{
  "task_id": "rcua_video_blender_001",
  "task_version": "0.1.9",
  "task_language": "zh-CN",
  "score_range": [
    0,
    1
  ],
  "trials": [
    {
      "model": "gpt",
      "model_name": "GPT-6.1 Sol",
      "budget_s": 210,
      "score": 0.0,
      "strict_success": false,
      "steps": 18,
      "saved_s": null,
      "sealed_s": 210.0,
      "api_median_s": 9.839520450914279,
      "video": "assets/gpt61sol_partial019_budget210_score0.000_replay_1x.mp4",
      "duration_s": 210,
      "video_sha256": "5b610db078a27552ad3223aefac6736c190ba003aeccdba009c7fd665eacdbe2",
      "trajectory": "assets/gptp19a_210_t01_trajectory.json",
      "raw_result": "assets/gptp19a_210_t01_result.json",
      "captured_frames": 90,
      "delivered_frames": 90,
      "criteria": [
        {
          "id": "body",
          "label": "机身形状与位置",
          "weight": 0.2,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "body"
        },
        {
          "id": "nose",
          "label": "尖头形状与位置",
          "weight": 0.15,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "nose"
        },
        {
          "id": "nozzle",
          "label": "喷口形状与位置",
          "weight": 0.1,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "nozzle"
        },
        {
          "id": "fins",
          "label": "四片尾翼整体形状与布局",
          "weight": 0.25,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "fins"
        },
        {
          "id": "window_shape",
          "label": "舷窗形状与位置",
          "weight": 0.1,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "window"
        },
        {
          "id": "window_recess",
          "label": "舷窗中心凹陷",
          "weight": 0.05,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "window_center_recess"
        },
        {
          "id": "silhouette",
          "label": "四视角整体轮廓",
          "weight": 0.1,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "multiview_silhouette"
        },
        {
          "id": "bounds",
          "label": "整体空间范围",
          "weight": 0.05,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "bounded_geometry"
        }
      ]
    },
    {
      "model": "gpt",
      "model_name": "GPT-6.1 Sol",
      "budget_s": 600,
      "score": 0.7,
      "strict_success": false,
      "steps": 24,
      "saved_s": 239.43348979321308,
      "sealed_s": 253.27517429017462,
      "api_median_s": 9.399636426358484,
      "video": "assets/gpt61sol_partial019_budget600_score0.700_replay_1x.mp4",
      "duration_s": 255,
      "video_sha256": "e5c85e9d18af2e2211b9e31eb258ddc3269ee2263e39bd6213ef89e07f72b387",
      "trajectory": "assets/gptp19_600_t01_trajectory.json",
      "raw_result": "assets/gptp19_600_t01_result.json",
      "captured_frames": 90,
      "delivered_frames": 90,
      "criteria": [
        {
          "id": "body",
          "label": "机身形状与位置",
          "weight": 0.2,
          "passed": true,
          "credit": 0.2,
          "geometry_check": "body"
        },
        {
          "id": "nose",
          "label": "尖头形状与位置",
          "weight": 0.15,
          "passed": true,
          "credit": 0.15,
          "geometry_check": "nose"
        },
        {
          "id": "nozzle",
          "label": "喷口形状与位置",
          "weight": 0.1,
          "passed": true,
          "credit": 0.1,
          "geometry_check": "nozzle"
        },
        {
          "id": "fins",
          "label": "四片尾翼整体形状与布局",
          "weight": 0.25,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "fins"
        },
        {
          "id": "window_shape",
          "label": "舷窗形状与位置",
          "weight": 0.1,
          "passed": true,
          "credit": 0.1,
          "geometry_check": "window"
        },
        {
          "id": "window_recess",
          "label": "舷窗中心凹陷",
          "weight": 0.05,
          "passed": true,
          "credit": 0.05,
          "geometry_check": "window_center_recess"
        },
        {
          "id": "silhouette",
          "label": "四视角整体轮廓",
          "weight": 0.1,
          "passed": true,
          "credit": 0.1,
          "geometry_check": "multiview_silhouette"
        },
        {
          "id": "bounds",
          "label": "整体空间范围",
          "weight": 0.05,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "bounded_geometry"
        }
      ]
    },
    {
      "model": "qwen",
      "model_name": "Original Qwen3.5-4B",
      "budget_s": 210,
      "score": 0.0,
      "strict_success": false,
      "steps": 29,
      "saved_s": 204.05192088894546,
      "sealed_s": 210.0,
      "api_median_s": 5.6570602029096335,
      "video": "assets/qwen35_4b_original_budget210_score0.000_1x_replay.mp4",
      "duration_s": 210,
      "video_sha256": "be714e36278f5f7474b95017f714d40ebece21d66874241fdbcd23d5ee24f1b5",
