361010830325,
          "wilson95": [
            0.8253226276877503,
            0.9046265787464898
          ],
          "correct": 241,
          "attempts": 277,
          "invalid": 0
        }
      },
      "wic": {
        "Jev": {
          "accuracy": 0.7398119122257053,
          "wilson95": [
            0.704404201462688,
            0.7723490462353305
          ],
          "correct": 472,
          "attempts": 638,
          "invalid": 0
        },
        "Luna": {
          "accuracy": 0.7288401253918495,
          "wilson95": [
            0.6930507977097585,
            0.761890209897501
          ],
          "correct": 465,
          "attempts": 638,
          "invalid": 0
        }
      },
      "policy": {
        "Jev": {
          "accuracy": 1.0,
          "wilson95": [
            0.9873570287754538,
            0.9999999999999998
          ],
          "correct": 300,
          "attempts": 300,
          "invalid": 0
        },
        "Luna": {
          "accuracy": 0.9966666666666667,
          "wilson95": [
            0.9813633063054688,
            0.9994113422781488
          ],
          "correct": 299,
          "attempts": 300,
          "invalid": 0
        }
      }
    },
    "exact_probability": {
      "Jev": {
        "attempts": 300,
        "valid": 299,
        "invalid": 1,
        "mae": 0.07640276740848177,
        "plotted_case_ids": [
          "expanded-probability-0145",
          "expanded-probability-0206",
          "expanded-probability-0130",
          "expanded-probability-0152",
          "expanded-probability-0262",
          "expanded-probability-0008",
          "expanded-probability-0069",
          "expanded-probability-0275",
          "expanded-probability-0185",
          "expanded-probability-0237",
          "expanded-probability-0180",
          "expanded-probability-0080",
          "expanded-probability-0006",
          "expanded-probability-0049",
          "expanded-probability-0186",
          "expanded-probability-0159",
          "expanded-probability-0271",
          "expanded-probability-0277",
          "expanded-probability-0245",
          "expanded-probability-0136",
          "expanded-probability-0288",
          "expanded-probability-0283",
          "expanded-probability-0229",
          "expanded-probability-0093",
          "expanded-probability-0048",
          "expanded-probability-0151",
          "expanded-probability-0181",
          "expanded-probability-0218",
          "expanded-probability-0154",
          "expanded-probability-0232",
          "expanded-probability-0276",
          "expanded-probability-0128",
          "expanded-probability-0067",
          "expanded-probability-0247",
          "expanded-probability-0138",
          "expanded-probability-0005",
          "expanded-probability-0268",
          "expanded-probability-0072",
          "expanded-probability-0220",
          "expanded-probability-0281",
          "expanded-probability-0222",
          "expanded-probability-0123",
          "expanded-probability-0226",
          "expanded-probability-0148",
          "expanded-probability-0280",
          "expanded-probability-0090",
          "expanded-probability-0234",
          "expanded-probability-0224",
          "expanded-probability-0225",
          "expanded-probability-0291",
          "expanded-probability-0267",
          "expanded-probability-0141",
          "expanded-probability-0025",
          "expanded-probability-0293",
          "expanded-probability-0104",
          "expanded-probability-0196",
          "expanded-probability-0058",
          "expanded-probability-0017",
          "expanded-probability-0211",
          "expanded-probability-0031",
          "expanded-probability-0176",
          "expanded-probability-0113",
          "expanded-probability-0249",
          "expanded-probability-0042",
          "expanded-probability-0114",
          "expanded-probability-0055",
          "expanded-probability-0212",
          "expanded-probability-0092",
          "expanded-probability-0200",
          "expanded-probability-0013",
          "expanded-probability-0111",
          "expanded-probability-0106",
          "expanded-probability-0018",
          "expanded-probability-0001",
          "expanded-probability-0024",
          "expanded-probability-0139",
          "expanded-probability-0137",
          "expanded-probability-0132",
          "expanded-probability-0235",
          "expanded-probability-0269",
          "expanded-probability-0170",
          "expanded-probability-0207",
          "expanded-probability-0045",
          "expanded-probability-0079",
          "expanded-probability-0056",
          "expanded-probability-0081",
          "expanded-probability-0242",
          "expanded-probability-0146",
          "expanded-probability-0027",
          "expanded-probability-0052",
          "expanded-probability-0019",
          "expanded-probability-0125",
          "expanded-probability-0164",
          "expanded-probability-0167",
          "expanded-probability-0265",
          "expanded-probability-0103",
          "expanded-probability-0292",
          "expanded-probability-0144",
          "expanded-probability-0240",
          "expanded-probability-0204",
          "expanded-probability-0026",
          "expanded-probability-0039",
          "expanded-probability-0066",
          "expanded-probability-0163",
          "expanded-probability-0023",
          "expanded-probability-0228",
          "expanded-probability-0044",
          "expanded-probability-0112",
          "expanded-probability-0248",
          "expanded-probability-0278",
          "expanded-probability-0160",
          "expanded-probability-0061",
          "expanded-probability-0193",
