[
  {
    "sample_id": "b_d5be48474254",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0001.JPG",
    "source_split": "train",
    "source_frame_id": "0001",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0001.JPG",
    "context_ids": [
      "b_1aede9f60837",
      "b_a38e3ce498d5",
      "b_63ebbf87a7c1",
      "b_442bffc55440"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_d5be48474254",
      "target_sha256": "c8e0464012d765884090d74c36c513738c8d3ff0de56534b10d97666e0dfae81",
      "context_ids": [
        "b_1aede9f60837",
        "b_a38e3ce498d5",
        "b_63ebbf87a7c1",
        "b_442bffc55440"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_1aede9f60837",
              "observation": "red umbrella partially obscured"
            },
            {
              "context_id": "b_a38e3ce498d5",
              "observation": "red umbrella fully visible"
            },
            {
              "context_id": "b_63ebbf87a7c1",
              "observation": "red umbrella partially obscured"
            },
            {
              "context_id": "b_442bffc5540",
              "observation": "red umbrella fully visible"
            }
          ],
          "uncertainty": "red umbrella's visibility varies across views",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "black and white umbrella",
          "grounding_queries": [
            "black and white umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_1aede9f60837",
              "observation": "black and white umbrella partially obscured"
            },
            {
              "context_id": "b_a38e3ce498d5",
              "observation": "black and white umbrella fully visible"
            },
            {
              "context_id": "b_63ebbf87a7c1",
              "observation": "black and white umbrella partially obscured"
            },
            {
              "context_id": "b_442bffc5540",
              "observation": "black and white umbrella fully visible"
            }
          ],
          "uncertainty": "black and white umbrella's visibility varies across views",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "mannequin shirt",
          "grounding_queries": [
            "mannequin shirt"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_1aede9f60837",
              "observation": "mannequin shirt partially obscured"
            },
            {
              "context_id": "b_a38e3ce498d5",
              "observation": "mannequin shirt fully visible"
            },
            {
              "context_id": "b_63ebbf87a7c1",
              "observation": "mannequin shirt partially obscured"
            },
            {
              "context_id": "b_442bffc5540",
              "observation": "mannequin shirt fully visible"
            }
          ],
          "uncertainty": "mannequin shirt's visibility varies across views",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_1aede9f60837\", \"observation\": \"red umbrella partially obscured\"},\n      {\"context_id\": \"b_a38e3ce498d5\", \"observation\": \"red umbrella fully visible\"},\n      {\"context_id\": \"b_63ebbf87a7c1\", \"observation\": \"red umbrella partially obscured\"},\n      {\"context_id\": \"b_442bffc5540\", \"observation\": \"red umbrella fully visible\"}\n    ],\n    \"uncertainty\": \"red umbrella's visibility varies across views\",\n    \"fallback_box_norm_xyxy\": [292, 275, 400, 370]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"black and white umbrella\",\n    \"grounding_queries\": [\"black and white umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_1aede9f60837\", \"observation\": \"black and white umbrella partially obscured\"},\n      {\"context_id\": \"b_a38e3ce498d5\", \"observation\": \"black and white umbrella fully visible\"},\n      {\"context_id\": \"b_63ebbf87a7c1\", \"observation\": \"black and white umbrella partially obscured\"},\n      {\"context_id\": \"b_442bffc5540\", \"observation\": \"black and white umbrella fully visible\"}\n    ],\n    \"uncertainty\": \"black and white umbrella's visibility varies across views\",\n    \"fallback_box_norm_xyxy\": [1, 320, 170, 448]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"mannequin shirt\",\n    \"grounding_queries\": [\"mannequin shirt\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_1aede9f60837\", \"observation\": \"mannequin shirt partially obscured\"},\n      {\"context_id\": \"b_a38e3ce498d5\", \"observation\": \"mannequin shirt fully visible\"},\n      {\"context_id\": \"b_63ebbf87a7c1\", \"observation\": \"mannequin shirt partially obscured\"},\n      {\"context_id\": \"b_442bffc5540\", \"observation\": \"mannequin shirt fully visible\"}\n    ],\n    \"uncertainty\": \"mannequin shirt's visibility varies across views\",\n    \"fallback_box_norm_xyxy\": [170, 280, 290, 448]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_d5be48474254",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "queries": [
            "red umbrella"
          ],
          "source": "grounding",
          "boxes_xyxy": [
            [
              113.717529296875,
              702.1360473632812,
              436.87554931640625,
              1007.55078125
            ],
            [
              753.3574829101562,
              598.4653930664062,
              1055.838623046875,
              894.8519287109375
            ]
          ],
          "box_scores": [
            0.3146287798881531,
            0.2932315170764923
          ],
          "box_queries": [
            "red umbrella",
            "red umbrella"
          ],
          "sam_predicted_iou": [
            0.99609375,
            0.97265625
          ],
          "mask_path": "masks/b_d5be48474254_r1.png"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "black and white umbrella",
          "queries": [
            "black and white umbrella"
          ],
          "source": "grounding",
          "boxes_xyxy": [
            [
              113.71705627441406,
              701.9348754882812,
              437.05767822265625,
              1007.9161376953125
            ]
          ],
          "box_scores": [
            0.27239349484443665
          ],
          "box_queries": [
            "black and white umbrella"
          ],
          "sam_predicted_iou": [
            0.99609375
          ],
          "mask_path": "masks/b_d5be48474254_r2.png"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "mannequin shirt",
          "queries": [
            "mannequin shirt"
          ],
          "source": "grounding",
          "boxes_xyxy": [],
          "box_scores": [],
          "box_queries": [],
          "sam_predicted_iou": [],
          "mask_path": "masks/b_d5be48474254_r3.png"
        }
      ],
      "final_mask_path": "masks/b_d5be48474254.png",
      "routing_mask_path": "masks/b_d5be48474254_routing.png",
      "predicted_fraction": 0.045033856966976794
    },
    "raw_mask_metrics": {
      "iou": 0.5666007355440863,
      "dice": 0.7233505291918604,
      "precision": 0.7930572077430518,
      "recall": 0.6649076995611649,
      "safety_recall": 0.6649076995611649,
      "leakage": 0.3350923004388351,
      "overmask": 0.00984842402570801,
      "boundary_f5": 0.12364454052526745,
      "iou_ignore_boundary5": 0.6084722406694116,
      "predicted_fraction": 0.045033856966976794,
      "gt_fraction": 0.05371335733320866
    },
    "routing_mask_metrics": {
      "iou": 0.5847536110654697,
      "dice": 0.7379741645419885,
      "precision": 0.7683739462398305,
      "recall": 0.7098882965219598,
      "safety_recall": 0.7098882965219598,
      "leakage": 0.2901117034780401,
      "overmask": 0.012146869975687733,
      "boundary_f5": 0.337117142558952,
      "iou_ignore_boundary5": 0.6080899308224442,
      "predicted_fraction": 0.04962490454595041,
      "gt_fraction": 0.05371335733320866
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.7976734486635477,
      "area_precision": 0.46590830857548365,
      "box_fraction": 0.0919616976015709
    },
    "frame_detected": true,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.00784313678741455,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_1aede9f60837",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0002.JPG",
    "source_split": "train",
    "source_frame_id": "0002",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0002.JPG",
    "context_ids": [
      "b_d5be48474254",
      "b_a38e3ce498d5",
      "b_63ebbf87a7c1",
      "b_442bffc55440"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_1aede9f60837",
      "target_sha256": "8fa7c6b67704a3493ad59f7ce40137e6dd08cf91d93521af022d99e65c613084",
      "context_ids": [
        "b_d5be48474254",
        "b_a38e3ce498d5",
        "b_63ebbf87a7c1",
        "b_442bffc55440"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_d5be48474254",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_a38e3ce498d5",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_63ebbf87a7c1",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_442bffc55440",
              "observation": "red umbrella is visible"
            }
          ],
          "uncertainty": "red umbrella is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "black and white umbrella",
          "grounding_queries": [
            "black and white umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_d5be48474254",
              "observation": "black and white umbrella is visible"
            },
            {
              "context_id": "b_a38e3ce498d5",
              "observation": "black and white umbrella is visible"
            },
            {
              "context_id": "b_63ebbf87a7c1",
              "observation": "black and white umbrella is visible"
            },
            {
