{
  "summary_sha256": "686213b31fd138f5007e1876f4431a77db0ab8e0b9b9e230c1cb2689fdc39fc8",
  "models": [
    "Jev",
    "Luna",
    "Qwen3.8 Flash",
    "DeepSeek V4.1 Flash",
    "MiMo V2.6 Flash",
    "Hy4 Preview"
  ],
  "quality_sources": {
    "benchmark/expanded/runs/quality-20261002T093341Z/responses.jsonl": "c508f6173a27bed9834a294329414cf499bdc078381eb3f41e4d66c08f9f3f8b",
    "benchmark/multimodel/runs/quality-20261002T155349573021Z/responses.jsonl": "a0bc839aaf80fb359b48e808962f21249bde21ef00d058ce163a6bab6ca750b6",
    "benchmark/multimodel/runs/quality-20261002T160544592288Z/responses.jsonl": "9b949a90dcc900c2228ef035987ae306a858c138eab153eb7cf4bb6056357a22"
  },
  "figures": {
    "quality": {
      "svg": "quality.svg",
      "png": "quality.png",
      "size_inches": [
        12.4,
        8.370000000000001
      ],
      "png_pixels": [
        3224,
        2176
      ],
      "title": "Binary-task quality",
      "caption": "Observed accuracy across attempted cases. Error bars show nominal Wilson 95% intervals, which do not incorporate passage clustering. Paired cluster contrasts are reported separately. Model quality observations may come from different collection windows.",
      "svg_sha256": "049b6ec50264e4c994fdc5e5d98629433d6e07b4b6bb56dd4f909f1d3c6dcdf9",
      "png_sha256": "aa2a209211cf3a747269c5707079f50e0122e3fabcb8f726ca53d015569a7b6a"
    },
    "probability": {
      "svg": "probability.svg",
      "png": "probability.png",
      "size_inches": [
        12.450000000000001,
        8.4
      ],
      "png_pixels": [
        3237,
        2184
      ],
      "title": "Exact-probability estimates",
      "caption": "Returned probabilities versus frozen mathematical targets. Each model uses its own valid-response denominator; invalid predictions remain in failure accounting. Points are not jittered. The paired MAE comparisons use common valid cases.",
      "svg_sha256": "1d4b6e70a744449ab98261140e2e99cdf0e242bd8985e86bf501e47e9d242703",
      "png_sha256": "9773251ee1c188928a15bdc94c068103fbd16f7f6ff1b3431228487cdb5a7a3c"
    },
    "timing": {
      "svg": "timing.svg",
      "png": "timing.png",
      "size_inches": [
        12.4,
        9.900000000000002
      ],
      "png_pixels": [
        3224,
        2574
      ],
      "title": "Interleaved complete-response latency",
      "caption": "Median of three observations for each case and model, then median across the same complete cases for every model. Only the new serial interleaved session is shown; no old or concurrent bulk latencies are pooled. Point summaries do not imply cross-day precision.",
      "svg_sha256": "388e4814cf88718028f2be7ffc1625bbfe256af5bd07829100936381d2f8c6da",
      "png_sha256": "30ce1f5d9a3b9ddec6ae3057824cbd5f287153d5c56f8b0dd1eafd2def4e4c96"
    },
    "cache-1": {
      "svg": "cache-1.svg",
      "png": "cache-1.png",
      "size_inches": [
        12.4,
        6.3100000000000005
      ],
      "png_pixels": [
        3224,
        1641
      ],
      "title": "Long-prefix service costs and latency",
      "caption": "Measured cost and latency within one cache session. Hatched increments allocate the separate primes to the measured cached decisions. Models without verified cache controls are uncontrolled service baselines. Session results and priming costs are not pooled across measurement dates.",
      "svg_sha256": "c0c0f5ed1e753e693c2527d0d34ade78074163c6f7b68107891160565106afac",
      "png_sha256": "be94451e4f1959db21013bd373b9c42ca4d7fd43d386efe19439e0a1ea0a8043"
    }
  },
  "metrics": {
    "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
        },
        "Qwen3.8 Flash": {
          "accuracy": 0.881651376146789,
          "wilson95": [
            0.8701295492173265,
            0.8922775591889281
          ],
          "correct": 2883,
          "attempts": 3270,
          "invalid": 74
        },
        "DeepSeek V4.1 Flash": {
          "accuracy": 0.8746177370030581,
          "wilson95": [
            0.8628261706084228,
            0.8855301657679089
          ],
          "correct": 2860,
          "attempts": 3270,
          "invalid": 0
        },
        "MiMo V2.6 Flash": {
          "accuracy": 0.8889908256880734,
          "wilson95": [
            0.8777638470589225,
            0.899304936508991
          ],
          "correct": 2907,
          "attempts": 3270,
          "invalid": 0
        },
        "Hy4 Preview": {
          "accuracy": 0.8896024464831804,
          "wilson95": [
            0.8784006906210592,
            0.8998898992103258
          ],
          "correct": 2909,
          "attempts": 3270,
          "invalid": 4
        }
      },
      "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
        },
        "Qwen3.8 Flash": {
          "accuracy": 0.8447653429602888,
          "wilson95": [
            0.7974351903182963,
            0.8826638264082396
          ],
          "correct": 234,
          "attempts": 277,
          "invalid": 7
        },
        "DeepSeek V4.1 Flash": {
          "accuracy": 0.851985559566787,
          "wilson95": [
            0.805360563813589,
            0.8889813642580053
          ],
          "correct": 236,
          "attempts": 277,
          "invalid": 0
        },
        "MiMo V2.6 Flash": {
          "accuracy": 0.8375451263537906,
          "wilson95": [
            0.789540911797291,
            0.8763151935841865
          ],
          "correct": 232,
          "attempts": 277,
          "invalid": 0
        },
        "Hy4 Preview": {
          "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
        },
        "Qwen3.8 Flash": {
          "accuracy": 0.731974921630094,
          "wilson95": [
            0.6962919010718486,
            0.7648811751327708
          ],
          "correct": 467,
          "attempts": 638,
          "invalid": 17
        },
        "DeepSeek V4.1 Flash": {
          "accuracy": 0.7445141065830722,
          "wilson95": [
            0.7092782498887528,
            0.7768231007053056
          ],
          "correct": 475,
          "attempts": 638,
          "invalid": 0
        },
        "MiMo V2.6 Flash": {
          "accuracy": 0.6959247648902821,
          "wilson95": [
            0.6591446418815946,
            0.7303596454533879
          ],
          "correct": 444,
          "attempts": 638,
          "invalid": 0
        },
        "Hy4 Preview": {
          "accuracy": 0.7210031347962382,
          "wilson95": [
            0.6849573750736453,
            0.754403461040215
          ],
          "correct": 460,
          "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
        },
        "Qwen3.8 Flash": {
          "accuracy": 0.9766666666666667,
          "wilson95": [
            0.9526280277317246,
            0.9886523397008747
          ],
          "correct": 293,
          "attempts": 300,
          "invalid": 7
        },
        "DeepSeek V4.1 Flash": {
          "accuracy": 0.9933333333333333,
          "wilson95": [
            0.9760224167496929,
            0.998169851642088
          ],
          "correct": 298,
          "attempts": 300,
          "invalid": 0
        },
        "MiMo V2.6 Flash": {
          "accuracy": 0.9866666666666667,
          "wilson95": [
            0.9662244697474675,
            0.9948030382606411
          ],
          "correct": 296,
          "attempts": 300,
          "invalid": 0
        },
        "Hy4 Preview": {
          "accuracy": 0.9533333333333334,
          "wilson95": [
            0.923202639772079,
            0.9720010663176659