          "expanded-probability-0155",
          "expanded-probability-0266",
          "expanded-probability-0033",
          "expanded-probability-0087",
          "expanded-probability-0108",
          "expanded-probability-0282",
          "expanded-probability-0260",
          "expanded-probability-0021",
          "expanded-probability-0003",
          "expanded-probability-0221",
          "expanded-probability-0135",
          "expanded-probability-0286",
          "expanded-probability-0126",
          "expanded-probability-0096",
          "expanded-probability-0197",
          "expanded-probability-0071",
          "expanded-probability-0043",
          "expanded-probability-0127",
          "expanded-probability-0251",
          "expanded-probability-0203",
          "expanded-probability-0158",
          "expanded-probability-0102",
          "expanded-probability-0120",
          "expanded-probability-0073",
          "expanded-probability-0188",
          "expanded-probability-0166",
          "expanded-probability-0199",
          "expanded-probability-0142",
          "expanded-probability-0020",
          "expanded-probability-0162",
          "expanded-probability-0243",
          "expanded-probability-0257",
          "expanded-probability-0082",
          "expanded-probability-0076",
          "expanded-probability-0095",
          "expanded-probability-0153",
          "expanded-probability-0261",
          "expanded-probability-0183",
          "expanded-probability-0007",
          "expanded-probability-0014",
          "expanded-probability-0129",
          "expanded-probability-0002",
          "expanded-probability-0273",
          "expanded-probability-0210",
          "expanded-probability-0230",
          "expanded-probability-0121",
          "expanded-probability-0300",
          "expanded-probability-0009",
          "expanded-probability-0077",
          "expanded-probability-0041",
          "expanded-probability-0053",
          "expanded-probability-0231",
          "expanded-probability-0147",
          "expanded-probability-0233",
          "expanded-probability-0030",
          "expanded-probability-0116",
          "expanded-probability-0238",
          "expanded-probability-0134",
          "expanded-probability-0022",
          "expanded-probability-0074",
          "expanded-probability-0246",
          "expanded-probability-0059",
          "expanded-probability-0016",
          "expanded-probability-0169",
          "expanded-probability-0037",
          "expanded-probability-0250",
          "expanded-probability-0124",
          "expanded-probability-0296",
          "expanded-probability-0239",
          "expanded-probability-0083",
          "expanded-probability-0122",
          "expanded-probability-0264",
          "expanded-probability-0109",
          "expanded-probability-0255",
          "expanded-probability-0168",
          "expanded-probability-0099",
          "expanded-probability-0177",
          "expanded-probability-0236",
          "expanded-probability-0091",
          "expanded-probability-0078",
          "expanded-probability-0201",
          "expanded-probability-0290",
          "expanded-probability-0064",
          "expanded-probability-0032",
          "expanded-probability-0088",
          "expanded-probability-0182",
          "expanded-probability-0010",
          "expanded-probability-0272",
          "expanded-probability-0034",
          "expanded-probability-0241",
          "expanded-probability-0085",
          "expanded-probability-0191",
          "expanded-probability-0198",
          "expanded-probability-0216",
          "expanded-probability-0223",
          "expanded-probability-0165",
          "expanded-probability-0219",
          "expanded-probability-0029",
          "expanded-probability-0060",
          "expanded-probability-0209",
          "expanded-probability-0036",
          "expanded-probability-0284",
          "expanded-probability-0047",
          "expanded-probability-0046",
          "expanded-probability-0208",
          "expanded-probability-0289",
          "expanded-probability-0131",
          "expanded-probability-0110",
          "expanded-probability-0094",
          "expanded-probability-0294",
          "expanded-probability-0161",
          "expanded-probability-0038",
          "expanded-probability-0195",
          "expanded-probability-0011",
          "expanded-probability-0063",
          "expanded-probability-0140",
          "expanded-probability-0107",
          "expanded-probability-0119",
          "expanded-probability-0189",
          "expanded-probability-0051",
          "expanded-probability-0287",
          "expanded-probability-0062",
          "expanded-probability-0115",
          "expanded-probability-0299",
          "expanded-probability-0217",
          "expanded-probability-0173",
          "expanded-probability-0270",
          "expanded-probability-0175",
          "expanded-probability-0040",
          "expanded-probability-0068",
          "expanded-probability-0178",
          "expanded-probability-0274",
          "expanded-probability-0184",
          "expanded-probability-0285",
          "expanded-probability-0065",
          "expanded-probability-0117",
          "expanded-probability-0174",
          "expanded-probability-0156",
          "expanded-probability-0244",
          "expanded-probability-0015",
          "expanded-probability-0004",
          "expanded-probability-0253",
          "expanded-probability-0105",
          "expanded-probability-0143",
          "expanded-probability-0028",
          "expanded-probability-0057",
          "expanded-probability-0263",
          "expanded-probability-0012",
          "expanded-probability-0190",