              "context_id": "b_442bffc55440",
              "observation": "black and white umbrella is visible"
            }
          ],
          "uncertainty": "black and white umbrella is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "mannequin shirt",
          "grounding_queries": [
            "mannequin shirt"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_a38e3ce498d5",
              "observation": "mannequin shirt is visible"
            },
            {
              "context_id": "b_63ebbf87a7c1",
              "observation": "mannequin shirt is visible"
            },
            {
              "context_id": "b_442bffc55440",
              "observation": "mannequin shirt is visible"
            }
          ],
          "uncertainty": "mannequin shirt is consistently visible across two context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_d5be48474254\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_a38e3ce498d5\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_63ebbf87a7c1\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_442bffc55440\", \"observation\": \"red umbrella is visible\"}\n    ],\n    \"uncertainty\": \"red umbrella is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [280, 258, 380, 348]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"black and white umbrella\",\n    \"grounding_queries\": [\"black and white umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_d5be48474254\", \"observation\": \"black and white umbrella is visible\"},\n      {\"context_id\": \"b_a38e3ce498d5\", \"observation\": \"black and white umbrella is visible\"},\n      {\"context_id\": \"b_63ebbf87a7c1\", \"observation\": \"black and white umbrella is visible\"},\n      {\"context_id\": \"b_442bffc55440\", \"observation\": \"black and white umbrella is visible\"}\n    ],\n    \"uncertainty\": \"black and white umbrella is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0, 138, 280, 476]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"mannequin shirt\",\n    \"grounding_queries\": [\"mannequin shirt\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_a38e3ce498d5\", \"observation\": \"mannequin shirt is visible\"},\n      {\"context_id\": \"b_63ebbf87a7c1\", \"observation\": \"mannequin shirt is visible\"},\n      {\"context_id\": \"b_442bffc55440\", \"observation\": \"mannequin shirt is visible\"}\n    ],\n    \"uncertainty\": \"mannequin shirt is consistently visible across two context images\",\n    \"fallback_box_norm_xyxy\": [168, 308, 310, 476]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_1aede9f60837",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_1aede9f60837.png",
      "routing_mask_path": "masks/b_1aede9f60837_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.24996883133074632
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.24996883133074632
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_a38e3ce498d5",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0003.JPG",
    "source_split": "train",
    "source_frame_id": "0003",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0003.JPG",
    "context_ids": [
      "b_63ebbf87a7c1",
      "b_1aede9f60837",
      "b_442bffc55440",
      "b_6c74696a698a"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_a38e3ce498d5",
      "target_sha256": "1f9ba7f01a8ec6049827973e61b97173ff02690eff8dcb5da68e289ff5445ea2",
      "context_ids": [
        "b_63ebbf87a7c1",
        "b_1aede9f60837",
        "b_442bffc55440",
        "b_6c74696a698a"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "mannequin with shirt",
          "grounding_queries": [
            "mannequin with shirt"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_63ebbf87a7c1",
              "observation": "mannequin with shirt is visible"
            },
            {
              "context_id": "b_1aede9f60837",
              "observation": "mannequin with shirt is visible"
            },
            {
              "context_id": "b_442bffc55440",
              "observation": "mannequin with shirt is visible"
            },
            {
              "context_id": "b_6c74696a698a",
              "observation": "mannequin with shirt is visible"
            }
          ],
          "uncertainty": "mannequin with shirt is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_63ebbf87a7c1",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_1aede9f60837",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_442bffc55440",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_6c74696a698a",
              "observation": "red umbrella is visible"
            }
          ],
          "uncertainty": "red umbrella is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "sign with Chinese characters",
          "grounding_queries": [
            "sign with Chinese characters"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_63ebbf87a7c1",
              "observation": "sign with Chinese characters is visible"
            },
            {
              "context_id": "b_1aede9f60837",
              "observation": "sign with Chinese characters is visible"
            },
            {
              "context_id": "b_442bffc55440",
              "observation": "sign with Chinese characters is visible"
            },
            {
              "context_id": "b_6c74696a698a",
              "observation": "sign with Chinese characters is visible"
            }
          ],
          "uncertainty": "sign with Chinese characters is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"mannequin with shirt\",\n    \"grounding_queries\": [\"mannequin with shirt\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_63ebbf87a7c1\", \"observation\": \"mannequin with shirt is visible\"},\n      {\"context_id\": \"b_1aede9f60837\", \"observation\": \"mannequin with shirt is visible\"},\n      {\"context_id\": \"b_442bffc55440\", \"observation\": \"mannequin with shirt is visible\"},\n      {\"context_id\": \"b_6c74696a698a\", \"observation\": \"mannequin with shirt is visible\"}\n    ],\n    \"uncertainty\": \"mannequin with shirt is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_63ebbf87a7c1\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_1aede9f60837\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_442bffc55440\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_6c74696a698a\", \"observation\": \"red umbrella is visible\"}\n    ],\n    \"uncertainty\": \"red umbrella is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"sign with Chinese characters\",\n    \"grounding_queries\": [\"sign with Chinese characters\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_63ebbf87a7c1\", \"observation\": \"sign with Chinese characters is visible\"},\n      {\"context_id\": \"b_1aede9f60837\", \"observation\": \"sign with Chinese characters is visible\"},\n      {\"context_id\": \"b_442bffc55440\", \"observation\": \"sign with Chinese characters is visible\"},\n      {\"context_id\": \"b_6c74696a698a\", \"observation\": \"sign with Chinese characters is visible\"}\n    ],\n    \"uncertainty\": \"sign with Chinese characters is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_a38e3ce498d5",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_a38e3ce498d5.png",
      "routing_mask_path": "masks/b_a38e3ce498d5_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.12777303754266212
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.12777303754266212
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.00784313678741455,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_63ebbf87a7c1",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0004.JPG",
    "source_split": "train",
    "source_frame_id": "0004",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0004.JPG",
    "context_ids": [
      "b_442bffc55440",
      "b_a38e3ce498d5",
      "b_6c74696a698a",
      "b_e21902361190"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_63ebbf87a7c1",
      "target_sha256": "2e7c8d2413cc7d1ba3ba9cc698dc06124e65069ca702187a418ca895b8afc20a",
      "context_ids": [
        "b_442bffc55440",
        "b_a38e3ce498d5",
        "b_6c74696a698a",
        "b_e21902361190"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_442bffc55440",
              "observation": "umbrella is visible"
            },
            {
              "context_id": "b_a38e3ce498d5",
              "observation": "umbrella is visible"
            },
            {
              "context_id": "b_6c74696a698a",
              "observation": "umbrella is visible"
            },
            {
              "context_id": "b_e21902361190",
              "observation": "umbrella is visible"
            }
          ],
          "uncertainty": "umbrella is consistently visible across all contexts",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "sign with Chinese characters",
          "grounding_queries": [
            "sign with Chinese characters"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_442bffc55440",
              "observation": "sign is visible"
            },
            {
              "context_id": "b_a38e3ce498d5",
              "observation": "sign is visible"
            },
            {
              "context_id": "b_6c74696a698a",
              "observation": "sign is visible"
            },
            {
              "context_id": "b_e21902361190",
              "observation": "sign is visible"
            }
          ],