          ],
          "correct": 286,
          "attempts": 300,
          "invalid": 0
        }
      }
    },
    "probability": {
      "Jev": {
        "valid": 299,
        "attempts": 300,
        "mae": 0.07640276740848177,
        "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": {
        "valid": 300,
        "attempts": 300,
        "mae": 0.024395889425565125,
        "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"
        ]
      },
      "Qwen3.8 Flash": {
        "valid": 291,
        "attempts": 300,
        "mae": 0.061031034066208786,
        "case_ids": [
          "expanded-probability-0214",
          "expanded-probability-0084",
          "expanded-probability-0035",
          "expanded-probability-0213",
          "expanded-probability-0098",
          "expanded-probability-0149",
          "expanded-probability-0172",
          "expanded-probability-0192",
          "expanded-probability-0279",
          "expanded-probability-0256",
          "expanded-probability-0259",
          "expanded-probability-0100",
          "expanded-probability-0252",
          "expanded-probability-0054",
          "expanded-probability-0179",
          "expanded-probability-0157",
          "expanded-probability-0298",
          "expanded-probability-0254",
          "expanded-probability-0097",
          "expanded-probability-0297",
          "expanded-probability-0295",
          "expanded-probability-0150",
          "expanded-probability-0227",
          "expanded-probability-0101",
          "expanded-probability-0070",
          "expanded-probability-0133",
          "expanded-probability-0089",
          "expanded-probability-0118",
          "expanded-probability-0258",
          "expanded-probability-0215",
          "expanded-probability-0171",
          "expanded-probability-0187",
          "expanded-probability-0086",
          "expanded-probability-0050",
          "expanded-probability-0205",
          "expanded-probability-0190",
          "expanded-probability-0012",
          "expanded-probability-0263",
          "expanded-probability-0057",
          "expanded-probability-0143",
          "expanded-probability-0028",
          "expanded-probability-0105",
          "expanded-probability-0004",
          "expanded-probability-0015",
          "expanded-probability-0117",
          "expanded-probability-0253",
          "expanded-probability-0065",
          "expanded-probability-0244",
          "expanded-probability-0156",
          "expanded-probability-0274",
          "expanded-probability-0178",
          "expanded-probability-0068",
          "expanded-probability-0285",
          "expanded-probability-0184",
          "expanded-probability-0040",
          "expanded-probability-0175",
          "expanded-probability-0270",
          "expanded-probability-0173",
          "expanded-probability-0217",
          "expanded-probability-0115",
          "expanded-probability-0287",
          "expanded-probability-0299",
          "expanded-probability-0062",
          "expanded-probability-0189",
          "expanded-probability-0051",
          "expanded-probability-0119",
          "expanded-probability-0140",
          "expanded-probability-0011",
          "expanded-probability-0063",
          "expanded-probability-0195",
          "expanded-probability-0107",
          "expanded-probability-0038",
          "expanded-probability-0294",
          "expanded-probability-0110",
          "expanded-probability-0161",
          "expanded-probability-0289",
          "expanded-probability-0094",
          "expanded-probability-0131",
          "expanded-probability-0046",
          "expanded-probability-0208",
          "expanded-probability-0047",
          "expanded-probability-0036",
          "expanded-probability-0284",
          "expanded-probability-0209",
          "expanded-probability-0060",
          "expanded-probability-0029",
          "expanded-probability-0165",
          "expanded-probability-0223",
          "expanded-probability-0198",
          "expanded-probability-0216",
          "expanded-probability-0219",
          "expanded-probability-0191",
          "expanded-probability-0085",
          "expanded-probability-0241",
          "expanded-probability-0034",
          "expanded-probability-0272",
          "expanded-probability-0010",
          "expanded-probability-0182",
          "expanded-probability-0032",
          "expanded-probability-0088",
          "expanded-probability-0064",
          "expanded-probability-0201",
          "expanded-probability-0290",
          "expanded-probability-0091",
          "expanded-probability-0177",
          "expanded-probability-0236",
          "expanded-probability-0099",
          "expanded-probability-0168",
          "expanded-probability-0078",
          "expanded-probability-0109",
          "expanded-probability-0264",
          "expanded-probability-0255",
          "expanded-probability-0122",
          "expanded-probability-0083",
          "expanded-probability-0239",
          "expanded-probability-0124",
          "expanded-probability-0296",
          "expanded-probability-0250",
          "expanded-probability-0037",
          "expanded-probability-0169",
          "expanded-probability-0016",
          "expanded-probability-0246",
          "expanded-probability-0074",
          "expanded-probability-0134",
          "expanded-probability-0116",
          "expanded-probability-0022",
          "expanded-probability-0238",
          "expanded-probability-0030",
          "expanded-probability-0147",
          "expanded-probability-0233",
          "expanded-probability-0053",
          "expanded-probability-0041",
          "expanded-probability-0077",
          "expanded-probability-0231",
          "expanded-probability-0009",
          "expanded-probability-0300",
          "expanded-probability-0230",
          "expanded-probability-0210",
          "expanded-probability-0121",
          "expanded-probability-0273",
          "expanded-probability-0129",
          "expanded-probability-0014",
          "expanded-probability-0007",
          "expanded-probability-0183",
          "expanded-probability-0261",
          "expanded-probability-0153",
          "expanded-probability-0095",
          "expanded-probability-0076",
          "expanded-probability-0082",
          "expanded-probability-0162",
          "expanded-probability-0243",
          "expanded-probability-0257",
          "expanded-probability-0020",
          "expanded-probability-0199",
          "expanded-probability-0142",
          "expanded-probability-0166",
          "expanded-probability-0188",
          "expanded-probability-0073",
          "expanded-probability-0102",
          "expanded-probability-0120",
          "expanded-probability-0158",
          "expanded-probability-0203",
          "expanded-probability-0251",
          "expanded-probability-0127",
          "expanded-probability-0043",
          "expanded-probability-0071",
          "expanded-probability-0197",
          "expanded-probability-0096",
          "expanded-probability-0126",
          "expanded-probability-0135",
          "expanded-probability-0221",
          "expanded-probability-0286",
          "expanded-probability-0021",
          "expanded-probability-0003",
          "expanded-probability-0282",
          "expanded-probability-0260",
          "expanded-probability-0033",
          "expanded-probability-0108",
          "expanded-probability-0087",
          "expanded-probability-0266",