      "trajectory": "assets/q2798s210_trajectory.json",
      "raw_result": "assets/q2798s210_result.json",
      "captured_frames": 90,
      "delivered_frames": 90,
      "criteria": [
        {
          "id": "body",
          "label": "机身形状与位置",
          "weight": 0.2,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "body"
        },
        {
          "id": "nose",
          "label": "尖头形状与位置",
          "weight": 0.15,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "nose"
        },
        {
          "id": "nozzle",
          "label": "喷口形状与位置",
          "weight": 0.1,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "nozzle"
        },
        {
          "id": "fins",
          "label": "四片尾翼整体形状与布局",
          "weight": 0.25,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "fins"
        },
        {
          "id": "window_shape",
          "label": "舷窗形状与位置",
          "weight": 0.1,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "window"
        },
        {
          "id": "window_recess",
          "label": "舷窗中心凹陷",
          "weight": 0.05,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "window_center_recess"
        },
        {
          "id": "silhouette",
          "label": "四视角整体轮廓",
          "weight": 0.1,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "multiview_silhouette"
        },
        {
          "id": "bounds",
          "label": "整体空间范围",
          "weight": 0.05,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "bounded_geometry"
        }
      ]
    },
    {
      "model": "qwen",
      "model_name": "Original Qwen3.5-4B",
      "budget_s": 600,
      "score": 0.0,
      "strict_success": false,
      "steps": 80,
      "saved_s": null,
      "sealed_s": 600.0,
      "api_median_s": 5.87324784649536,
      "video": "assets/qwen35_4b_original_budget600_score0.000_1x_replay.mp4",
      "duration_s": 600,
      "video_sha256": "ffccf21434afac88027bbad0d6184f7d035c4744d154995556c7babb3cc2bec3",
      "trajectory": "assets/q2798s600_trajectory.json",
      "raw_result": "assets/q2798s600_result.json",
      "captured_frames": 90,
      "delivered_frames": 90,
      "criteria": [
        {
          "id": "body",
          "label": "机身形状与位置",
          "weight": 0.2,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "body"
        },
        {
          "id": "nose",
          "label": "尖头形状与位置",
          "weight": 0.15,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "nose"
        },
        {
          "id": "nozzle",
          "label": "喷口形状与位置",
          "weight": 0.1,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "nozzle"
        },
        {
          "id": "fins",
          "label": "四片尾翼整体形状与布局",
          "weight": 0.25,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "fins"
        },
        {
          "id": "window_shape",
          "label": "舷窗形状与位置",
          "weight": 0.1,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "window"
        },
        {
          "id": "window_recess",
          "label": "舷窗中心凹陷",
          "weight": 0.05,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "window_center_recess"
        },
        {
          "id": "silhouette",
          "label": "四视角整体轮廓",
          "weight": 0.1,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "multiview_silhouette"
        },
        {
          "id": "bounds",
          "label": "整体空间范围",
          "weight": 0.05,
          "passed": false,
          "credit": 0.0,
          "geometry_check": "bounded_geometry"
        }
      ]
    }
  ],
  "backend_qwen_http_stream_medians": [
    {
      "budget_s": 210,
      "n": 30,
      "backend_http_stream_median_s": 1.7500330774928443,
      "definition": "On-pod localhost HTTP request through full/cancelled SSE stream. Includes image processing/prefill/decode; different boundary from client first-complete-action timing."
    },
    {
      "budget_s": 600,
      "n": 81,
      "backend_http_stream_median_s": 1.9323997870087624,
      "definition": "On-pod localhost HTTP request through full/cancelled SSE stream. Includes image processing/prefill/decode; different boundary from client first-complete-action timing."
    }
  ],
  "limitations": [
    "One valid trial per budget; not a success-rate estimate",
    "Same task/scorer/budgets as GPT; reasoning/action protocols and GUI history differ",
    "Qwen endpoint requires authorized Taiji exec relay; measured end-to-end includes transfer/relay; not pure vLLM latency",
    "Video sampled1Hz from visible GUI, not native continuous video inference",
    "No CUA-trained4B or4B+27B used; existing production harness untouched"
  ],
  "source_manifest_sha256": "9818fe53f9d777394d0d3f5f5222beec35276519d2f2ebd81fe5d4b9e081649f"
}