{
  "best_of_4": {
    "original": {
      "accuracy": [
        0.7430555555555557,
        0.6819069913587377,
        0.8042041197523737
      ],
      "grounding_accuracy": [
        0.5190972222222222,
        0.21280729166666662,
        0.8253871527777779
      ],
      "mean_iou": [
        0.22690152350473317,
        0.0236844939549421,
        0.43011855305452423
      ]
    }
  },
  "rl_answer_only": {
    "original": {
      "accuracy": [
        0.5364583333333334,
        0.38949657307254426,
        0.6834200935941225
      ],
      "grounding_accuracy": [
        0.5208333333333334,
        0.05001487321457421,
        0.9916517934520925
      ],
      "mean_iou": [
        0.23086301129744013,
        -0.06686375431282249,
        0.5285897769077028
      ]
    },
    "no_image": {
      "accuracy": [
        0.4131944444444444,
        0.3086076388888888,
        0.51778125
      ],
      "grounding_accuracy": [
        0.07291666666666667,
        0.034098882161620415,
        0.11173445117171293
      ],
      "mean_iou": [
        0.03260412443041397,
        -0.005733858267078201,
        0.07094210712790613
      ]
    },
    "occluded": {
      "accuracy": [
        0.3298611111111111,
        0.16602055973214927,
        0.49370166249007297
      ],
      "grounding_accuracy": [
        0.04340277777777778,
        0.001808871435316703,
        0.08499668412023886
      ],
      "mean_iou": [
        0.02670755135928329,
        0.006590274572173455,
        0.046824828146393126
      ]
    },
    "swapped": {
      "accuracy": [
        0.328125,
        0.09200563490351388,
        0.5642443650964861
      ],
      "grounding_accuracy": [
        0.08333333333333333,
        0.060921875,
        0.10574479166666666
      ],
      "mean_iou": [
        0.053155430214447526,
        0.04589849012140375,
        0.0604123703074913
      ]
    },
    "conflict": {
      "accuracy": [
        0.2378472222222222,
        0.19240602922244232,
        0.2832884152220021
      ],
      "grounding_accuracy": [
        0.5104166666666666,
        0.02852688139854026,
        0.992306451934793
      ],
      "mean_iou": [
        0.2287077154345012,
        -0.05673769535283485,
        0.5141531262218373
      ]
    },
    "visual_dependency_score": [
      0.21006944444444445,
      0.1620409383975085,
      0.2580979504913804
    ]
  },
  "rl_grounding": {
    "original": {
      "accuracy": [
        0.515625,
        0.2578120410182256,
        0.7734379589817744
      ],
      "grounding_accuracy": [
        0.45659722222222227,
        -0.1768101363640902,
        1.0900045808085348
      ],
      "mean_iou": [
        0.1974988696774276,
        -0.1659054055042223,
        0.5609031448590776
      ]
    },
    "no_image": {
      "accuracy": [
        0.3836805555555556,
        0.2148079078050944,
        0.5525532033060168
      ],
      "grounding_accuracy": [
        0.109375,
        0.03733224092959603,
        0.18141775907040397
      ],
      "mean_iou": [
        0.03225827991452992,
        0.005568456602232045,
        0.058948103226827794
      ]
    },
    "occluded": {
      "accuracy": [
        0.296875,
        0.10149643589525503,
        0.49225356410474497
      ],
      "grounding_accuracy": [
        0.05555555555555556,
        0.013961649213094486,
        0.09714946189801663
      ],
      "mean_iou": [
        0.02752357587400886,
        -0.0005820447617792511,
        0.055629196509796974
      ]
    },
    "swapped": {
      "accuracy": [
        0.2934027777777778,
        0.11395554347519118,
        0.47285001208036437
      ],
      "grounding_accuracy": [
        0.08854166666666667,
        0.04371875,
        0.13336458333333334
      ],
      "mean_iou": [
        0.045983211692925546,
        0.020723401547054942,
        0.07124302183879615
      ]
    },
    "conflict": {
      "accuracy": [
        0.27777777777777785,
        0.24521468376032032,
        0.3103408717952354
      ],
      "grounding_accuracy": [
        0.45486111111111116,
        -0.18773098041034042,
        1.0974532026325627
      ],