          "uncertainty": "sign is consistently visible across all contexts",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_442bffc55440\", \"observation\": \"umbrella is visible\"},\n      {\"context_id\": \"b_a38e3ce498d5\", \"observation\": \"umbrella is visible\"},\n      {\"context_id\": \"b_6c74696a698a\", \"observation\": \"umbrella is visible\"},\n      {\"context_id\": \"b_e21902361190\", \"observation\": \"umbrella is visible\"}\n    ],\n    \"uncertainty\": \"umbrella is consistently visible across all contexts\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"sign with Chinese characters\",\n    \"grounding_queries\": [\"sign with Chinese characters\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_442bffc55440\", \"observation\": \"sign is visible\"},\n      {\"context_id\": \"b_a38e3ce498d5\", \"observation\": \"sign is visible\"},\n      {\"context_id\": \"b_6c74696a698a\", \"observation\": \"sign is visible\"},\n      {\"context_id\": \"b_e21902361190\", \"observation\": \"sign is visible\"}\n    ],\n    \"uncertainty\": \"sign is consistently visible across all contexts\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_63ebbf87a7c1",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_63ebbf87a7c1.png",
      "routing_mask_path": "masks/b_63ebbf87a7c1_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.23910216700172987
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.23910216700172987
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_6c74696a698a",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0005.JPG",
    "source_split": "train",
    "source_frame_id": "0005",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0005.JPG",
    "context_ids": [
      "b_c7168d6511cd",
      "b_63ebbf87a7c1",
      "b_e21902361190",
      "b_442bffc55440"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_6c74696a698a",
      "target_sha256": "692c1ff802d24d43d30ea964cbb5401b228209287c7e3cb45fe06d378039eadc",
      "context_ids": [
        "b_c7168d6511cd",
        "b_63ebbf87a7c1",
        "b_e21902361190",
        "b_442bffc55440"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_c7168d6511cd",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_63ebbf87a7c1",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_e21902361190",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_442bffc55440",
              "observation": "red umbrella is visible"
            }
          ],
          "uncertainty": "red umbrella is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "white sign with red text",
          "grounding_queries": [
            "white sign with red text"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_c7168d6511cd",
              "observation": "white sign with red text is visible"
            },
            {
              "context_id": "b_63ebbf87a7c1",
              "observation": "white sign with red text is visible"
            },
            {
              "context_id": "b_e21902361190",
              "observation": "white sign with red text is visible"
            },
            {
              "context_id": "b_442bffc55440",
              "observation": "white sign with red text is visible"
            }
          ],
          "uncertainty": "white sign with red text is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_c7168d6511cd\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_63ebbf87a7c1\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_e21902361190\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_442bffc55440\", \"observation\": \"red umbrella is visible\"}\n    ],\n    \"uncertainty\": \"red umbrella is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [278, 250, 382, 340]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"white sign with red text\",\n    \"grounding_queries\": [\"white sign with red text\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_c7168d6511cd\", \"observation\": \"white sign with red text is visible\"},\n      {\"context_id\": \"b_63ebbf87a7c1\", \"observation\": \"white sign with red text is visible\"},\n      {\"context_id\": \"b_e21902361190\", \"observation\": \"white sign with red text is visible\"},\n      {\"context_id\": \"b_442bffc55440\", \"observation\": \"white sign with red text is visible\"}\n    ],\n    \"uncertainty\": \"white sign with red text is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [170, 250, 234, 300]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_6c74696a698a",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_6c74696a698a.png",
      "routing_mask_path": "masks/b_6c74696a698a_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.020396971963782005
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.020396971963782005
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.00784313678741455,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_c7168d6511cd",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0006.JPG",
    "source_split": "train",
    "source_frame_id": "0006",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0006.JPG",
    "context_ids": [
      "b_6c74696a698a",
      "b_e21902361190",
      "b_f2d05c389498",
      "b_f9646cc9c954"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_c7168d6511cd",
      "target_sha256": "a6ed36e8527621f968cb684b2277d0e217f8744eec4d5bbc6be1c9a93d1e5b11",
      "context_ids": [
        "b_6c74696a698a",
        "b_e21902361190",
        "b_f2d05c389498",
        "b_f9646cc9c954"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_6c74696a698a",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_e21902361190",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_f2d05c389498",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_f9646cc9c954",
              "observation": "red umbrella is visible"
            }
          ],
          "uncertainty": "red umbrella is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "black and white umbrella",
          "grounding_queries": [
            "black and white umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_f9646cc9c954",
              "observation": "black and white umbrella is visible"
            }
          ],
          "uncertainty": "black and white umbrella is only visible in one context image",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_6c74696a698a\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_e21902361190\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_f2d05c389498\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_f9646cc9c954\", \"observation\": \"red umbrella is visible\"}\n    ],\n    \"uncertainty\": \"red umbrella is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"black and white umbrella\",\n    \"grounding_queries\": [\"black and white umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_f9646cc9c954\", \"observation\": \"black and white umbrella is visible\"}\n    ],\n    \"uncertainty\": \"black and white umbrella is only visible in one context image\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_c7168d6511cd",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "black and white umbrella",
          "queries": [
            "black and white umbrella"
          ],
          "source": "grounding",
          "boxes_xyxy": [],
          "box_scores": [],
          "box_queries": [],
          "sam_predicted_iou": [],
          "mask_path": "masks/b_c7168d6511cd_r2.png"
        }
      ],
      "final_mask_path": "masks/b_c7168d6511cd.png",
      "routing_mask_path": "masks/b_c7168d6511cd_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.45545412751102593
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.45545412751102593
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.027450978755950928,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_f9646cc9c954",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0007.JPG",
    "source_split": "train",
    "source_frame_id": "0007",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0007.JPG",
    "context_ids": [
      "b_f2d05c389498",
      "b_1ea940c8ac4e",
      "b_39a6a990391f",
      "b_c7168d6511cd"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_f9646cc9c954",
      "target_sha256": "0f951d7fb7e2f9b4f41f6adca3e1ca2cb66f91a97ab2476d57248824398134e6",
      "context_ids": [
        "b_f2d05c389498",
        "b_1ea940c8ac4e",
        "b_39a6a990391f",
        "b_c7168d6511cd"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "black and white umbrella",
          "grounding_queries": [
            "black and white umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_f2d05c389498",
              "observation": "umbrella is visible"
            },
            {
              "context_id": "b_1ea940c8ac4e",
              "observation": "umbrella is visible"
            },
            {
              "context_id": "b_39a6a990391f",
              "observation": "umbrella is visible"
            },
            {
              "context_id": "b_c7168d6511cd",
              "observation": "umbrella is visible"
            }
          ],
          "uncertainty": "umbrella is consistently visible across all contexts",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_f2d05c389498",
              "observation": "umbrella is visible"
            },
            {
              "context_id": "b_1ea940c8ac4e",
              "observation": "umbrella is visible"
            },
            {
              "context_id": "b_39a6a990391f",
              "observation": "umbrella is visible"
            },
            {
              "context_id": "b_c7168d6511cd",
              "observation": "umbrella is visible"
            }
          ],
          "uncertainty": "umbrella is consistently visible across all contexts",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "flower arrangement",
          "grounding_queries": [
            "flower arrangement"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_f2d05c389498",
              "observation": "flower arrangement is visible"
            },
            {