          "expanded-probability-0205",
          "expanded-probability-0050",
          "expanded-probability-0086",
          "expanded-probability-0215",
          "expanded-probability-0187",
          "expanded-probability-0171",
          "expanded-probability-0258",
          "expanded-probability-0118",
          "expanded-probability-0089",
          "expanded-probability-0133",
          "expanded-probability-0101",
          "expanded-probability-0070",
          "expanded-probability-0227",
          "expanded-probability-0150",
          "expanded-probability-0295",
          "expanded-probability-0297",
          "expanded-probability-0097",
          "expanded-probability-0254",
          "expanded-probability-0298",
          "expanded-probability-0157",
          "expanded-probability-0054",
          "expanded-probability-0194",
          "expanded-probability-0202",
          "expanded-probability-0100",
          "expanded-probability-0252",
          "expanded-probability-0179",
          "expanded-probability-0259",
          "expanded-probability-0279",
          "expanded-probability-0256",
          "expanded-probability-0172",
          "expanded-probability-0192",
          "expanded-probability-0149",
          "expanded-probability-0098",
          "expanded-probability-0213",
          "expanded-probability-0035",
          "expanded-probability-0084",
          "expanded-probability-0214"
        ]
      },
      "Luna": {
        "attempts": 300,
        "valid": 300,
        "invalid": 0,
        "mae": 0.024395889425565125,
        "plotted_case_ids": [
          "expanded-probability-0206",
          "expanded-probability-0130",
          "expanded-probability-0152",
          "expanded-probability-0145",
          "expanded-probability-0262",
          "expanded-probability-0008",
          "expanded-probability-0069",
          "expanded-probability-0275",
          "expanded-probability-0185",
          "expanded-probability-0237",
          "expanded-probability-0075",
          "expanded-probability-0180",
          "expanded-probability-0080",
          "expanded-probability-0006",
          "expanded-probability-0049",
          "expanded-probability-0186",
          "expanded-probability-0277",
          "expanded-probability-0159",
          "expanded-probability-0271",
          "expanded-probability-0245",
          "expanded-probability-0136",
          "expanded-probability-0288",
          "expanded-probability-0283",
          "expanded-probability-0229",
          "expanded-probability-0093",
          "expanded-probability-0048",
          "expanded-probability-0151",
          "expanded-probability-0181",
          "expanded-probability-0218",
          "expanded-probability-0154",
          "expanded-probability-0232",
          "expanded-probability-0276",
          "expanded-probability-0128",
          "expanded-probability-0067",
          "expanded-probability-0247",
          "expanded-probability-0005",
          "expanded-probability-0138",
          "expanded-probability-0268",
          "expanded-probability-0072",
          "expanded-probability-0220",
          "expanded-probability-0281",
          "expanded-probability-0222",
          "expanded-probability-0123",
          "expanded-probability-0226",
          "expanded-probability-0148",
          "expanded-probability-0280",
          "expanded-probability-0090",
          "expanded-probability-0234",
          "expanded-probability-0225",
          "expanded-probability-0224",
          "expanded-probability-0291",
          "expanded-probability-0267",
          "expanded-probability-0141",
          "expanded-probability-0025",
          "expanded-probability-0293",
          "expanded-probability-0104",
          "expanded-probability-0196",
          "expanded-probability-0058",
          "expanded-probability-0017",
          "expanded-probability-0211",
          "expanded-probability-0176",
          "expanded-probability-0031",
          "expanded-probability-0113",
          "expanded-probability-0042",
          "expanded-probability-0249",
          "expanded-probability-0114",
          "expanded-probability-0055",
          "expanded-probability-0212",
          "expanded-probability-0092",
          "expanded-probability-0200",
          "expanded-probability-0013",
          "expanded-probability-0111",
          "expanded-probability-0106",
          "expanded-probability-0018",
          "expanded-probability-0001",
          "expanded-probability-0139",
          "expanded-probability-0024",
          "expanded-probability-0132",
          "expanded-probability-0137",
          "expanded-probability-0235",
          "expanded-probability-0269",
          "expanded-probability-0170",
          "expanded-probability-0207",
          "expanded-probability-0045",
          "expanded-probability-0079",
          "expanded-probability-0081",
          "expanded-probability-0056",
          "expanded-probability-0027",
          "expanded-probability-0242",
          "expanded-probability-0052",
          "expanded-probability-0146",
          "expanded-probability-0019",
          "expanded-probability-0125",
          "expanded-probability-0164",
          "expanded-probability-0265",
          "expanded-probability-0103",
          "expanded-probability-0292",
          "expanded-probability-0167",
          "expanded-probability-0240",
          "expanded-probability-0144",
          "expanded-probability-0204",
          "expanded-probability-0026",
          "expanded-probability-0066",
          "expanded-probability-0039",
          "expanded-probability-0163",
          "expanded-probability-0023",
          "expanded-probability-0228",
          "expanded-probability-0044",