          "expanded-probability-0155",
          "expanded-probability-0193",
          "expanded-probability-0061",
          "expanded-probability-0160",
          "expanded-probability-0278",
          "expanded-probability-0248",
          "expanded-probability-0112",
          "expanded-probability-0044",
          "expanded-probability-0228",
          "expanded-probability-0023",
          "expanded-probability-0163",
          "expanded-probability-0039",
          "expanded-probability-0026",
          "expanded-probability-0066",
          "expanded-probability-0204",
          "expanded-probability-0144",
          "expanded-probability-0240",
          "expanded-probability-0167",
          "expanded-probability-0103",
          "expanded-probability-0292",
          "expanded-probability-0265",
          "expanded-probability-0164",
          "expanded-probability-0125",
          "expanded-probability-0052",
          "expanded-probability-0146",
          "expanded-probability-0019",
          "expanded-probability-0027",
          "expanded-probability-0242",
          "expanded-probability-0056",
          "expanded-probability-0079",
          "expanded-probability-0081",
          "expanded-probability-0045",
          "expanded-probability-0207",
          "expanded-probability-0170",
          "expanded-probability-0269",
          "expanded-probability-0235",
          "expanded-probability-0137",
          "expanded-probability-0132",
          "expanded-probability-0024",
          "expanded-probability-0139",
          "expanded-probability-0001",
          "expanded-probability-0018",
          "expanded-probability-0106",
          "expanded-probability-0111",
          "expanded-probability-0013",
          "expanded-probability-0200",
          "expanded-probability-0092",
          "expanded-probability-0212",
          "expanded-probability-0249",
          "expanded-probability-0042",
          "expanded-probability-0113",
          "expanded-probability-0176",
          "expanded-probability-0031",
          "expanded-probability-0211",
          "expanded-probability-0017",
          "expanded-probability-0058",
          "expanded-probability-0196",
          "expanded-probability-0104",
          "expanded-probability-0293",
          "expanded-probability-0141",
          "expanded-probability-0267",
          "expanded-probability-0291",
          "expanded-probability-0224",
          "expanded-probability-0225",
          "expanded-probability-0234",
          "expanded-probability-0090",
          "expanded-probability-0280",
          "expanded-probability-0148",
          "expanded-probability-0123",
          "expanded-probability-0226",
          "expanded-probability-0222",
          "expanded-probability-0281",
          "expanded-probability-0072",
          "expanded-probability-0268",
          "expanded-probability-0138",
          "expanded-probability-0005",
          "expanded-probability-0247",
          "expanded-probability-0067",
          "expanded-probability-0128",
          "expanded-probability-0276",
          "expanded-probability-0232",
          "expanded-probability-0154",
          "expanded-probability-0218",
          "expanded-probability-0181",
          "expanded-probability-0151",
          "expanded-probability-0048",
          "expanded-probability-0093",
          "expanded-probability-0229",
          "expanded-probability-0283",
          "expanded-probability-0288",
          "expanded-probability-0136",
          "expanded-probability-0245",
          "expanded-probability-0271",
          "expanded-probability-0277",
          "expanded-probability-0159",
          "expanded-probability-0186",
          "expanded-probability-0049",
          "expanded-probability-0006",
          "expanded-probability-0080",
          "expanded-probability-0180",
          "expanded-probability-0075",
          "expanded-probability-0237",
          "expanded-probability-0185",
          "expanded-probability-0275",
          "expanded-probability-0069",
          "expanded-probability-0008",
          "expanded-probability-0262",
          "expanded-probability-0152",
          "expanded-probability-0130",
          "expanded-probability-0145",
          "expanded-probability-0206"
        ]
      },
      "DeepSeek V4.1 Flash": {
        "valid": 300,
        "attempts": 300,
        "mae": 0.03834838842462868,
        "case_ids": [
          "expanded-probability-0214",
          "expanded-probability-0084",
          "expanded-probability-0035",
          "expanded-probability-0213",
          "expanded-probability-0149",
          "expanded-probability-0098",
          "expanded-probability-0172",
          "expanded-probability-0192",
          "expanded-probability-0279",
          "expanded-probability-0259",
          "expanded-probability-0179",
          "expanded-probability-0100",
          "expanded-probability-0256",
          "expanded-probability-0252",
          "expanded-probability-0054",
          "expanded-probability-0194",
          "expanded-probability-0202",
          "expanded-probability-0157",
          "expanded-probability-0298",
          "expanded-probability-0254",
          "expanded-probability-0097",
          "expanded-probability-0297",
          "expanded-probability-0295",
          "expanded-probability-0150",
          "expanded-probability-0227",
          "expanded-probability-0070",
          "expanded-probability-0133",
          "expanded-probability-0101",
          "expanded-probability-0089",
          "expanded-probability-0118",
          "expanded-probability-0258",
          "expanded-probability-0215",
          "expanded-probability-0086",
          "expanded-probability-0187",
          "expanded-probability-0171",
          "expanded-probability-0050",
          "expanded-probability-0205",
          "expanded-probability-0190",
          "expanded-probability-0012",
          "expanded-probability-0263",
          "expanded-probability-0028",
          "expanded-probability-0057",
          "expanded-probability-0143",
          "expanded-probability-0105",
          "expanded-probability-0004",
          "expanded-probability-0253",
          "expanded-probability-0015",
          "expanded-probability-0244",
          "expanded-probability-0174",
          "expanded-probability-0156",
          "expanded-probability-0117",
          "expanded-probability-0285",
          "expanded-probability-0065",
          "expanded-probability-0184",
          "expanded-probability-0274",
          "expanded-probability-0178",
          "expanded-probability-0068",
          "expanded-probability-0040",
          "expanded-probability-0175",
          "expanded-probability-0217",
          "expanded-probability-0173",
          "expanded-probability-0270",
          "expanded-probability-0115",
          "expanded-probability-0299",
          "expanded-probability-0287",
          "expanded-probability-0062",
          "expanded-probability-0189",
          "expanded-probability-0119",
          "expanded-probability-0107",
          "expanded-probability-0051",
          "expanded-probability-0011",
          "expanded-probability-0063",
          "expanded-probability-0140",
          "expanded-probability-0195",
          "expanded-probability-0038",
          "expanded-probability-0161",
          "expanded-probability-0294",
          "expanded-probability-0110",
          "expanded-probability-0289",
          "expanded-probability-0094",