              "context_id": "b_1ea940c8ac4e",
              "observation": "flower arrangement is visible"
            },
            {
              "context_id": "b_39a6a990391f",
              "observation": "flower arrangement is visible"
            },
            {
              "context_id": "b_c7168d6511cd",
              "observation": "flower arrangement is visible"
            }
          ],
          "uncertainty": "flower arrangement is consistently visible across all contexts",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"black and white umbrella\",\n    \"grounding_queries\": [\"black and white umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_f2d05c389498\", \"observation\": \"umbrella is visible\"},\n      {\"context_id\": \"b_1ea940c8ac4e\", \"observation\": \"umbrella is visible\"},\n      {\"context_id\": \"b_39a6a990391f\", \"observation\": \"umbrella is visible\"},\n      {\"context_id\": \"b_c7168d6511cd\", \"observation\": \"umbrella is visible\"}\n    ],\n    \"uncertainty\": \"umbrella is consistently visible across all contexts\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_f2d05c389498\", \"observation\": \"umbrella is visible\"},\n      {\"context_id\": \"b_1ea940c8ac4e\", \"observation\": \"umbrella is visible\"},\n      {\"context_id\": \"b_39a6a990391f\", \"observation\": \"umbrella is visible\"},\n      {\"context_id\": \"b_c7168d6511cd\", \"observation\": \"umbrella is visible\"}\n    ],\n    \"uncertainty\": \"umbrella is consistently visible across all contexts\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"flower arrangement\",\n    \"grounding_queries\": [\"flower arrangement\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_f2d05c389498\", \"observation\": \"flower arrangement is visible\"},\n      {\"context_id\": \"b_1ea940c8ac4e\", \"observation\": \"flower arrangement is visible\"},\n      {\"context_id\": \"b_39a6a990391f\", \"observation\": \"flower arrangement is visible\"},\n      {\"context_id\": \"b_c7168d6511cd\", \"observation\": \"flower arrangement is visible\"}\n    ],\n    \"uncertainty\": \"flower arrangement is consistently visible across all contexts\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_f9646cc9c954",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_f9646cc9c954.png",
      "routing_mask_path": "masks/b_f9646cc9c954_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.3984047485467608
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.3984047485467608
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.019607841968536377,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_1ea940c8ac4e",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0008.JPG",
    "source_split": "train",
    "source_frame_id": "0008",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0008.JPG",
    "context_ids": [
      "b_39a6a990391f",
      "b_31faa8f0af47",
      "b_f9646cc9c954",
      "b_0ed8205374dd"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_1ea940c8ac4e",
      "target_sha256": "5939628718d60686b21ce724529ae903572b899fa39971c25aa12f559a30b34d",
      "context_ids": [
        "b_39a6a990391f",
        "b_31faa8f0af47",
        "b_f9646cc9c954",
        "b_0ed8205374dd"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_39a6a990391f",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_31faa8f0af47",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_f9646cc9c954",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_0ed8205374dd",
              "observation": "red umbrella is visible"
            }
          ],
          "uncertainty": "red umbrella is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "black umbrella",
          "grounding_queries": [
            "black umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_39a6a990391f",
              "observation": "black umbrella is visible"
            },
            {
              "context_id": "b_31faa8f0af47",
              "observation": "black umbrella is visible"
            },
            {
              "context_id": "b_f9646cc9c954",
              "observation": "black umbrella is visible"
            },
            {
              "context_id": "b_0ed8205374dd",
              "observation": "black umbrella is visible"
            }
          ],
          "uncertainty": "black umbrella is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "green wall",
          "grounding_queries": [
            "green wall"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_39a6a990391f",
              "observation": "green wall is visible"
            },
            {
              "context_id": "b_31faa8f0af47",
              "observation": "green wall is visible"
            },
            {
              "context_id": "b_f9646cc9c954",
              "observation": "green wall is visible"
            },
            {
              "context_id": "b_0ed8205374dd",
              "observation": "green wall is visible"
            }
          ],
          "uncertainty": "green wall is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_39a6a990391f\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_31faa8f0af47\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_f9646cc9c954\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_0ed8205374dd\", \"observation\": \"red umbrella is visible\"}\n    ],\n    \"uncertainty\": \"red umbrella is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"black umbrella\",\n    \"grounding_queries\": [\"black umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_39a6a990391f\", \"observation\": \"black umbrella is visible\"},\n      {\"context_id\": \"b_31faa8f0af47\", \"observation\": \"black umbrella is visible\"},\n      {\"context_id\": \"b_f9646cc9c954\", \"observation\": \"black umbrella is visible\"},\n      {\"context_id\": \"b_0ed8205374dd\", \"observation\": \"black umbrella is visible\"}\n    ],\n    \"uncertainty\": \"black umbrella is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"green wall\",\n    \"grounding_queries\": [\"green wall\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_39a6a990391f\", \"observation\": \"green wall is visible\"},\n      {\"context_id\": \"b_31faa8f0af47\", \"observation\": \"green wall is visible\"},\n      {\"context_id\": \"b_f9646cc9c954\", \"observation\": \"green wall is visible\"},\n      {\"context_id\": \"b_0ed8205374dd\", \"observation\": \"green wall is visible\"}\n    ],\n    \"uncertainty\": \"green wall is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_1ea940c8ac4e",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_1ea940c8ac4e.png",
      "routing_mask_path": "masks/b_1ea940c8ac4e_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.11082263858369566
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.11082263858369566
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_31faa8f0af47",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0009.JPG",
    "source_split": "train",
    "source_frame_id": "0009",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0009.JPG",
    "context_ids": [
      "b_0ed8205374dd",
      "b_1ea940c8ac4e",
      "b_39a6a990391f",
      "b_57cf54a3bbc0"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_31faa8f0af47",
      "target_sha256": "8522b8e9cf92d396f1857e8197dbba361ef53ce127af536586946128b5e576d0",
      "context_ids": [
        "b_0ed8205374dd",
        "b_1ea940c8ac4e",
        "b_39a6a990391f",
        "b_57cf54a3bbc0"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "black curtain",
          "grounding_queries": [
            "black curtain"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_0ed8205374dd",
              "observation": "black curtain is absent"
            },
            {
              "context_id": "b_1ea940c8ac4e",
              "observation": "black curtain is absent"
            },
            {
              "context_id": "b_39a6a990391f",
              "observation": "black curtain is absent"
            },
            {
              "context_id": "b_57cf54a3bbc0",
              "observation": "black curtain is absent"
            }
          ],
          "uncertainty": "black curtain is consistently absent across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"black curtain\",\n    \"grounding_queries\": [\"black curtain\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_0ed8205374dd\", \"observation\": \"black curtain is absent\"},\n      {\"context_id\": \"b_1ea940c8ac4e\", \"observation\": \"black curtain is absent\"},\n      {\"context_id\": \"b_39a6a990391f\", \"observation\": \"black curtain is absent\"},\n      {\"context_id\": \"b_57cf54a3bbc0\", \"observation\": \"black curtain is absent\"}\n    ],\n    \"uncertainty\": \"black curtain is consistently absent across all context images\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_31faa8f0af47",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "black curtain",
          "queries": [
            "black curtain"
          ],
          "source": "grounding",
          "boxes_xyxy": [],
          "box_scores": [],
          "box_queries": [],
          "sam_predicted_iou": [],
          "mask_path": "masks/b_31faa8f0af47_r1.png"
        }
      ],
      "final_mask_path": "masks/b_31faa8f0af47.png",
      "routing_mask_path": "masks/b_31faa8f0af47_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.38210548256892174
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.38210548256892174
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.019607841968536377,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_0ed8205374dd",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0010.JPG",
    "source_split": "train",
    "source_frame_id": "0010",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0010.JPG",