          "expanded-probability-0112",
          "expanded-probability-0248",
          "expanded-probability-0278",
          "expanded-probability-0160",
          "expanded-probability-0061",
          "expanded-probability-0193",
          "expanded-probability-0155",
          "expanded-probability-0266",
          "expanded-probability-0087",
          "expanded-probability-0033",
          "expanded-probability-0282",
          "expanded-probability-0108",
          "expanded-probability-0021",
          "expanded-probability-0003",
          "expanded-probability-0260",
          "expanded-probability-0221",
          "expanded-probability-0135",
          "expanded-probability-0126",
          "expanded-probability-0286",
          "expanded-probability-0096",
          "expanded-probability-0197",
          "expanded-probability-0071",
          "expanded-probability-0043",
          "expanded-probability-0251",
          "expanded-probability-0127",
          "expanded-probability-0203",
          "expanded-probability-0158",
          "expanded-probability-0102",
          "expanded-probability-0120",
          "expanded-probability-0073",
          "expanded-probability-0188",
          "expanded-probability-0166",
          "expanded-probability-0142",
          "expanded-probability-0199",
          "expanded-probability-0020",
          "expanded-probability-0162",
          "expanded-probability-0082",
          "expanded-probability-0257",
          "expanded-probability-0243",
          "expanded-probability-0076",
          "expanded-probability-0095",
          "expanded-probability-0153",
          "expanded-probability-0261",
          "expanded-probability-0183",
          "expanded-probability-0007",
          "expanded-probability-0014",
          "expanded-probability-0129",
          "expanded-probability-0002",
          "expanded-probability-0273",
          "expanded-probability-0210",
          "expanded-probability-0121",
          "expanded-probability-0230",
          "expanded-probability-0300",
          "expanded-probability-0009",
          "expanded-probability-0077",
          "expanded-probability-0041",
          "expanded-probability-0231",
          "expanded-probability-0053",
          "expanded-probability-0233",
          "expanded-probability-0147",
          "expanded-probability-0030",
          "expanded-probability-0238",
          "expanded-probability-0116",
          "expanded-probability-0022",
          "expanded-probability-0134",
          "expanded-probability-0074",
          "expanded-probability-0016",
          "expanded-probability-0246",
          "expanded-probability-0169",
          "expanded-probability-0059",
          "expanded-probability-0037",
          "expanded-probability-0250",
          "expanded-probability-0296",
          "expanded-probability-0124",
          "expanded-probability-0239",
          "expanded-probability-0083",
          "expanded-probability-0122",
          "expanded-probability-0255",
          "expanded-probability-0109",
          "expanded-probability-0264",
          "expanded-probability-0099",
          "expanded-probability-0236",
          "expanded-probability-0177",
          "expanded-probability-0078",
          "expanded-probability-0091",
          "expanded-probability-0168",
          "expanded-probability-0201",
          "expanded-probability-0290",
          "expanded-probability-0064",
          "expanded-probability-0088",
          "expanded-probability-0032",
          "expanded-probability-0182",
          "expanded-probability-0010",
          "expanded-probability-0272",
          "expanded-probability-0034",
          "expanded-probability-0085",
          "expanded-probability-0241",
          "expanded-probability-0191",
          "expanded-probability-0198",
          "expanded-probability-0216",
          "expanded-probability-0223",
          "expanded-probability-0165",
          "expanded-probability-0029",
          "expanded-probability-0219",
          "expanded-probability-0209",
          "expanded-probability-0036",
          "expanded-probability-0284",
          "expanded-probability-0047",
          "expanded-probability-0208",
          "expanded-probability-0046",
          "expanded-probability-0131",
          "expanded-probability-0094",
          "expanded-probability-0060",
          "expanded-probability-0289",
          "expanded-probability-0110",
          "expanded-probability-0294",
          "expanded-probability-0161",
          "expanded-probability-0038",
          "expanded-probability-0195",
          "expanded-probability-0063",
          "expanded-probability-0011",
          "expanded-probability-0140",
          "expanded-probability-0119",
          "expanded-probability-0107",
          "expanded-probability-0051",
          "expanded-probability-0189",
          "expanded-probability-0062",
          "expanded-probability-0115",
          "expanded-probability-0287",
          "expanded-probability-0299",
          "expanded-probability-0217",
          "expanded-probability-0173",
          "expanded-probability-0270",
          "expanded-probability-0175",
          "expanded-probability-0040",
          "expanded-probability-0068",
          "expanded-probability-0178",
          "expanded-probability-0184",
          "expanded-probability-0274",
          "expanded-probability-0065",
          "expanded-probability-0285",
          "expanded-probability-0156",
          "expanded-probability-0117",
          "expanded-probability-0174",
          "expanded-probability-0244",
          "expanded-probability-0015",
          "expanded-probability-0253",
          "expanded-probability-0004",
          "expanded-probability-0105",