      "mean_iou": [
        0.19725477430555552,
        -0.16192624331204628,
        0.5564357919231573
      ]
    },
    "visual_dependency_score": [
      0.1973379629629629,
      0.05233011649372793,
      0.3423458094321979
    ]
  },
  "rl_weak": {
    "original": {
      "accuracy": [
        0.5625,
        0.5400885416666666,
        0.5849114583333334
      ],
      "grounding_accuracy": [
        0.2222222222222222,
        -0.05114041221034077,
        0.4955848566547852
      ],
      "mean_iou": [
        0.08113149714241581,
        -0.06994513907640533,
        0.23220813336123697
      ]
    },
    "no_image": {
      "accuracy": [
        0.421875,
        0.354640625,
        0.489109375
      ],
      "grounding_accuracy": [
        0.0625,
        0.0625,
        0.0625
      ],
      "mean_iou": [
        0.04171843624497992,
        0.04171843624497992,
        0.04171843624497992
      ]
    },
    "occluded": {
      "accuracy": [
        0.359375,
        0.29467869249158957,
        0.42407130750841043
      ],
      "grounding_accuracy": [
        0.0625,
        0.03662147699663583,
        0.08837852300336417
      ],
      "mean_iou": [
        0.03504709945551437,
        0.028662311242719216,
        0.04143188766830952
      ]
    },
    "swapped": {
      "accuracy": [
        0.3420138888888889,
        0.23196675665102473,
        0.45206102112675306
      ],
      "grounding_accuracy": [
        0.07465277777777778,
        0.044770833333333315,
        0.10453472222222224
      ],
      "mean_iou": [
        0.047049556398671616,
        0.04652120471066962,
        0.047577908086673615
      ]
    },
    "conflict": {
      "accuracy": [
        0.23611111111111113,
        0.18381770833333333,
        0.28840451388888894
      ],
      "grounding_accuracy": [
        0.18923611111111113,
        -0.008414373116509277,
        0.38688659533873154
      ],
      "mean_iou": [
        0.0692737295335187,
        -0.030094169288709602,
        0.16864162835574698
      ]
    },
    "visual_dependency_score": [
      0.2291666666666666,
      0.17210988652952838,
      0.28622344680380485
    ]
  },
  "sft": {
    "original": {
      "accuracy": [
        0.5520833333333334,
        0.5003262873266051,
        0.6038403793400616
      ],
      "grounding_accuracy": [
        0.5225694444444444,
        -0.0598889843502568,
        1.1050278732391456
      ],
      "mean_iou": [
        0.23201804326969042,
        -0.11357878915834377,
        0.5776148756977246
      ]
    },
    "no_image": {
      "accuracy": [
        0.4375,
        0.424560738498318,
        0.450439261501682
      ],
      "grounding_accuracy": [
        0.08159722222222222,
        0.02930381944444444,
        0.133890625
      ],
      "mean_iou": [
        0.023127776099820526,
        0.0010634344679384816,
        0.04519211773170257
      ]
    },
    "occluded": {
      "accuracy": [
        0.3697916666666667,
        0.32496875000000003,
        0.41461458333333334
      ],
      "grounding_accuracy": [
        0.03298611111111111,
        0.006050890384858426,
        0.0599213318373638
      ],
      "mean_iou": [
        0.01776157112424055,
        0.002707743050087967,
        0.03281539919839313
      ]
    },
    "swapped": {
      "accuracy": [
        0.34201388888888884,
        0.2558600401998811,
        0.42816773757789656
      ],
      "grounding_accuracy": [
        0.09027777777777778,
        0.07533680555555555,
        0.10521875
      ],
      "mean_iou": [
        0.0497187266694985,
        0.042926324065146994,
        0.05651112927385001
      ]
    },
    "conflict": {
      "accuracy": [
        0.23611111111111108,
        0.11870320594341344,
        0.3535190162788087
      ],
      "grounding_accuracy": [
        0.49826388888888884,
        -0.11663472270367259,
        1.1131625004814503
      ],
      "mean_iou": [
        0.2192636316212652,
        -0.1455860695649616,
        0.584113332807492
      ]
    },
    "visual_dependency_score": [
      0.21354166666666666,
      0.19860069444444434,
      0.22848263888888898
    ]
  }
}