    "context_ids": [
      "b_31faa8f0af47",
      "b_57cf54a3bbc0",
      "b_c7dd5015d023",
      "b_1ea940c8ac4e"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_0ed8205374dd",
      "target_sha256": "9fc4d940f95eac42ee267d66fd919fd55d45c2ccc498dcf4dfece29c22bbdabd",
      "context_ids": [
        "b_31faa8f0af47",
        "b_57cf54a3bbc0",
        "b_c7dd5015d023",
        "b_1ea940c8ac4e"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "black curtain",
          "grounding_queries": [
            "black curtain"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_31faa8f0af47",
              "observation": "black curtain partially obscures the teddy bear"
            },
            {
              "context_id": "b_57cf54a3bbc0",
              "observation": "black curtain fully covers the teddy bear"
            },
            {
              "context_id": "b_c7dd5015d023",
              "observation": "black curtain is not visible"
            },
            {
              "context_id": "b_1ea940c8ac4e",
              "observation": "black curtain is not visible"
            }
          ],
          "uncertainty": "black curtain is only visible in one context image",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "brown cushion",
          "grounding_queries": [
            "brown cushion"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_31faa8f0af47",
              "observation": "brown cushion is visible"
            },
            {
              "context_id": "b_57cf54a3bbc0",
              "observation": "brown cushion is visible"
            },
            {
              "context_id": "b_c7dd5015d023",
              "observation": "brown cushion is visible"
            },
            {
              "context_id": "b_1ea940c8ac4e",
              "observation": "brown cushion is visible"
            }
          ],
          "uncertainty": "brown cushion is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_31faa8f0af47",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_57cf54a3bbc0",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_c7dd5015d023",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_1ea940c8ac4e",
              "observation": "red umbrella is visible"
            }
          ],
          "uncertainty": "red umbrella is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"black curtain\",\n    \"grounding_queries\": [\"black curtain\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_31faa8f0af47\", \"observation\": \"black curtain partially obscures the teddy bear\"},\n      {\"context_id\": \"b_57cf54a3bbc0\", \"observation\": \"black curtain fully covers the teddy bear\"},\n      {\"context_id\": \"b_c7dd5015d023\", \"observation\": \"black curtain is not visible\"},\n      {\"context_id\": \"b_1ea940c8ac4e\", \"observation\": \"black curtain is not visible\"}\n    ],\n    \"uncertainty\": \"black curtain is only visible in one context image\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"brown cushion\",\n    \"grounding_queries\": [\"brown cushion\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_31faa8f0af47\", \"observation\": \"brown cushion is visible\"},\n      {\"context_id\": \"b_57cf54a3bbc0\", \"observation\": \"brown cushion is visible\"},\n      {\"context_id\": \"b_c7dd5015d023\", \"observation\": \"brown cushion is visible\"},\n      {\"context_id\": \"b_1ea940c8ac4e\", \"observation\": \"brown cushion is visible\"}\n    ],\n    \"uncertainty\": \"brown cushion is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_31faa8f0af47\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_57cf54a3bbc0\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_c7dd5015d023\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_1ea940c8ac4e\", \"observation\": \"red umbrella is visible\"}\n    ],\n    \"uncertainty\": \"red umbrella is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_0ed8205374dd",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "black curtain",
          "queries": [
            "black curtain"
          ],
          "source": "grounding",
          "boxes_xyxy": [],
          "box_scores": [],
          "box_queries": [],
          "sam_predicted_iou": [],
          "mask_path": "masks/b_0ed8205374dd_r1.png"
        }
      ],
      "final_mask_path": "masks/b_0ed8205374dd.png",
      "routing_mask_path": "masks/b_0ed8205374dd_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.10644295354309848
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.10644295354309848
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_c7dd5015d023",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0011.JPG",
    "source_split": "train",
    "source_frame_id": "0011",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0011.JPG",
    "context_ids": [
      "b_3fad5029fb0d",
      "b_57cf54a3bbc0",
      "b_0a52973d1b8d",
      "b_0ed8205374dd"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_c7dd5015d023",
      "target_sha256": "1a7c6617a66bee68a33a1c94e36afbcc21f1f40af2a2d54f760cdb68d0fe29c3",
      "context_ids": [
        "b_3fad5029fb0d",
        "b_57cf54a3bbc0",
        "b_0a52973d1b8d",
        "b_0ed8205374dd"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_3fad5029fb0d",
              "observation": "red umbrella partially obscured"
            },
            {
              "context_id": "b_57cf54a3bbc0",
              "observation": "red umbrella fully visible"
            },
            {
              "context_id": "b_0a52973d1b8d",
              "observation": "red umbrella fully visible"
            },
            {
              "context_id": "b_0ed8205374dd",
              "observation": "red umbrella fully visible"
            }
          ],
          "uncertainty": "red umbrella consistently visible across all contexts",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "green curtain",
          "grounding_queries": [
            "green curtain"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_3fad5029fb0d",
              "observation": "green curtain partially obscures the teddy bear"
            },
            {
              "context_id": "b_57cf54a3bbc0",
              "observation": "green curtain fully visible"
            },
            {
              "context_id": "b_0a52973d1b8d",
              "observation": "green curtain fully visible"
            },
            {
              "context_id": "b_0ed8205374dd",
              "observation": "green curtain fully visible"
            }
          ],
          "uncertainty": "green curtain consistently visible across all contexts",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_3fad5029fb0d\", \"observation\": \"red umbrella partially obscured\"},\n      {\"context_id\": \"b_57cf54a3bbc0\", \"observation\": \"red umbrella fully visible\"},\n      {\"context_id\": \"b_0a52973d1b8d\", \"observation\": \"red umbrella fully visible\"},\n      {\"context_id\": \"b_0ed8205374dd\", \"observation\": \"red umbrella fully visible\"}\n    ],\n    \"uncertainty\": \"red umbrella consistently visible across all contexts\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"green curtain\",\n    \"grounding_queries\": [\"green curtain\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_3fad5029fb0d\", \"observation\": \"green curtain partially obscures the teddy bear\"},\n      {\"context_id\": \"b_57cf54a3bbc0\", \"observation\": \"green curtain fully visible\"},\n      {\"context_id\": \"b_0a52973d1b8d\", \"observation\": \"green curtain fully visible\"},\n      {\"context_id\": \"b_0ed8205374dd\", \"observation\": \"green curtain fully visible\"}\n    ],\n    \"uncertainty\": \"green curtain consistently visible across all contexts\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_c7dd5015d023",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_c7dd5015d023.png",
      "routing_mask_path": "masks/b_c7dd5015d023_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.15088606682562689
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.15088606682562689
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_3fad5029fb0d",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0012.JPG",
    "source_split": "train",
    "source_frame_id": "0012",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0012.JPG",
    "context_ids": [
      "b_6125c5000375",
      "b_48b647235f33",
      "b_c7dd5015d023",
      "b_0a52973d1b8d"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_3fad5029fb0d",
      "target_sha256": "e986422f654342f73f45ef0a77076801f4c6ba007336b65b72a5b0254189af77",
      "context_ids": [
        "b_6125c5000375",
        "b_48b647235f33",
        "b_c7dd5015d023",
        "b_0a52973d1b8d"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "curtain",
          "grounding_queries": [
            "curtain"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_6125c5000375",
              "observation": "curtain partially visible"
            },
            {
              "context_id": "b_48b647235f33",
              "observation": "curtain fully visible"
            },
            {
              "context_id": "b_c7dd5015d023",
              "observation": "curtain fully visible"
            },
            {
              "context_id": "b_0a52973d1b8d",
              "observation": "curtain fully visible"
            }
          ],
