processed/external_benchmarks.json

{
  "benchmarks": {
    "DocVQA": {
      "metric": "ANLS",
      "SOTA": {
        "Claude 3.5 Sonnet": 95.2,
        "InternVL2-76B": 94.1,
        "Molmo-7B-D": 93.2,
        "GPT-4o": 92.8,
        "Qwen2-VL-72B": 94.5,
        "Llama 3.2 Vision (90B)": 90.1,
        "Gemini 1.5 Pro": 90.9
      }
    },
    "TextVQA": {
      "metric": "Accuracy",
      "SOTA": {
        "Claude 3.5 Sonnet": 94.9,
        "GPT-4o": 93.7,
        "Molmo-7B-D": 92.2,
        "Qwen2-VL-72B": 91.2,
        "InternVL2-76B": 84.4,
        "Gemini 1.5 Pro": 82.3,
        "Llama 3.2 Vision (11B)": 73.5
      }
    },
    "InfographicVQA": {
      "metric": "ANLS",
      "SOTA": {
        "Claude 3.5 Sonnet": 90.7,
        "Molmo-7B-D": 85.6,
        "Qwen2-VL-72B": 85.5,
        "InternVL2-76B": 82.0,
        "GPT-4o": 80.3,
        "Gemini 1.5 Pro": 75.8,
        "Llama 3.2 Vision (90B)": 56.8
      }
    },
    "VMCBench": {
      "metric": "Accuracy",
      "SOTA": {
        "Qwen2-VL-72B": 85.0,
        "GPT-4o": 80.3,
        "Claude 3.5 Sonnet": 78.5,
        "Gemini 1.5 Pro": 75.0,
        "InternVL2-76B": 74.5,
        "Llama 3.2 Vision (11B)": 60.3,
        "Molmo-7B-D": 72.0
      }
    }
  },
  "sources": [
    "https://arxiv.org/abs/2409.11340 (Llama 3.2)",
    "https://arxiv.org/abs/2410.05248 (VMCBench)",
    "https://molmo.allenai.org/blog (Molmo Release)",
    "https://github.com/OpenGVLab/InternVL (InternVL2)",
    "https://artificialanalysis.ai/models"
  ]
}