          "expanded-probability-0131",
          "expanded-probability-0046",
          "expanded-probability-0208",
          "expanded-probability-0047",
          "expanded-probability-0036",
          "expanded-probability-0284",
          "expanded-probability-0060",
          "expanded-probability-0209",
          "expanded-probability-0219",
          "expanded-probability-0029",
          "expanded-probability-0165",
          "expanded-probability-0223",
          "expanded-probability-0216",
          "expanded-probability-0198",
          "expanded-probability-0191",
          "expanded-probability-0241",
          "expanded-probability-0085",
          "expanded-probability-0034",
          "expanded-probability-0272",
          "expanded-probability-0010",
          "expanded-probability-0182",
          "expanded-probability-0290",
          "expanded-probability-0032",
          "expanded-probability-0064",
          "expanded-probability-0088",
          "expanded-probability-0201",
          "expanded-probability-0091",
          "expanded-probability-0078",
          "expanded-probability-0177",
          "expanded-probability-0236",
          "expanded-probability-0168",
          "expanded-probability-0099",
          "expanded-probability-0109",
          "expanded-probability-0264",
          "expanded-probability-0255",
          "expanded-probability-0122",
          "expanded-probability-0083",
          "expanded-probability-0239",
          "expanded-probability-0296",
          "expanded-probability-0124",
          "expanded-probability-0250",
          "expanded-probability-0037",
          "expanded-probability-0059",
          "expanded-probability-0169",
          "expanded-probability-0074",
          "expanded-probability-0246",
          "expanded-probability-0016",
          "expanded-probability-0134",
          "expanded-probability-0238",
          "expanded-probability-0116",
          "expanded-probability-0022",
          "expanded-probability-0030",
          "expanded-probability-0233",
          "expanded-probability-0147",
          "expanded-probability-0053",
          "expanded-probability-0231",
          "expanded-probability-0041",
          "expanded-probability-0077",
          "expanded-probability-0009",
          "expanded-probability-0300",
          "expanded-probability-0230",
          "expanded-probability-0121",
          "expanded-probability-0210",
          "expanded-probability-0273",
          "expanded-probability-0002",
          "expanded-probability-0129",
          "expanded-probability-0014",
          "expanded-probability-0007",
          "expanded-probability-0183",
          "expanded-probability-0261",
          "expanded-probability-0153",
          "expanded-probability-0095",
          "expanded-probability-0076",
          "expanded-probability-0257",
          "expanded-probability-0243",
          "expanded-probability-0082",
          "expanded-probability-0162",
          "expanded-probability-0020",
          "expanded-probability-0199",
          "expanded-probability-0142",
          "expanded-probability-0166",
          "expanded-probability-0188",
          "expanded-probability-0073",
          "expanded-probability-0120",
          "expanded-probability-0102",
          "expanded-probability-0203",
          "expanded-probability-0158",
          "expanded-probability-0251",
          "expanded-probability-0127",
          "expanded-probability-0043",
          "expanded-probability-0071",
          "expanded-probability-0197",
          "expanded-probability-0096",
          "expanded-probability-0126",
          "expanded-probability-0286",
          "expanded-probability-0135",
          "expanded-probability-0221",
          "expanded-probability-0260",
          "expanded-probability-0021",
          "expanded-probability-0282",
          "expanded-probability-0003",
          "expanded-probability-0108",
          "expanded-probability-0033",
          "expanded-probability-0087",
          "expanded-probability-0266",
          "expanded-probability-0155",
          "expanded-probability-0193",
          "expanded-probability-0061",
          "expanded-probability-0160",
          "expanded-probability-0278",
          "expanded-probability-0112",
          "expanded-probability-0248",
          "expanded-probability-0044",
          "expanded-probability-0228",
          "expanded-probability-0023",
          "expanded-probability-0163",
          "expanded-probability-0039",
          "expanded-probability-0026",
          "expanded-probability-0066",
          "expanded-probability-0204",
          "expanded-probability-0240",
          "expanded-probability-0167",
          "expanded-probability-0144",
          "expanded-probability-0292",
          "expanded-probability-0103",
          "expanded-probability-0265",
          "expanded-probability-0164",
          "expanded-probability-0019",
          "expanded-probability-0125",
          "expanded-probability-0052",
          "expanded-probability-0027",
          "expanded-probability-0242",
          "expanded-probability-0056",
          "expanded-probability-0146",
          "expanded-probability-0081",
          "expanded-probability-0079",
          "expanded-probability-0045",
          "expanded-probability-0207",
          "expanded-probability-0170",
          "expanded-probability-0269",
          "expanded-probability-0235",
          "expanded-probability-0137",
          "expanded-probability-0132",
          "expanded-probability-0024",
          "expanded-probability-0139",
          "expanded-probability-0001",
          "expanded-probability-0018",
          "expanded-probability-0111",
          "expanded-probability-0106",
          "expanded-probability-0013",
          "expanded-probability-0200",
          "expanded-probability-0092",
          "expanded-probability-0212",
          "expanded-probability-0055",
          "expanded-probability-0114",
          "expanded-probability-0042",
          "expanded-probability-0249",
          "expanded-probability-0113",
          "expanded-probability-0176",
          "expanded-probability-0031",
          "expanded-probability-0211",
          "expanded-probability-0017",
          "expanded-probability-0196",
          "expanded-probability-0104",
          "expanded-probability-0058",
          "expanded-probability-0293",
          "expanded-probability-0025",
          "expanded-probability-0141",
          "expanded-probability-0267",
          "expanded-probability-0291",
          "expanded-probability-0225",
          "expanded-probability-0224",
          "expanded-probability-0234",
          "expanded-probability-0090",
          "expanded-probability-0280",
          "expanded-probability-0148",
          "expanded-probability-0123",
          "expanded-probability-0226",
          "expanded-probability-0222",
          "expanded-probability-0281",
          "expanded-probability-0220",
          "expanded-probability-0268",
          "expanded-probability-0072",
          "expanded-probability-0138",
          "expanded-probability-0247",
          "expanded-probability-0005",
          "expanded-probability-0067",
          "expanded-probability-0128",
          "expanded-probability-0276",
          "expanded-probability-0232",
          "expanded-probability-0154",
          "expanded-probability-0181",
          "expanded-probability-0218",
          "expanded-probability-0151",
          "expanded-probability-0048",