          "uncertainty": "curtain is consistently visible across all contexts",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"curtain\",\n    \"grounding_queries\": [\"curtain\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_6125c5000375\", \"observation\": \"curtain partially visible\"},\n      {\"context_id\": \"b_48b647235f33\", \"observation\": \"curtain fully visible\"},\n      {\"context_id\": \"b_c7dd5015d023\", \"observation\": \"curtain fully visible\"},\n      {\"context_id\": \"b_0a52973d1b8d\", \"observation\": \"curtain fully visible\"}\n    ],\n    \"uncertainty\": \"curtain is consistently visible across all contexts\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_3fad5029fb0d",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_3fad5029fb0d.png",
      "routing_mask_path": "masks/b_3fad5029fb0d_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.19238763694734054
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.19238763694734054
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.0156862735748291,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_6125c5000375",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0013.JPG",
    "source_split": "train",
    "source_frame_id": "0013",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0013.JPG",
    "context_ids": [
      "b_3fad5029fb0d",
      "b_48b647235f33",
      "b_2962b6747c5c",
      "b_0a52973d1b8d"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_6125c5000375",
      "target_sha256": "ceef144931ee2abde82fa3b7d28510e0c313a7be5c40c58dde0e35f1ce5d52ef",
      "context_ids": [
        "b_3fad5029fb0d",
        "b_48b647235f33",
        "b_2962b6747c5c",
        "b_0a52973d1b8d"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red cloth",
          "grounding_queries": [
            "red cloth"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_3fad5029fb0d",
              "observation": "red cloth partially visible"
            },
            {
              "context_id": "b_48b647235f33",
              "observation": "red cloth completely absent"
            },
            {
              "context_id": "b_2962b6747c5c",
              "observation": "red cloth partially visible"
            },
            {
              "context_id": "b_0a52973d1b8d",
              "observation": "red cloth partially visible"
            }
          ],
          "uncertainty": "red cloth may be partially obscured in some views",
          "fallback_box_norm_xyxy": [
            0.0,
            0.0,
            0.3,
            1.0
          ],
          "routing_eligible": true,
          "consistency_filter": "passed"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red cloth\",\n    \"grounding_queries\": [\"red cloth\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_3fad5029fb0d\", \"observation\": \"red cloth partially visible\"},\n      {\"context_id\": \"b_48b647235f33\", \"observation\": \"red cloth completely absent\"},\n      {\"context_id\": \"b_2962b6747c5c\", \"observation\": \"red cloth partially visible\"},\n      {\"context_id\": \"b_0a52973d1b8d\", \"observation\": \"red cloth partially visible\"}\n    ],\n    \"uncertainty\": \"red cloth may be partially obscured in some views\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 0.3, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_6125c5000375",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red cloth",
          "queries": [
            "red cloth"
          ],
          "source": "grounding",
          "boxes_xyxy": [
            [
              106.89315795898438,
              247.7123260498047,
              676.0635986328125,
              1165.3028564453125
            ]
          ],
          "box_scores": [
            0.3131753206253052
          ],
          "box_queries": [
            "red cloth"
          ],
          "sam_predicted_iou": [
            0.99609375
          ],
          "mask_path": "masks/b_6125c5000375_r1.png"
        }
      ],
      "final_mask_path": "masks/b_6125c5000375.png",
      "routing_mask_path": "masks/b_6125c5000375_routing.png",
      "predicted_fraction": 0.23016406408278398
    },
    "raw_mask_metrics": {
      "iou": 0.8724448636983735,
      "dice": 0.9318777611161885,
      "precision": 1.0,
      "recall": 0.8724448636983735,
      "safety_recall": 0.8724448636983735,
      "leakage": 0.12755513630162654,
      "overmask": 0.0,
      "boundary_f5": 0.025880600641205082,
      "iou_ignore_boundary5": 0.9033726206814819,
      "predicted_fraction": 0.23016406408278398,
      "gt_fraction": 0.26381502563623044
    },
    "routing_mask_metrics": {
      "iou": 0.8967687269478278,
      "dice": 0.9455751923861133,
      "precision": 0.9999938244666398,
      "recall": 0.896773693333801,
      "safety_recall": 0.896773693333801,
      "leakage": 0.103226306666199,
      "overmask": 1.9845981949417884e-06,
      "boundary_f5": 0.45161147598369206,
      "iou_ignore_boundary5": 0.9236515442871507,
      "predicted_fraction": 0.23658383592812504,
      "gt_fraction": 0.26381502563623044
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.9123726470039856,
      "area_precision": 0.9472222594883215,
      "box_fraction": 0.25410890722645596
    },
    "frame_detected": true,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_2962b6747c5c",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0014.JPG",
    "source_split": "train",
    "source_frame_id": "0014",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0014.JPG",
    "context_ids": [
      "b_2d3a471eb3b2",
      "b_6125c5000375",
      "b_48b647235f33",
      "b_42d9140f0598"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_2962b6747c5c",
      "target_sha256": "694505997ce3fd02c783fc8b4875ecd868966c155a15378d4caa53c76d6f7fef",
      "context_ids": [
        "b_2d3a471eb3b2",
        "b_6125c5000375",
        "b_48b647235f33",
        "b_42d9140f0598"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "black and white umbrella",
          "grounding_queries": [
            "black and white umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_2d3a471eb3b2",
              "observation": "absent"
            },
            {
              "context_id": "b_6125c5000375",
              "observation": "absent"
            },
            {
              "context_id": "b_48b647235f33",
              "observation": "absent"
            },
            {
              "context_id": "b_42d9140f0598",
              "observation": "absent"
            }
          ],
          "uncertainty": "no other views confirm absence",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_2d3a471eb3b2",
              "observation": "present"
            },
            {
              "context_id": "b_6125c5000375",
              "observation": "absent"
            },
            {
              "context_id": "b_48b647235f33",
              "observation": "present"
            },
            {
              "context_id": "b_42d9140f0598",
              "observation": "absent"
            }
          ],
          "uncertainty": "no other views confirm absence",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "blue umbrella",
          "grounding_queries": [
            "blue umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_2d3a471eb3b2",
              "observation": "absent"
            },
            {
              "context_id": "b_6125c5000375",
              "observation": "absent"
            },
            {
              "context_id": "b_48b647235f33",
              "observation": "absent"
            },
            {
              "context_id": "b_42d9140f0598",
              "observation": "absent"
            }
          ],
          "uncertainty": "no other views confirm absence",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"black and white umbrella\",\n    \"grounding_queries\": [\"black and white umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_2d3a471eb3b2\", \"observation\": \"absent\"},\n      {\"context_id\": \"b_6125c5000375\", \"observation\": \"absent\"},\n      {\"context_id\": \"b_48b647235f33\", \"observation\": \"absent\"},\n      {\"context_id\": \"b_42d9140f0598\", \"observation\": \"absent\"}\n    ],\n    \"uncertainty\": \"no other views confirm absence\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_2d3a471eb3b2\", \"observation\": \"present\"},\n      {\"context_id\": \"b_6125c5000375\", \"observation\": \"absent\"},\n      {\"context_id\": \"b_48b647235f33\", \"observation\": \"present\"},\n      {\"context_id\": \"b_42d9140f0598\", \"observation\": \"absent\"}\n    ],\n    \"uncertainty\": \"no other views confirm absence\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"blue umbrella\",\n    \"grounding_queries\": [\"blue umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_2d3a471eb3b2\", \"observation\": \"absent\"},\n      {\"context_id\": \"b_6125c5000375\", \"observation\": \"absent\"},\n      {\"context_id\": \"b_48b647235f33\", \"observation\": \"absent\"},\n      {\"context_id\": \"b_42d9140f0598\", \"observation\": \"absent\"}\n    ],\n    \"uncertainty\": \"no other views confirm absence\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_2962b6747c5c",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "black and white umbrella",
          "queries": [
            "black and white umbrella"
          ],
          "source": "grounding",
          "boxes_xyxy": [],
          "box_scores": [],
          "box_queries": [],
          "sam_predicted_iou": [],
          "mask_path": "masks/b_2962b6747c5c_r1.png"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "queries": [
            "red umbrella"
          ],
          "source": "grounding",
          "boxes_xyxy": [
            [
              646.9236450195312,
              822.9298706054688,
              966.198486328125,
              1159.1925048828125
            ]
          ],
          "box_scores": [
            0.28780749440193176
          ],
          "box_queries": [
            "red umbrella"
          ],
          "sam_predicted_iou": [