          "expanded-probability-0143",
          "expanded-probability-0057",
          "expanded-probability-0028",
          "expanded-probability-0263",
          "expanded-probability-0012",
          "expanded-probability-0190",
          "expanded-probability-0205",
          "expanded-probability-0050",
          "expanded-probability-0187",
          "expanded-probability-0171",
          "expanded-probability-0215",
          "expanded-probability-0086",
          "expanded-probability-0258",
          "expanded-probability-0118",
          "expanded-probability-0089",
          "expanded-probability-0133",
          "expanded-probability-0101",
          "expanded-probability-0070",
          "expanded-probability-0150",
          "expanded-probability-0227",
          "expanded-probability-0295",
          "expanded-probability-0297",
          "expanded-probability-0097",
          "expanded-probability-0254",
          "expanded-probability-0298",
          "expanded-probability-0157",
          "expanded-probability-0202",
          "expanded-probability-0054",
          "expanded-probability-0194",
          "expanded-probability-0252",
          "expanded-probability-0100",
          "expanded-probability-0259",
          "expanded-probability-0179",
          "expanded-probability-0256",
          "expanded-probability-0279",
          "expanded-probability-0192",
          "expanded-probability-0172",
          "expanded-probability-0098",
          "expanded-probability-0149",
          "expanded-probability-0213",
          "expanded-probability-0035",
          "expanded-probability-0084",
          "expanded-probability-0214"
        ]
      }
    },
    "timing_case_points": {
      "boolq": [
        {
          "case_id": "boolq-dev-0221",
          "Jev": 0.25169820900191553,
          "Luna": 1.2845857919892296
        },
        {
          "case_id": "boolq-dev-0519",
          "Jev": 0.24054708299809135,
          "Luna": 1.157200790999923
        },
        {
          "case_id": "boolq-dev-0715",
          "Jev": 0.23157149998587556,
          "Luna": 1.4557699160068296
        },
        {
          "case_id": "boolq-dev-0781",
          "Jev": 0.2534800409921445,
          "Luna": 1.1642412920191418
        },
        {
          "case_id": "boolq-dev-1124",
          "Jev": 0.2412845000217203,
          "Luna": 1.063191374996677
        },
        {
          "case_id": "boolq-dev-1240",
          "Jev": 0.2520102499984205,
          "Luna": 1.2243095420126338
        },
        {
          "case_id": "boolq-dev-1609",
          "Jev": 0.23422570902039297,
          "Luna": 1.325058792019263
        },
        {
          "case_id": "boolq-dev-2142",
          "Jev": 0.22483870800351724,
          "Luna": 0.9864088330068626
        },
        {
          "case_id": "boolq-dev-2750",
          "Jev": 0.22290137500385754,
          "Luna": 1.0341606670117471
        },
        {
          "case_id": "boolq-dev-2890",
          "Jev": 0.24011887499364093,
          "Luna": 1.1334215420065448
        },
        {
          "case_id": "boolq-dev-3044",
          "Jev": 0.2700931250001304,
          "Luna": 1.6243063339788932
        },
        {
          "case_id": "boolq-dev-3090",
          "Jev": 0.22763195799780078,
          "Luna": 1.1052920839865692
        }
      ],
      "policy": [
        {
          "case_id": "expanded-policy-0049",
          "Jev": 0.2810377910209354,
          "Luna": 1.2475500000000466
        },
        {
          "case_id": "expanded-policy-0075",
          "Jev": 0.24832725001033396,
          "Luna": 1.05442454200238
        },
        {
          "case_id": "expanded-policy-0092",
          "Jev": 0.29515929101035,
          "Luna": 1.2884278339915909
        },
        {
          "case_id": "expanded-policy-0107",
          "Jev": 0.2347486249927897,
          "Luna": 1.3038568749907427
        },
        {
          "case_id": "expanded-policy-0110",
          "Jev": 0.2282464159943629,
          "Luna": 1.098671290994389
        },
        {
          "case_id": "expanded-policy-0128",
          "Jev": 0.24591062500257976,
          "Luna": 1.0194633750070352
        },
        {
          "case_id": "expanded-policy-0134",
          "Jev": 0.23933391601894982,
          "Luna": 1.0143275829905178
        },
        {
          "case_id": "expanded-policy-0137",
          "Jev": 0.27156699998886324,
          "Luna": 1.3740167500218377
        },
        {
          "case_id": "expanded-policy-0185",
          "Jev": 0.2281115840014536,
          "Luna": 1.069131333992118
        },
        {
          "case_id": "expanded-policy-0202",
          "Jev": 0.23152870900230482,
          "Luna": 1.273702125006821
        },
        {
          "case_id": "expanded-policy-0204",
          "Jev": 0.25338991699391045,
          "Luna": 1.3299457920074929
        },
        {
          "case_id": "expanded-policy-0244",
          "Jev": 0.24512266600504518,
          "Luna": 1.2077263749961276
        }
      ],
      "probability": [
        {
          "case_id": "expanded-probability-0032",
          "Jev": 0.22996724999393336,
          "Luna": 1.39172829099698
        },
        {
          "case_id": "expanded-probability-0038",
          "Jev": 0.24057024999638088,
          "Luna": 1.0481641670048703
        },
        {
          "case_id": "expanded-probability-0076",
          "Jev": 0.23776570800691843,
          "Luna": 1.4567598749999888
        },
        {
          "case_id": "expanded-probability-0080",
          "Jev": 0.22648120901430957,
          "Luna": 1.2807568330026697
        },
        {
          "case_id": "expanded-probability-0098",
          "Jev": 0.2663675830117427,
          "Luna": 1.1293373329972383
        },
        {
