{
  "source": "Supplied manuscript: From Visual Affinity to Learning Credit in Multimodal Reinforcement Learning, Table 1, page 6",
  "unit": "percentage",
  "runs": 4,
  "benchmarks": ["MMMU", "MathVision", "We-Math", "MathVista", "MMK12-val", "MathVerse"],
  "models": {
    "4b": {
      "name": "Qwen3-VL-4B",
      "trainingData": "ViRL39K",
      "checkpoint": "step 76",
      "rows": [
        {"method": "Base", "scores": [46.56, 21.71, 75.52, 73.83, 54.73, 63.12], "average": 55.91},
        {"method": "GSPO", "scores": [44.67, 24.58, 76.52, 74.83, 67.14, 64.29], "average": 58.67},
        {"method": "DelTA", "scores": [44.33, 25.00, 76.00, 74.00, 74.58, 64.67], "average": 59.76},
        {"method": "VPPO", "scores": [45.21, 23.12, 76.43, 74.19, 69.23, 64.23], "average": 58.74},
        {"method": "AT-RL", "scores": [43.52, 22.78, 76.33, 74.22, 70.22, 63.45], "average": 58.42},
        {"method": "GSPO + OVC", "scores": [46.83, 26.64, 77.05, 74.37, 74.58, 64.80], "average": 60.71}
      ]
    },
    "8b": {
      "name": "Qwen3-VL-8B",
      "trainingData": "ViRL39K",
      "checkpoint": "step 76",
      "rows": [
        {"method": "Base", "scores": [48.78, 24.67, 80.23, 74.40, 64.46, 62.84], "average": 59.23},
        {"method": "GSPO", "scores": [44.56, 27.30, 83.33, 76.10, 74.73, 63.53], "average": 61.59},
        {"method": "DelTA", "scores": [49.21, 24.01, 85.40, 77.03, 77.06, 64.82], "average": 62.92},
        {"method": "VPPO", "scores": [49.43, 27.69, 84.36, 76.21, 75.29, 64.09], "average": 62.85},
        {"method": "AT-RL", "scores": [null, null, null, null, null, null], "average": null},
        {"method": "GSPO + OVC", "scores": [49.56, 29.93, 83.45, 76.97, 78.94, 65.05], "average": 63.98}
      ]
    },
    "9b": {
      "name": "Qwen3.5-9B",
      "trainingData": "Math77K",
      "checkpoint": "step 300",
      "rows": [
        {"method": "Base", "scores": [45.67, 36.18, 87.24, 79.20, 79.09, 77.36], "average": 67.46},
        {"method": "GSPO", "scores": [46.78, 34.87, 87.18, 79.20, 87.74, 77.74], "average": 68.92},
        {"method": "GSPO + OVC", "scores": [50.33, 40.46, 88.51, 80.30, 88.28, 76.80], "average": 70.78}
      ]
    }
  }
}