          "expanded-probability-0093",
          "expanded-probability-0229",
          "expanded-probability-0283",
          "expanded-probability-0288",
          "expanded-probability-0136",
          "expanded-probability-0271",
          "expanded-probability-0245",
          "expanded-probability-0277",
          "expanded-probability-0159",
          "expanded-probability-0186",
          "expanded-probability-0049",
          "expanded-probability-0006",
          "expanded-probability-0080",
          "expanded-probability-0180",
          "expanded-probability-0075",
          "expanded-probability-0185",
          "expanded-probability-0237",
          "expanded-probability-0275",
          "expanded-probability-0069",
          "expanded-probability-0008",
          "expanded-probability-0262",
          "expanded-probability-0130",
          "expanded-probability-0152",
          "expanded-probability-0206",
          "expanded-probability-0145"
        ]
      },
      "MiMo V2.6 Flash": {
        "valid": 300,
        "attempts": 300,
        "mae": 0.05892329014672775,
        "case_ids": [
          "expanded-probability-0214",
          "expanded-probability-0084",
          "expanded-probability-0035",
          "expanded-probability-0213",
          "expanded-probability-0149",
          "expanded-probability-0098",
          "expanded-probability-0172",
          "expanded-probability-0192",
          "expanded-probability-0279",
          "expanded-probability-0259",
          "expanded-probability-0179",
          "expanded-probability-0100",
          "expanded-probability-0256",
          "expanded-probability-0252",
          "expanded-probability-0194",
          "expanded-probability-0054",
          "expanded-probability-0202",
          "expanded-probability-0157",
          "expanded-probability-0298",
          "expanded-probability-0254",
          "expanded-probability-0097",
          "expanded-probability-0297",
          "expanded-probability-0295",
          "expanded-probability-0227",
          "expanded-probability-0150",
          "expanded-probability-0070",
          "expanded-probability-0133",
          "expanded-probability-0101",
          "expanded-probability-0089",
          "expanded-probability-0118",
          "expanded-probability-0258",
          "expanded-probability-0215",
          "expanded-probability-0086",
          "expanded-probability-0171",
          "expanded-probability-0187",
          "expanded-probability-0050",
          "expanded-probability-0205",
          "expanded-probability-0190",
          "expanded-probability-0012",
          "expanded-probability-0057",
          "expanded-probability-0028",
          "expanded-probability-0263",
          "expanded-probability-0143",
          "expanded-probability-0105",
          "expanded-probability-0004",
          "expanded-probability-0253",
          "expanded-probability-0015",
          "expanded-probability-0244",
          "expanded-probability-0174",
          "expanded-probability-0156",
          "expanded-probability-0117",
          "expanded-probability-0065",
          "expanded-probability-0285",
          "expanded-probability-0274",
          "expanded-probability-0184",
          "expanded-probability-0178",
          "expanded-probability-0068",
          "expanded-probability-0040",
          "expanded-probability-0175",
          "expanded-probability-0173",
          "expanded-probability-0217",
          "expanded-probability-0270",
          "expanded-probability-0299",
          "expanded-probability-0115",
          "expanded-probability-0287",
          "expanded-probability-0062",
          "expanded-probability-0189",
          "expanded-probability-0107",
          "expanded-probability-0119",
          "expanded-probability-0051",
          "expanded-probability-0011",
          "expanded-probability-0140",
          "expanded-probability-0063",
          "expanded-probability-0195",
          "expanded-probability-0038",
          "expanded-probability-0294",
          "expanded-probability-0110",
          "expanded-probability-0161",
          "expanded-probability-0131",
          "expanded-probability-0289",
          "expanded-probability-0094",
          "expanded-probability-0208",
          "expanded-probability-0046",
          "expanded-probability-0047",
          "expanded-probability-0284",
          "expanded-probability-0060",
          "expanded-probability-0036",
          "expanded-probability-0209",
          "expanded-probability-0219",
          "expanded-probability-0029",
          "expanded-probability-0165",
          "expanded-probability-0223",
          "expanded-probability-0216",
          "expanded-probability-0198",
          "expanded-probability-0191",
          "expanded-probability-0241",
          "expanded-probability-0085",
          "expanded-probability-0034",
          "expanded-probability-0010",
          "expanded-probability-0272",
          "expanded-probability-0032",
          "expanded-probability-0182",
          "expanded-probability-0088",
          "expanded-probability-0290",
          "expanded-probability-0201",
          "expanded-probability-0064",
          "expanded-probability-0078",
          "expanded-probability-0091",
          "expanded-probability-0236",
          "expanded-probability-0177",
          "expanded-probability-0168",
          "expanded-probability-0099",
          "expanded-probability-0264",
          "expanded-probability-0109",
          "expanded-probability-0255",
          "expanded-probability-0122",
          "expanded-probability-0083",
          "expanded-probability-0239",
          "expanded-probability-0124",
          "expanded-probability-0296",
          "expanded-probability-0250",
          "expanded-probability-0037",
          "expanded-probability-0169",
          "expanded-probability-0059",
          "expanded-probability-0246",
          "expanded-probability-0016",
          "expanded-probability-0134",
          "expanded-probability-0074",
          "expanded-probability-0238",
          "expanded-probability-0022",
          "expanded-probability-0030",
          "expanded-probability-0116",
          "expanded-probability-0233",
          "expanded-probability-0147",
          "expanded-probability-0053",
          "expanded-probability-0231",
          "expanded-probability-0041",
          "expanded-probability-0077",
          "expanded-probability-0009",
          "expanded-probability-0300",
          "expanded-probability-0230",
          "expanded-probability-0210",
          "expanded-probability-0121",
          "expanded-probability-0273",
          "expanded-probability-0002",
          "expanded-probability-0129",
          "expanded-probability-0014",
          "expanded-probability-0007",
          "expanded-probability-0183",
          "expanded-probability-0261",
          "expanded-probability-0153",
          "expanded-probability-0095",
          "expanded-probability-0076",
          "expanded-probability-0257",
          "expanded-probability-0082",
          "expanded-probability-0162",
          "expanded-probability-0243",
          "expanded-probability-0020",
          "expanded-probability-0199",
          "expanded-probability-0142",
          "expanded-probability-0166",
          "expanded-probability-0188",
          "expanded-probability-0073",
          "expanded-probability-0120",
          "expanded-probability-0102",
          "expanded-probability-0158",
          "expanded-probability-0203",