          "case_id": "expanded-probability-0113",
          "Jev": 0.23824504102231003,
          "Luna": 1.1270600830030162
        },
        {
          "case_id": "expanded-probability-0118",
          "Jev": 0.25837749999482185,
          "Luna": 1.085922500002198
        },
        {
          "case_id": "expanded-probability-0127",
          "Jev": 0.24378150000120513,
          "Luna": 1.1287698749802075
        },
        {
          "case_id": "expanded-probability-0143",
          "Jev": 0.2761282499996014,
          "Luna": 1.2286522909998894
        },
        {
          "case_id": "expanded-probability-0190",
          "Jev": 0.22918841600767337,
          "Luna": 1.1643393330159597
        },
        {
          "case_id": "expanded-probability-0216",
          "Jev": 0.24926133299595676,
          "Luna": 1.3274669169913977
        },
        {
          "case_id": "expanded-probability-0219",
          "Jev": 0.22696479101432487,
          "Luna": 1.4273357919882983
        }
      ],
      "rte": [
        {
          "case_id": "rte-validation-0009",
          "Jev": 0.23938729098881595,
          "Luna": 1.284242624999024
        },
        {
          "case_id": "rte-validation-0011",
          "Jev": 0.22858245801762678,
          "Luna": 1.4206044170132373
        },
        {
          "case_id": "rte-validation-0064",
          "Jev": 0.23325454202131368,
          "Luna": 0.9846211659896653
        },
        {
          "case_id": "rte-validation-0095",
          "Jev": 0.24200683299568482,
          "Luna": 1.168572166992817
        },
        {
          "case_id": "rte-validation-0130",
          "Jev": 0.25848716698237695,
          "Luna": 1.2075708339980338
        },
        {
          "case_id": "rte-validation-0143",
          "Jev": 0.23283624998293817,
          "Luna": 1.2101580000016838
        },
        {
          "case_id": "rte-validation-0148",
          "Jev": 0.270281042001443,
          "Luna": 1.3711336669803131
        },
        {
          "case_id": "rte-validation-0167",
          "Jev": 0.2595445000042673,
          "Luna": 1.424521334003657
        },
        {
          "case_id": "rte-validation-0170",
          "Jev": 0.26847645902307704,
          "Luna": 1.329112334002275
        },
        {
          "case_id": "rte-validation-0188",
          "Jev": 0.2563239169830922,
          "Luna": 1.1399602919991594
        },
        {
          "case_id": "rte-validation-0198",
          "Jev": 0.24650837501394562,
          "Luna": 1.2521744169935118
        },
        {
          "case_id": "rte-validation-0220",
          "Jev": 0.23679791600443423,
          "Luna": 1.2734092500177212
        }
      ],
      "wic": [
        {
          "case_id": "wic-validation-0031",
          "Jev": 0.23972412501461804,
          "Luna": 1.272503832995426
        },
        {
          "case_id": "wic-validation-0057",
          "Jev": 0.23112724997918122,
          "Luna": 1.1100165419920813
        },
        {
          "case_id": "wic-validation-0120",
          "Jev": 0.22684354198281653,
          "Luna": 1.1431510410038754
        },
        {
          "case_id": "wic-validation-0134",
          "Jev": 0.2368912090023514,
          "Luna": 1.1289562500023749
        },
        {
          "case_id": "wic-validation-0214",
          "Jev": 0.22400670900242403,
          "Luna": 1.4311250419996213
        },
        {
          "case_id": "wic-validation-0223",
          "Jev": 0.24040924999280833,
          "Luna": 1.1144308329967316
        },
        {
          "case_id": "wic-validation-0297",
          "Jev": 0.24664945900440216,
          "Luna": 1.0126336249813903
        },
        {
          "case_id": "wic-validation-0364",
          "Jev": 0.2517865829868242,
          "Luna": 1.1085810840013437
        },
        {
          "case_id": "wic-validation-0383",
          "Jev": 0.2194919169996865,
          "Luna": 1.1375581660249736
        },
        {
          "case_id": "wic-validation-0455",
          "Jev": 0.2566521250118967,
          "Luna": 1.292783832992427
        },
        {
          "case_id": "wic-validation-0566",
          "Jev": 0.24260079200030304,
          "Luna": 1.1629298749903683
        },
        {
          "case_id": "wic-validation-0625",
          "Jev": 0.23573424998903647,
          "Luna": 1.3218399159959517
        }
      ]
    },
    "cache_cost_by_block": [
      {
        "block": 0,
        "arm": "Jev",
        "measured_cost_per_1000_usd": 0.13468875,
        "allocated_prime_cost_per_1000_usd": 0
      },
      {
        "block": 0,
        "arm": "Luna uncached",
        "measured_cost_per_1000_usd": 0.30266666666666664,
        "allocated_prime_cost_per_1000_usd": 0
      },
      {
        "block": 0,
        "arm": "Luna cached",
        "measured_cost_per_1000_usd": 0.07173666666666666,
        "allocated_prime_cost_per_1000_usd": 0.015261458333333335
      },
      {
        "block": 1,
        "arm": "Jev",
        "measured_cost_per_1000_usd": 0.13482524999999998,
        "allocated_prime_cost_per_1000_usd": 0
      },
      {
        "block": 1,
        "arm": "Luna uncached",
        "measured_cost_per_1000_usd": 0.30244166666666666,
        "allocated_prime_cost_per_1000_usd": 0
      },
      {
        "block": 1,
        "arm": "Luna cached",
        "measured_cost_per_1000_usd": 0.07169166666666665,
        "allocated_prime_cost_per_1000_usd": 0.015251041666666666
      },
      {
        "block": 2,
        "arm": "Jev",
        "measured_cost_per_1000_usd": 0.13476575,
        "allocated_prime_cost_per_1000_usd": 0
      },
      {
        "block": 2,
        "arm": "Luna uncached",
        "measured_cost_per_1000_usd": 0.30265,
        "allocated_prime_cost_per_1000_usd": 0
      },