            0.984375
          ],
          "mask_path": "masks/b_2962b6747c5c_r2.png"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "blue umbrella",
          "queries": [
            "blue umbrella"
          ],
          "source": "grounding",
          "boxes_xyxy": [
            [
              646.8992309570312,
              822.8531494140625,
              966.4400024414062,
              1159.5849609375
            ]
          ],
          "box_scores": [
            0.2961234748363495
          ],
          "box_queries": [
            "blue umbrella"
          ],
          "sam_predicted_iou": [
            0.984375
          ],
          "mask_path": "masks/b_2962b6747c5c_r3.png"
        }
      ],
      "final_mask_path": "masks/b_2962b6747c5c.png",
      "routing_mask_path": "masks/b_2962b6747c5c_routing.png",
      "predicted_fraction": 0.037878212320974956
    },
    "raw_mask_metrics": {
      "iou": 0.3983456099984043,
      "dice": 0.5697384211030045,
      "precision": 0.9949856641423557,
      "recall": 0.39914689938673725,
      "safety_recall": 0.39914689938673725,
      "leakage": 0.6008531006132628,
      "overmask": 0.00020973797273729028,
      "boundary_f5": 0.06760471954187672,
      "iou_ignore_boundary5": 0.5162138060558565,
      "predicted_fraction": 0.037878212320974956,
      "gt_fraction": 0.09442207443078218
    },
    "routing_mask_metrics": {
      "iou": 0.42483968604114297,
      "dice": 0.5963333141309983,
      "precision": 0.9874915040005724,
      "recall": 0.42713829617136284,
      "safety_recall": 0.42713829617136284,
      "leakage": 0.5728617038286371,
      "overmask": 0.0005641413676959424,
      "boundary_f5": 0.12949185840615218,
      "iou_ignore_boundary5": 0.5377809870228029,
      "predicted_fraction": 0.040842157962815775,
      "gt_fraction": 0.09442207443078218
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.46893713153944944,
      "area_precision": 0.8457252355748212,
      "box_fraction": 0.052355085168388736
    },
    "frame_detected": true,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_2d3a471eb3b2",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0015.JPG",
    "source_split": "train",
    "source_frame_id": "0015",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0015.JPG",
    "context_ids": [
      "b_2962b6747c5c",
      "b_3d50f98d854f",
      "b_42d9140f0598",
      "b_66a730d6ef90"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_2d3a471eb3b2",
      "target_sha256": "f014481f7f0078f0da0b52176b2bee217e86d6f0eadbe198bdd21db93b673616",
      "context_ids": [
        "b_2962b6747c5c",
        "b_3d50f98d854f",
        "b_42d9140f0598",
        "b_66a730d6ef90"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_2962b6747c5c",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_3d50f98d854f",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_42d9140f0598",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_66a730d6ef90",
              "observation": "red umbrella is visible"
            }
          ],
          "uncertainty": "red umbrella is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "black and white umbrella",
          "grounding_queries": [
            "black and white umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_2962b6747c5c",
              "observation": "black and white umbrella is visible"
            },
            {
              "context_id": "b_3d50f98d854f",
              "observation": "black and white umbrella is visible"
            },
            {
              "context_id": "b_42d9140f0598",
              "observation": "black and white umbrella is visible"
            },
            {
              "context_id": "b_66a730d6ef90",
              "observation": "black and white umbrella is visible"
            }
          ],
          "uncertainty": "black and white umbrella is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_2962b6747c5c\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_3d50f98d854f\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_42d9140f0598\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_66a730d6ef90\", \"observation\": \"red umbrella is visible\"}\n    ],\n    \"uncertainty\": \"red umbrella is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"black and white umbrella\",\n    \"grounding_queries\": [\"black and white umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_2962b6747c5c\", \"observation\": \"black and white umbrella is visible\"},\n      {\"context_id\": \"b_3d50f98d854f\", \"observation\": \"black and white umbrella is visible\"},\n      {\"context_id\": \"b_42d9140f0598\", \"observation\": \"black and white umbrella is visible\"},\n      {\"context_id\": \"b_66a730d6ef90\", \"observation\": \"black and white umbrella is visible\"}\n    ],\n    \"uncertainty\": \"black and white umbrella is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_2d3a471eb3b2",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_2d3a471eb3b2.png",
      "routing_mask_path": "masks/b_2d3a471eb3b2_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.2572437935387349
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.2572437935387349
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_3d50f98d854f",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0016.JPG",
    "source_split": "train",
    "source_frame_id": "0016",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0016.JPG",
    "context_ids": [
      "b_42d9140f0598",
      "b_66a730d6ef90",
      "b_4dd28c1b3b60",
      "b_2d3a471eb3b2"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_3d50f98d854f",
      "target_sha256": "b54ae4644bc87c5f8c5e614d9d4fd35577effb77828e182c45f19a0f901fdbbb",
      "context_ids": [
        "b_42d9140f0598",
        "b_66a730d6ef90",
        "b_4dd28c1b3b60",
        "b_2d3a471eb3b2"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "mannequin",
          "grounding_queries": [
            "mannequin"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_42d9140f0598",
              "observation": "mannequin appears in the foreground"
            },
            {
              "context_id": "b_66a730d6ef90",
              "observation": "mannequin appears in the foreground"
            },
            {
              "context_id": "b_4dd28c1b3b60",
              "observation": "mannequin appears in the foreground"
            },
            {
              "context_id": "b_2d3a471eb3b2",
              "observation": "mannequin appears in the foreground"
            }
          ],
          "uncertainty": "mannequin is consistently visible across all context images",
          "fallback_box_norm_xyxy": [
            0.7,
            0.3,
            1.0,
            0.8
          ],
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "mirror",
          "grounding_queries": [
            "mirror"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_4dd28c1b3b60",
              "observation": "mirror appears in the foreground"
            }
          ],
          "uncertainty": "mirror is only visible in one context image",
          "fallback_box_norm_xyxy": [
            0.6,
            0.5,
            0.9,
            0.8
          ],
          "routing_eligible": true,
          "consistency_filter": "passed"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"mannequin\",\n    \"grounding_queries\": [\"mannequin\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_42d9140f0598\", \"observation\": \"mannequin appears in the foreground\"},\n      {\"context_id\": \"b_66a730d6ef90\", \"observation\": \"mannequin appears in the foreground\"},\n      {\"context_id\": \"b_4dd28c1b3b60\", \"observation\": \"mannequin appears in the foreground\"},\n      {\"context_id\": \"b_2d3a471eb3b2\", \"observation\": \"mannequin appears in the foreground\"}\n    ],\n    \"uncertainty\": \"mannequin is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0.7, 0.3, 1.0, 0.8]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"mirror\",\n    \"grounding_queries\": [\"mirror\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_4dd28c1b3b60\", \"observation\": \"mirror appears in the foreground\"}\n    ],\n    \"uncertainty\": \"mirror is only visible in one context image\",\n    \"fallback_box_norm_xyxy\": [0.6, 0.5, 0.9, 0.8]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_3d50f98d854f",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "mirror",
          "queries": [
            "mirror"
          ],
          "source": "grounding",
          "boxes_xyxy": [
            [
              707.57080078125,
              487.44549560546875,
              1005.8277587890625,
              673.9874877929688
            ],
            [
              939.605712890625,
              390.13330078125,
              1174.2569580078125,
              565.254150390625
            ]
          ],
          "box_scores": [
            0.3109780251979828,
            0.2871554493904114
          ],
          "box_queries": [
            "mirror",
            "mirror"
          ],
          "sam_predicted_iou": [
            0.9453125,
            0.94921875
          ],
          "mask_path": "masks/b_3d50f98d854f_r2.png"
        }
      ],
      "final_mask_path": "masks/b_3d50f98d854f.png",
      "routing_mask_path": "masks/b_3d50f98d854f_routing.png",
      "predicted_fraction": 0.02520717424844546
    },
    "raw_mask_metrics": {
      "iou": 0.06761076176326317,
      "dice": 0.12665807461812656,
      "precision": 0.7376494909097935,
      "recall": 0.0692766056277716,
      "safety_recall": 0.0692766056277716,
      "leakage": 0.9307233943722284,
      "overmask": 0.009039288519657573,
      "boundary_f5": 0.18058851221590533,
