{
  "generated": "2026-09-08",
  "benchmark": "fieldworkarena",
  "results": [
    {
      "system": "GPT-4o (2024-08-06)",
      "developer": "OpenAI",
      "value": 35.0,
      "date": "2026-06-07",
      "source": {
        "url": "https://arxiv.org/abs/2505.19662",
        "kind": "paper",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "notes": "Table 2 total score 0.35 in arXiv v4 (7 Jun 2026); models run as agents with default parameters."
      }
    },
    {
      "system": "Gemini 2.5 Pro",
      "developer": "Google DeepMind",
      "value": 46.0,
      "date": "2026-06-07",
      "source": {
        "url": "https://arxiv.org/abs/2505.19662",
        "kind": "paper",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "notes": "Table 2 total score 0.46 in arXiv v4."
      }
    },
    {
      "system": "GPT-5.2 (2025-12-11)",
      "developer": "OpenAI",
      "value": 52.0,
      "date": "2026-06-07",
      "source": {
        "url": "https://arxiv.org/abs/2505.19662",
        "kind": "paper",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "notes": "Table 2 total score 0.52 in arXiv v4; best model in the paper."
      }
    }
  ]
}