      {
        "block": 2,
        "arm": "Luna cached",
        "measured_cost_per_1000_usd": 0.07171999999999999,
        "allocated_prime_cost_per_1000_usd": 0.015261458333333335
      },
      {
        "block": 3,
        "arm": "Jev",
        "measured_cost_per_1000_usd": 0.13479025,
        "allocated_prime_cost_per_1000_usd": 0
      },
      {
        "block": 3,
        "arm": "Luna uncached",
        "measured_cost_per_1000_usd": 0.303075,
        "allocated_prime_cost_per_1000_usd": 0
      },
      {
        "block": 3,
        "arm": "Luna cached",
        "measured_cost_per_1000_usd": 0.071785,
        "allocated_prime_cost_per_1000_usd": 0.015282291666666666
      }
    ],
    "cache_latency_by_block": [
      {
        "block": 0,
        "arm": "Jev",
        "median_latency_s": 0.2450790205184603,
        "observations": [
          0.24539658302091993,
          0.23788479101494886,
          0.23040612498880364,
          0.2342282079916913,
          0.26152791699860245,
          0.25977641699137166,
          0.2635777919786051,
          0.2447614580160007,
          0.24438258400186896,
          0.24433287500869483,
          0.2584773749986198,
          0.7945039999904111,
          0.47106525002163835,
          0.25936816600733437,
          0.2361520419945009,
          0.23159608399146236,
          0.29143133299658075,
          0.27733820799039677,
          0.27556783400359564,
          0.27056212499155663,
          0.22641379202832468,
          0.23605645800125785,
          0.23936579198925756,
          0.24125399999320507
        ]
      },
      {
        "block": 1,
        "arm": "Jev",
        "median_latency_s": 0.25583479198394343,
        "observations": [
          0.23981179200927727,
          0.2376095420040656,
          0.2381258330133278,
          0.25884262498584576,
          0.23900504200719297,
          0.2324664999905508,
          0.24018854199675843,
          0.2339921249949839,
          0.43054199998732656,
          0.28614612499950454,
          0.29595262501970865,
          0.2343116250121966,
          0.3339771250030026,
          0.2541687499906402,
          0.2430049999966286,
          0.256192499975441,
          0.26261933299247175,
          0.2594481250271201,
          0.3370368749892805,
          0.23708558297948912,
          0.278676624991931,
          0.2624618329864461,
          0.25547708399244584,
          0.42461395799182355
        ]
      },
      {
        "block": 2,
        "arm": "Jev",
        "median_latency_s": 0.2504945209948346,
        "observations": [
          0.2811455000191927,
          0.23107629199512303,
          0.26945683400845155,
          0.2652759159973357,
          0.23915808400488459,
          0.25629699998535216,
          0.3019130419997964,
          0.24813699998776428,
          0.2994925419916399,
          0.25870945799397305,
          0.2400733329996001,
          0.27837033302057534,
          0.24110833299346268,
          0.2773529169789981,
          0.23066383399418555,
          0.23761845898116007,
          0.24324462498771027,
          0.24190537497634068,
          0.2322770830069203,
          0.23828433401649818,
          0.2410115829843562,
          0.2631688749825116,
          0.3199245000141673,
          0.2528520420019049
        ]
      },
      {
        "block": 3,
        "arm": "Jev",
        "median_latency_s": 0.2538313749973895,
        "observations": [
          0.24112416699063033,
          0.40450112501275726,
          0.23738662502728403,
          0.2697485419921577,
          0.3112395829812158,
          0.23599508401821367,
          0.2750560830172617,
          0.2337047919863835,
          0.2333839579951018,
          0.2515162500203587,
          0.23262387499562465,
          0.3047420410148334,
          0.24152020897599868,
          0.23324408297776245,
          0.2645022080105264,
          0.24286458300775848,
          0.24783433400443755,
          0.32362183299846947,
          0.25855687499279156,
          0.448163166991435,
          0.28264683397719637,
          0.2422857500205282,
          0.2561464999744203,
          0.28064424998592585
        ]
      },
      {
        "block": 0,
        "arm": "Luna uncached",
        "median_latency_s": 1.1054907084908336,
        "observations": [
          1.9841262080008164,
          1.131418166012736,
          1.6521988750027958,
          1.083750791993225,
          1.1089782920025755,
          1.023326167021878,
          1.07623291600612,
          1.074110625020694,
          1.2577742080029566,
          1.0860667090164497,
          1.0845622089982498,
          1.9354566670081113,
          0.9811242090072483,
          1.2466680410143454,
          1.102634874987416,
          1.069755709002493,
          1.1083465419942513,
          1.0190802500001155,
          1.2650049169897102,
          1.1006010409910232,
          1.0889721670246217,
          1.1883936249942053,
          1.1107912080187816,
          1.270560208009556
        ]
      },
      {
        "block": 1,
        "arm": "Luna uncached",
        "median_latency_s": 0.9795645830017747,
        "observations": [
          0.9286506250209641,
          1.0849806669866666,
          0.8395205000124406,
          1.149269124987768,
          0.9618379580206238,
          1.00926316701225,
          0.9338597909954842,
          0.9043411249876954,
          2.511469917022623,
          0.9933747920149472,
          1.0351697920123115,
          0.9278682919975836,
          0.9909200830152258,
          0.9104973750072531,
          0.8985622089821845,
          1.1559770409949124,
          1.3179980409913696,
          0.9682090829883236,
          1.012561541981995,
          2.5922650409920607,