          "expanded-probability-0251",
          "expanded-probability-0127",
          "expanded-probability-0043",
          "expanded-probability-0071",
          "expanded-probability-0197",
          "expanded-probability-0096",
          "expanded-probability-0126",
          "expanded-probability-0286",
          "expanded-probability-0135",
          "expanded-probability-0221",
          "expanded-probability-0003",
          "expanded-probability-0260",
          "expanded-probability-0021",
          "expanded-probability-0282",
          "expanded-probability-0033",
          "expanded-probability-0108",
          "expanded-probability-0087",
          "expanded-probability-0266",
          "expanded-probability-0155",
          "expanded-probability-0193",
          "expanded-probability-0061",
          "expanded-probability-0160",
          "expanded-probability-0278",
          "expanded-probability-0248",
          "expanded-probability-0112",
          "expanded-probability-0044",
          "expanded-probability-0228",
          "expanded-probability-0023",
          "expanded-probability-0163",
          "expanded-probability-0039",
          "expanded-probability-0026",
          "expanded-probability-0066",
          "expanded-probability-0204",
          "expanded-probability-0144",
          "expanded-probability-0240",
          "expanded-probability-0103",
          "expanded-probability-0167",
          "expanded-probability-0265",
          "expanded-probability-0164",
          "expanded-probability-0292",
          "expanded-probability-0019",
          "expanded-probability-0125",
          "expanded-probability-0052",
          "expanded-probability-0242",
          "expanded-probability-0146",
          "expanded-probability-0056",
          "expanded-probability-0027",
          "expanded-probability-0081",
          "expanded-probability-0079",
          "expanded-probability-0045",
          "expanded-probability-0207",
          "expanded-probability-0170",
          "expanded-probability-0269",
          "expanded-probability-0235",
          "expanded-probability-0137",
          "expanded-probability-0132",
          "expanded-probability-0024",
          "expanded-probability-0139",
          "expanded-probability-0001",
          "expanded-probability-0018",
          "expanded-probability-0111",
          "expanded-probability-0106",
          "expanded-probability-0013",
          "expanded-probability-0200",
          "expanded-probability-0092",
          "expanded-probability-0212",
          "expanded-probability-0055",
          "expanded-probability-0114",
          "expanded-probability-0042",
          "expanded-probability-0249",
          "expanded-probability-0176",
          "expanded-probability-0031",
          "expanded-probability-0113",
          "expanded-probability-0211",
          "expanded-probability-0017",
          "expanded-probability-0104",
          "expanded-probability-0196",
          "expanded-probability-0058",
          "expanded-probability-0293",
          "expanded-probability-0025",
          "expanded-probability-0141",
          "expanded-probability-0267",
          "expanded-probability-0291",
          "expanded-probability-0224",
          "expanded-probability-0225",
          "expanded-probability-0234",
          "expanded-probability-0090",
          "expanded-probability-0280",
          "expanded-probability-0148",
          "expanded-probability-0123",
          "expanded-probability-0226",
          "expanded-probability-0222",
          "expanded-probability-0220",
          "expanded-probability-0281",
          "expanded-probability-0072",
          "expanded-probability-0268",
          "expanded-probability-0138",
          "expanded-probability-0247",
          "expanded-probability-0005",
          "expanded-probability-0067",
          "expanded-probability-0128",
          "expanded-probability-0276",
          "expanded-probability-0232",
          "expanded-probability-0154",
          "expanded-probability-0181",
          "expanded-probability-0218",
          "expanded-probability-0151",
          "expanded-probability-0048",
          "expanded-probability-0093",
          "expanded-probability-0229",
          "expanded-probability-0283",
          "expanded-probability-0288",
          "expanded-probability-0136",
          "expanded-probability-0245",
          "expanded-probability-0159",
          "expanded-probability-0271",
          "expanded-probability-0277",
          "expanded-probability-0186",
          "expanded-probability-0049",
          "expanded-probability-0006",
          "expanded-probability-0080",
          "expanded-probability-0180",
          "expanded-probability-0237",
          "expanded-probability-0185",
          "expanded-probability-0069",
          "expanded-probability-0275",
          "expanded-probability-0008",
          "expanded-probability-0075",
          "expanded-probability-0262",
          "expanded-probability-0130",
          "expanded-probability-0152",
          "expanded-probability-0145",
          "expanded-probability-0206"
        ]
      },
      "Hy4 Preview": {
        "valid": 300,
        "attempts": 300,
        "mae": 0.04887760465541143,
        "case_ids": [
          "expanded-probability-0214",
          "expanded-probability-0084",
          "expanded-probability-0035",
          "expanded-probability-0213",
          "expanded-probability-0149",
          "expanded-probability-0098",
          "expanded-probability-0172",
          "expanded-probability-0192",
          "expanded-probability-0279",
          "expanded-probability-0256",
          "expanded-probability-0179",
          "expanded-probability-0259",
          "expanded-probability-0100",
          "expanded-probability-0252",
          "expanded-probability-0202",
          "expanded-probability-0194",
          "expanded-probability-0054",
          "expanded-probability-0157",
          "expanded-probability-0298",
          "expanded-probability-0254",
          "expanded-probability-0097",
          "expanded-probability-0297",
          "expanded-probability-0295",
          "expanded-probability-0227",
          "expanded-probability-0150",
          "expanded-probability-0070",
          "expanded-probability-0101",
          "expanded-probability-0133",
          "expanded-probability-0089",
          "expanded-probability-0118",
          "expanded-probability-0258",
          "expanded-probability-0086",
          "expanded-probability-0215",
          "expanded-probability-0171",
          "expanded-probability-0187",
          "expanded-probability-0050",
          "expanded-probability-0205",
          "expanded-probability-0190",
          "expanded-probability-0012",
          "expanded-probability-0263",
          "expanded-probability-0057",
          "expanded-probability-0028",
          "expanded-probability-0143",
          "expanded-probability-0105",
          "expanded-probability-0004",
          "expanded-probability-0253",
          "expanded-probability-0015",
          "expanded-probability-0244",
          "expanded-probability-0174",
          "expanded-probability-0156",
          "expanded-probability-0117",
          "expanded-probability-0065",
          "expanded-probability-0285",
          "expanded-probability-0274",
          "expanded-probability-0184",
          "expanded-probability-0068",
          "expanded-probability-0178",
          "expanded-probability-0040",
