{
  "version": "xarts.intervals/1",
  "family": "intervals",
  "operation": "paired-values",
  "provenance": {
    "legacyId": "dumbbell"
  },
  "data": {
    "rows": [
      {
        "Department": "FinanceQA · info incompleta",
        "Before": 8.4,
        "After": 99
      },
      {
        "Department": "SWE-Bench Pro · issues reales",
        "Before": 23,
        "After": 95
      },
      {
        "Department": "SWE-bench Verified · curado",
        "Before": 77.2,
        "After": 95
      },
      {
        "Department": "Legal RAG · sin alucinación",
        "Before": 83,
        "After": 99
      },
      {
        "Department": "Vectara · resumen de doc corto",
        "Before": 96.7,
        "After": 99
      },
      {
        "Department": "Waymo · reducción lesiones graves",
        "Before": 90,
        "After": 90
      },
      {
        "Department": "MMLU-Pro · conocimiento general",
        "Before": 89.8,
        "After": 85
      },
      {
        "Department": "GPQA Diamond · nivel doctorado",
        "Before": 94.1,
        "After": 85
      }
    ]
  },
  "encoding": {
    "category": "Department",
    "start": "Before",
    "end": "After"
  },
  "calculation": {
    "order": "input",
    "difference": "end-minus-start",
    "percentBase": "start",
    "zeroBase": "null",
    "tolerance": 1e-9,
    "domain": {
      "origin": "zero",
      "paddingRatio": [
        0,
        0.1
      ],
      "paddingAbsolute": 0,
      "nice": true
    }
  },
  "layout": {
    "exportMeasurement": "shaped",
    "view": "paired-dots",
    "margin": {
      "top": 30,
      "right": 50,
      "bottom": 50,
      "left": 260
    },
    "measureCategory": "axis",
    "categoryMaxWidth": 200,
    "categoryWidthRatio": 0.28,
    "bandPadding": 0.3
  },
  "marks": {
    "radius": 5,
    "lineWidth": 2,
    "lineColor": "border",
    "startColor": "cyan",
    "endColor": "blue",
    "colorBy": "endpoint",
    "positiveColor": "emerald",
    "negativeColor": "red",
    "neutralColor": "cyan",
    "opacity": 0.9,
    "rowGuides": false,
    "guideColor": "borderLight",
    "baseline": "left",
    "arrow": {
      "length": 11,
      "width": 9,
      "startFill": "surface",
      "lineOpacity": 0.75
    }
  },
  "labels": {
    "start": "Frontera",
    "end": "Umbral",
    "delta": false,
    "deltaMode": "absolute",
    "deltaOffset": 8,
    "deltaFontSize": 11,
    "categoryFontSize": 11,
    "cutFontSize": 12,
    "cutFontWeight": 600,
    "categoryOffset": 8,
    "cutOffset": 8,
    "orderVisible": false,
    "orderText": {
      "input": "Filas en el orden del fichero",
      "end-desc": "Filas ordenadas por el valor de llegada",
      "difference-desc": "Filas ordenadas por el cambio (Δ)",
      "category": "Filas ordenadas por categoría"
    }
  },
  "axes": {
    "ticks": 5,
    "grid": true
  },
  "legend": {
    "visible": true,
    "position": "right",
    "shape": "circle"
  },
  "annotations": [],
  "document": {
    "header": {
      "title": "La brecha 85/95 — los benchmarks favorecen demos",
      "subtitle": "Mejor puntuación de frontera frente al umbral de despliegue, por caso de uso"
    },
    "spacing": {
      "horizontalPadding": 24,
      "topPadding": 22,
      "bottomPadding": 16,
      "minimumTextGap": 4
    }
  },
  "presentation": {
    "theme": "anlak",
    "dimensions": {
      "width": 720,
      "height": 440
    },
    "locale": "es-ES",
    "fonts": "reference",
    "surface": "theme",
    "framePadding": 24
  },
  "output": {
    "filename": "dumbbell-editorial-quality",
    "embedSpec": true,
    "geometryDecimals": 2
  }
}