          0.9289074169937521,
          0.8796593750012107,
          1.0165543749753851,
          0.8931765000161249
        ]
      },
      {
        "block": 2,
        "arm": "Luna uncached",
        "median_latency_s": 1.1332991039962508,
        "observations": [
          1.0381244579912163,
          1.0306172909913585,
          1.305153249995783,
          1.0647129170247354,
          1.3414884580124635,
          1.1088151669828221,
          1.001708249998046,
          1.1292876670195255,
          1.2606851249875035,
          1.0585155419830699,
          1.339511666999897,
          1.2498765000200365,
          0.9690516669943463,
          1.0080685000284575,
          2.0104987919912674,
          1.2520691670069937,
          1.0064260830113199,
          1.267900417005876,
          1.3365647919999901,
          1.1858831250283401,
          1.137310540972976,
          1.2747337079781573,
          1.1195639590150677,
          1.0458953749912325
        ]
      },
      {
        "block": 3,
        "arm": "Luna uncached",
        "median_latency_s": 1.1306785625056364,
        "observations": [
          1.2666927919781301,
          1.2717372920014895,
          1.2778859579993878,
          1.1251040000061039,
          0.9489551669976208,
          1.0987611669988837,
          1.2940932089986745,
          1.1362531250051688,
          1.2030409160070121,
          1.0090238330012653,
          1.2194578330090735,
          1.2785679170046933,
          1.1566200830275193,
          1.4958669590123463,
          1.0740031669847667,
          1.0187090419931337,
          1.0646702089870814,
          1.1687544579908717,
          1.0253124999871943,
          1.45644895802252,
          1.07929300001706,
          1.042271874990547,
          0.9755440830194857,
          1.0604597909841686
        ]
      },
      {
        "block": 0,
        "arm": "Luna cached",
        "median_latency_s": 0.9378486454952508,
        "observations": [
          0.9139409999770578,
          0.8663207499776036,
          1.1211672910139896,
          0.9179877080023289,
          1.0301940840145107,
          0.8660317079920787,
          1.0057736669841688,
          1.335720915987622,
          1.020078667002963,
          0.8562467089795973,
          0.9356298749917187,
          0.8540836669853888,
          1.0179858329938725,
          1.055924040992977,
          0.9691529999836348,
          1.5338783749903087,
          0.8994435839995276,
          1.1487766249920242,
          0.8657527919858694,
          0.8601249579805881,
          0.9497549580119085,
          0.873281791980844,
          0.940067415998783,
          0.9017277089878917
        ]
      },
      {
        "block": 1,
        "arm": "Luna cached",
        "median_latency_s": 0.8966984790022252,
        "observations": [
          0.8542473750130739,
          0.877776334004011,
          1.2937770829885267,
          0.8315495419956278,
          0.8426342500024475,
          0.9668593329843134,
          0.9111149170203134,
          0.8754214580112603,
          0.811486334016081,
          0.8875342500105035,
          0.8815272089850623,
          1.1397044160112273,
          0.8876853749970905,
          0.9516281669784803,
          0.9982330420170911,
          0.9220799159957096,
          0.9460331669833977,
          1.0059623749984894,
          1.0285100410110317,
          1.590495042008115,
          0.8830672079930082,
          0.9057115830073599,
          0.8516503330029082,
          0.8632939999806695
        ]
      },
      {
        "block": 2,
        "arm": "Luna cached",
        "median_latency_s": 0.992883917002473,
        "observations": [
          0.8275123329949565,
          0.9205572089995258,
          0.9665574170066975,
          0.8923758330056444,
          1.490707040997222,
          1.0841012079908978,
          0.8594330420019105,
          0.9969987920194399,
          0.9201487080135848,
          1.0022358330024872,
          1.0300292079919018,
          1.073435791011434,
          0.9887690419855062,
          0.9831305419793352,
          0.823666125012096,
          0.8723171669989824,
          1.4144206659984775,
          1.1073606660065707,
          0.9369677909999155,
          1.4138185420015361,
          0.9336446660163347,
          1.1220345420006197,
          1.835956333001377,
          1.0083840419829357
        ]
      },
      {
        "block": 3,
        "arm": "Luna cached",
        "median_latency_s": 1.0119312500028173,
        "observations": [
          1.0185388329846319,
          1.0810444579983596,
          0.9570607499917969,
          1.6558541669801343,
          1.1400549160025548,
          0.9414322500233538,
          1.4891138749953825,
          1.0097248750098515,
          1.1112381250131875,
          1.69068962498568,
          0.9653789169969968,
          0.9801852910022717,
          0.939113875007024,
          1.2710592089861166,
          1.014137624995783,
          1.079126124997856,
          1.2560004999977536,
          0.9101734159921762,
          1.0608138750249054,
          0.9487537920067552,
          0.9677097920211963,
          0.8272430830111261,
          0.994844708009623,
          0.9579493339988403
        ]
      }
    ],
    "cache_verification": {
      "cold_primed_blocks": 4,
      "blocks_with_all_24_cache_hits": 4,
      "blocks_with_verified_uncached_control": 4,
      "fully_verified": true,
      "prime_count": 4,
      "priming_cost": {
        "total_usd": 0.00146535,
        "known_subtotal_usd": 0.00146535,
        "missing_n": 0
      },
      "setup_inclusive_cached_cost_usd": 0.00835175,
      "setup_inclusive_cached_cost_per_1000_usd": 0.08699739583333334
    }
  }
}