          "expanded-probability-0175",
          "expanded-probability-0173",
          "expanded-probability-0217",
          "expanded-probability-0270",
          "expanded-probability-0299",
          "expanded-probability-0062",
          "expanded-probability-0115",
          "expanded-probability-0287",
          "expanded-probability-0189",
          "expanded-probability-0107",
          "expanded-probability-0119",
          "expanded-probability-0051",
          "expanded-probability-0011",
          "expanded-probability-0140",
          "expanded-probability-0063",
          "expanded-probability-0195",
          "expanded-probability-0038",
          "expanded-probability-0161",
          "expanded-probability-0294",
          "expanded-probability-0110",
          "expanded-probability-0289",
          "expanded-probability-0094",
          "expanded-probability-0046",
          "expanded-probability-0131",
          "expanded-probability-0208",
          "expanded-probability-0047",
          "expanded-probability-0284",
          "expanded-probability-0036",
          "expanded-probability-0060",
          "expanded-probability-0209",
          "expanded-probability-0219",
          "expanded-probability-0029",
          "expanded-probability-0165",
          "expanded-probability-0223",
          "expanded-probability-0216",
          "expanded-probability-0198",
          "expanded-probability-0191",
          "expanded-probability-0241",
          "expanded-probability-0085",
          "expanded-probability-0034",
          "expanded-probability-0272",
          "expanded-probability-0010",
          "expanded-probability-0182",
          "expanded-probability-0032",
          "expanded-probability-0064",
          "expanded-probability-0290",
          "expanded-probability-0201",
          "expanded-probability-0088",
          "expanded-probability-0091",
          "expanded-probability-0078",
          "expanded-probability-0236",
          "expanded-probability-0099",
          "expanded-probability-0177",
          "expanded-probability-0168",
          "expanded-probability-0109",
          "expanded-probability-0264",
          "expanded-probability-0255",
          "expanded-probability-0122",
          "expanded-probability-0083",
          "expanded-probability-0239",
          "expanded-probability-0124",
          "expanded-probability-0296",
          "expanded-probability-0250",
          "expanded-probability-0037",
          "expanded-probability-0059",
          "expanded-probability-0016",
          "expanded-probability-0169",
          "expanded-probability-0246",
          "expanded-probability-0134",
          "expanded-probability-0074",
          "expanded-probability-0116",
          "expanded-probability-0238",
          "expanded-probability-0022",
          "expanded-probability-0030",
          "expanded-probability-0147",
          "expanded-probability-0233",
          "expanded-probability-0053",
          "expanded-probability-0231",
          "expanded-probability-0041",
          "expanded-probability-0077",
          "expanded-probability-0009",
          "expanded-probability-0300",
          "expanded-probability-0121",
          "expanded-probability-0230",
          "expanded-probability-0210",
          "expanded-probability-0273",
          "expanded-probability-0002",
          "expanded-probability-0129",
          "expanded-probability-0014",
          "expanded-probability-0007",
          "expanded-probability-0183",
          "expanded-probability-0261",
          "expanded-probability-0153",
          "expanded-probability-0095",
          "expanded-probability-0076",
          "expanded-probability-0257",
          "expanded-probability-0082",
          "expanded-probability-0162",
          "expanded-probability-0243",
          "expanded-probability-0020",
          "expanded-probability-0199",
          "expanded-probability-0142",
          "expanded-probability-0166",
          "expanded-probability-0188",
          "expanded-probability-0073",
          "expanded-probability-0102",
          "expanded-probability-0120",
          "expanded-probability-0158",
          "expanded-probability-0203",
          "expanded-probability-0251",
          "expanded-probability-0127",
          "expanded-probability-0043",
          "expanded-probability-0071",
          "expanded-probability-0197",
          "expanded-probability-0096",
          "expanded-probability-0126",
          "expanded-probability-0286",
          "expanded-probability-0135",
          "expanded-probability-0221",
          "expanded-probability-0021",
          "expanded-probability-0260",
          "expanded-probability-0003",
          "expanded-probability-0108",
          "expanded-probability-0282",
          "expanded-probability-0033",
          "expanded-probability-0087",
          "expanded-probability-0266",
          "expanded-probability-0155",
          "expanded-probability-0193",
          "expanded-probability-0160",
          "expanded-probability-0061",
          "expanded-probability-0278",
          "expanded-probability-0248",
          "expanded-probability-0112",
          "expanded-probability-0044",
          "expanded-probability-0228",
          "expanded-probability-0023",
          "expanded-probability-0163",
          "expanded-probability-0039",
          "expanded-probability-0066",
          "expanded-probability-0026",
          "expanded-probability-0204",
          "expanded-probability-0144",
          "expanded-probability-0240",
          "expanded-probability-0103",
          "expanded-probability-0265",
          "expanded-probability-0167",
          "expanded-probability-0292",
          "expanded-probability-0164",
          "expanded-probability-0019",
          "expanded-probability-0125",
          "expanded-probability-0052",
          "expanded-probability-0242",
          "expanded-probability-0146",
          "expanded-probability-0027",
          "expanded-probability-0056",
          "expanded-probability-0081",
          "expanded-probability-0079",
          "expanded-probability-0045",
          "expanded-probability-0207",
          "expanded-probability-0170",
          "expanded-probability-0269",
          "expanded-probability-0235",
          "expanded-probability-0137",
          "expanded-probability-0132",
          "expanded-probability-0024",
          "expanded-probability-0139",
          "expanded-probability-0001",
          "expanded-probability-0018",
          "expanded-probability-0106",
          "expanded-probability-0013",
          "expanded-probability-0111",
          "expanded-probability-0200",
          "expanded-probability-0092",
          "expanded-probability-0212",
          "expanded-probability-0055",
          "expanded-probability-0114",
          "expanded-probability-0249",
          "expanded-probability-0042",
          "expanded-probability-0031",
          "expanded-probability-0113",
          "expanded-probability-0176",
          "expanded-probability-0211",
          "expanded-probability-0017",
          "expanded-probability-0058",
          "expanded-probability-0196",
          "expanded-probability-0104",
          "expanded-probability-0293",
          "expanded-probability-0025",
          "expanded-probability-0141",
          "expanded-probability-0267",
          "expanded-probability-0291",
          "expanded-probability-0224",
          "expanded-probability-0225",
          "expanded-probability-0234",
          "expanded-probability-0090",
          "expanded-probability-0280",
          "expanded-probability-0148",
          "expanded-probability-0123",
          "expanded-probability-0226",
          "expanded-probability-0222",
          "expanded-probability-0281",
          "expanded-probability-0220",
          "expanded-probability-0072",
          "expanded-probability-0268",
          "expanded-probability-0138",
          "expanded-probability-0005",
          "expanded-probability-0247",
          "expanded-probability-0067",
          "expanded-probability-0128",
          "expanded-probability-0276",
          "expanded-probability-0232",
          "expanded-probability-0154",
          "expanded-probability-0218",
          "expanded-probability-0181",
          "expanded-probability-0151",
          "expanded-probability-0048",
          "expanded-probability-0093",
          "expanded-probability-0229",
          "expanded-probability-0283",
          "expanded-probability-0136",
          "expanded-probability-0288",
          "expanded-probability-0245",
          "expanded-probability-0271",
          "expanded-probability-0277",
          "expanded-probability-0159",
          "expanded-probability-0186",
          "expanded-probability-0049",
          "expanded-probability-0006",
          "expanded-probability-0080",
          "expanded-probability-0180",
          "expanded-probability-0237",
          "expanded-probability-0185",
          "expanded-probability-0275",
          "expanded-probability-0069",
          "expanded-probability-0008",
          "expanded-probability-0262",
          "expanded-probability-0075",
          "expanded-probability-0130",
          "expanded-probability-0145",
          "expanded-probability-0152",
          "expanded-probability-0206"
        ]
      }
    },
    "timing": {
      "boolq": {
        "Jev": {
          "median_of_case_medians_s": 0.2667752709967317,
          "common_complete_cases": 12
        },
        "Luna": {
          "median_of_case_medians_s": 1.3766149159928318,
          "common_complete_cases": 12
        },
        "Qwen3.8 Flash": {
          "median_of_case_medians_s": 1.1149753329955274,
          "common_complete_cases": 12
        },
        "DeepSeek V4.1 Flash": {
          "median_of_case_medians_s": 0.9122722080064705,
          "common_complete_cases": 12
        },
        "MiMo V2.6 Flash": {
          "median_of_case_medians_s": 1.6746404794976115,
          "common_complete_cases": 12
        },
        "Hy4 Preview": {
          "median_of_case_medians_s": 2.823063853997155,
          "common_complete_cases": 12
        }
      },
      "rte": {
        "Jev": {
          "median_of_case_medians_s": 0.2597315624880139,
          "common_complete_cases": 12
        },
        "Luna": {
          "median_of_case_medians_s": 1.3477908544882666,
          "common_complete_cases": 12
        },
        "Qwen3.8 Flash": {
          "median_of_case_medians_s": 1.168272834009258,
          "common_complete_cases": 12
        },
        "DeepSeek V4.1 Flash": {
          "median_of_case_medians_s": 0.9440041039924836,
          "common_complete_cases": 12
        },
        "MiMo V2.6 Flash": {
          "median_of_case_medians_s": 1.9624968334974255,
          "common_complete_cases": 12
        },
        "Hy4 Preview": {
          "median_of_case_medians_s": 2.9834769580193097,
          "common_complete_cases": 12
        }
      },
      "wic": {
        "Jev": {
          "median_of_case_medians_s": 0.25368537550093606,
          "common_complete_cases": 12
        },
        "Luna": {
          "median_of_case_medians_s": 1.204547208501026,
          "common_complete_cases": 12
        },
        "Qwen3.8 Flash": {
          "median_of_case_medians_s": 1.232746749999933,
          "common_complete_cases": 12
        },
        "DeepSeek V4.1 Flash": {
          "median_of_case_medians_s": 0.8423815209971508,
          "common_complete_cases": 12
        },
        "MiMo V2.6 Flash": {
          "median_of_case_medians_s": 1.451285146002192,
          "common_complete_cases": 12
        },
        "Hy4 Preview": {
          "median_of_case_medians_s": 2.783512812500703,
          "common_complete_cases": 12
        }
      },
      "policy": {
        "Jev": {
          "median_of_case_medians_s": 0.2589071249967674,
          "common_complete_cases": 12
        },
        "Luna": {
          "median_of_case_medians_s": 1.250741875017411,
          "common_complete_cases": 12
        },
        "Qwen3.8 Flash": {
          "median_of_case_medians_s": 1.3677198960067471,
          "common_complete_cases": 12
        },
        "DeepSeek V4.1 Flash": {
          "median_of_case_medians_s": 0.7922371245076647,
          "common_complete_cases": 12
        },
        "MiMo V2.6 Flash": {
          "median_of_case_medians_s": 1.477291499992134,
          "common_complete_cases": 12
        },
        "Hy4 Preview": {
          "median_of_case_medians_s": 3.0067971455137013,
          "common_complete_cases": 12
        }
      },
      "probability": {
        "Jev": {
          "median_of_case_medians_s": 0.25876891701773275,
          "common_complete_cases": 12
        },
        "Luna": {
          "median_of_case_medians_s": 1.2770443120098207,
          "common_complete_cases": 12
        },
        "Qwen3.8 Flash": {
          "median_of_case_medians_s": 1.2737293960090028,
          "common_complete_cases": 12
        },
        "DeepSeek V4.1 Flash": {
          "median_of_case_medians_s": 0.8004174170055194,
          "common_complete_cases": 12
        },
        "MiMo V2.6 Flash": {
          "median_of_case_medians_s": 1.969626062491443,
          "common_complete_cases": 12
        },
        "Hy4 Preview": {
          "median_of_case_medians_s": 3.103338145490852,
          "common_complete_cases": 12
        }
      }
    },
    "cache_session_0": {
      "DeepSeek V4.1 Flash": {
        "measured_usd_per_1000": 0.079519825,
        "prime_inclusive_usd_per_1000": 0.079519825,
        "median_latency_s": 1.0940515204856638,
        "attempts": 96,
        "valid": 96
      },
      "Hy4 Preview": {
        "measured_usd_per_1000": 0.828287371875,
        "prime_inclusive_usd_per_1000": 0.828287371875,
        "median_latency_s": 3.3181042079959298,
        "attempts": 96,
        "valid": 96
      },
      "Jev": {
        "measured_usd_per_1000": 0.134715,
        "prime_inclusive_usd_per_1000": 0.134715,
        "median_latency_s": 0.29669247899437323,
        "attempts": 96,
        "valid": 96
      },
      "Luna cached": {
        "measured_usd_per_1000": 0.07175833333333334,
        "prime_inclusive_usd_per_1000": 0.08703541666666667,
        "median_latency_s": 1.137883604504168,
        "attempts": 96,
        "valid": 96
      },
      "Luna uncached": {
        "measured_usd_per_1000": 0.30295833333333333,
        "prime_inclusive_usd_per_1000": 0.30295833333333333,
        "median_latency_s": 1.092991729005007,
        "attempts": 96,
        "valid": 96
      },
      "MiMo V2.6 Flash": {
        "measured_usd_per_1000": 0.08102616666666666,
        "prime_inclusive_usd_per_1000": 0.08102616666666666,
        "median_latency_s": 2.0327555624971865,
        "attempts": 96,
        "valid": 96
      },
      "Qwen3.8 Flash": {
        "measured_usd_per_1000": 0.12042362499999999,
        "prime_inclusive_usd_per_1000": 0.12042362499999999,
        "median_latency_s": 1.489837562010507,
        "attempts": 96,
        "valid": 96
      }
    }
  }
}