      "iou_ignore_boundary5": 0.05759626117444637,
      "predicted_fraction": 0.02520717424844546,
      "gt_fraction": 0.26840315115246155
    },
    "routing_mask_metrics": {
      "iou": 0.07158868030260578,
      "dice": 0.1336122368937116,
      "precision": 0.6724126587275479,
      "recall": 0.07417568460092465,
      "safety_recall": 0.07417568460092465,
      "leakage": 0.9258243153990754,
      "overmask": 0.013257711919692189,
      "boundary_f5": 0.15425897400045405,
      "iou_ignore_boundary5": 0.05783324938404651,
      "predicted_fraction": 0.02960828774915455,
      "gt_fraction": 0.26840315115246155
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.07870279646685682,
      "area_precision": 0.4738986976662879,
      "box_fraction": 0.04457509311639939
    },
    "frame_detected": true,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.0156862735748291,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_4dd28c1b3b60",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0017.JPG",
    "source_split": "train",
    "source_frame_id": "0017",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0017.JPG",
    "context_ids": [
      "b_d95203c68daa",
      "b_3d50f98d854f",
      "b_8ffaf4db41bf",
      "b_17642d7e1164"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_4dd28c1b3b60",
      "target_sha256": "b1bcc7035d539193f181a7bb44e8eae918a065ce8bb46ebee596b324287323c8",
      "context_ids": [
        "b_d95203c68daa",
        "b_3d50f98d854f",
        "b_8ffaf4db41bf",
        "b_17642d7e1164"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "mirror",
          "grounding_queries": [
            "mirror"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_d95203c68daa",
              "observation": "mirror is visible"
            },
            {
              "context_id": "b_3d50f98d854f",
              "observation": "mirror is visible"
            },
            {
              "context_id": "b_8ffaf4db41bf",
              "observation": "mirror is visible"
            },
            {
              "context_id": "b_17642d7e1164",
              "observation": "mirror is visible"
            }
          ],
          "uncertainty": "mirror is consistently visible across all contexts",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "mirror",
          "grounding_queries": [
            "mirror"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_d95203c68daa",
              "observation": "mirror is visible"
            },
            {
              "context_id": "b_3d50f98d854f",
              "observation": "mirror is visible"
            },
            {
              "context_id": "b_8ffaf4db41bf",
              "observation": "mirror is visible"
            },
            {
              "context_id": "b_17642d7e1164",
              "observation": "mirror is visible"
            }
          ],
          "uncertainty": "mirror is consistently visible across all contexts",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"mirror\",\n    \"grounding_queries\": [\"mirror\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_d95203c68daa\", \"observation\": \"mirror is visible\"},\n      {\"context_id\": \"b_3d50f98d854f\", \"observation\": \"mirror is visible\"},\n      {\"context_id\": \"b_8ffaf4db41bf\", \"observation\": \"mirror is visible\"},\n      {\"context_id\": \"b_17642d7e1164\", \"observation\": \"mirror is visible\"}\n    ],\n    \"uncertainty\": \"mirror is consistently visible across all contexts\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"mirror\",\n    \"grounding_queries\": [\"mirror\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_d95203c68daa\", \"observation\": \"mirror is visible\"},\n      {\"context_id\": \"b_3d50f98d854f\", \"observation\": \"mirror is visible\"},\n      {\"context_id\": \"b_8ffaf4db41bf\", \"observation\": \"mirror is visible\"},\n      {\"context_id\": \"b_17642d7e1164\", \"observation\": \"mirror is visible\"}\n    ],\n    \"uncertainty\": \"mirror is consistently visible across all contexts\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_4dd28c1b3b60",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_4dd28c1b3b60.png",
      "routing_mask_path": "masks/b_4dd28c1b3b60_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.11984012420714697
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.11984012420714697
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.00784313678741455,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_d95203c68daa",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0018.JPG",
    "source_split": "train",
    "source_frame_id": "0018",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0018.JPG",
    "context_ids": [
      "b_4dd28c1b3b60",
      "b_17642d7e1164",
      "b_8ffaf4db41bf",
      "b_42d3a74baab7"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_d95203c68daa",
      "target_sha256": "01bc9f3c3918db7763a12902c0f0e86339b42f1d5c6a592d7fe1465a67e76508",
      "context_ids": [
        "b_4dd28c1b3b60",
        "b_17642d7e1164",
        "b_8ffaf4db41bf",
        "b_42d3a74baab7"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "mirror",
          "grounding_queries": [
            "mirror"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_4dd28c1b3b60",
              "observation": "mirror partially visible"
            },
            {
              "context_id": "b_17642d7e1164",
              "observation": "mirror fully visible"
            },
            {
              "context_id": "b_8ffaf4db41bf",
              "observation": "mirror partially visible"
            },
            {
              "context_id": "b_42d3a74baab7",
              "observation": "mirror fully visible"
            }
          ],
          "uncertainty": "mirror's position varies slightly across views",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "person holding mirror",
          "grounding_queries": [
            "person holding mirror"
          ],
          "is_spatial": true,
          "confidence": 0.8,
          "context_evidence": [
            {
              "context_id": "b_4dd28c1b3b60",
              "observation": "person not visible"
            },
            {
              "context_id": "b_17642d7e1164",
              "observation": "person not visible"
            },
            {
              "context_id": "b_8ffaf4db41bf",
              "observation": "person not visible"
            },
            {
              "context_id": "b_42d3a74baab7",
              "observation": "person partially visible"
            }
          ],
          "uncertainty": "person's visibility is inconsistent",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"mirror\",\n    \"grounding_queries\": [\"mirror\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_4dd28c1b3b60\", \"observation\": \"mirror partially visible\"},\n      {\"context_id\": \"b_17642d7e1164\", \"observation\": \"mirror fully visible\"},\n      {\"context_id\": \"b_8ffaf4db41bf\", \"observation\": \"mirror partially visible\"},\n      {\"context_id\": \"b_42d3a74baab7\", \"observation\": \"mirror fully visible\"}\n    ],\n    \"uncertainty\": \"mirror's position varies slightly across views\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"person holding mirror\",\n    \"grounding_queries\": [\"person holding mirror\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.8,\n    \"context_evidence\": [\n      {\"context_id\": \"b_4dd28c1b3b60\", \"observation\": \"person not visible\"},\n      {\"context_id\": \"b_17642d7e1164\", \"observation\": \"person not visible\"},\n      {\"context_id\": \"b_8ffaf4db41bf\", \"observation\": \"person not visible\"},\n      {\"context_id\": \"b_42d3a74baab7\", \"observation\": \"person partially visible\"}\n    ],\n    \"uncertainty\": \"person's visibility is inconsistent\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_d95203c68daa",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "mirror",
          "queries": [
            "mirror"
          ],
          "source": "grounding",
          "boxes_xyxy": [
            [
              1013.426025390625,
              278.7547912597656,
              1307.5675048828125,
              1166.7349853515625
            ],
            [
              639.6337280273438,
              490.9345703125,
              934.0805053710938,
              699.3328247070312
            ]
          ],
          "box_scores": [
            0.3015284538269043,
            0.26977986097335815
          ],
          "box_queries": [
            "mirror",
            "mirror"
          ],
          "sam_predicted_iou": [
            0.9296875,
            0.9296875
          ],
          "mask_path": "masks/b_d95203c68daa_r1.png"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "person holding mirror",
          "queries": [
            "person holding mirror"
          ],
          "source": "grounding",
          "boxes_xyxy": [
            [
              1014.04736328125,
              278.1952209472656,
              1307.2713623046875,
              1166.5721435546875
            ]
          ],
          "box_scores": [
            0.314229816198349
          ],
          "box_queries": [
            "person holding mirror"
          ],
          "sam_predicted_iou": [
            0.9296875
          ],
          "mask_path": "masks/b_d95203c68daa_r2.png"
        }
      ],
      "final_mask_path": "masks/b_d95203c68daa.png",
      "routing_mask_path": "masks/b_d95203c68daa_routing.png",
      "predicted_fraction": 0.09696378200632724
    },
    "raw_mask_metrics": {
      "iou": 0.6082124257635363,
      "dice": 0.756383194185026,
      "precision": 0.8440833751883475,
      "recall": 0.6851918294124842,
      "safety_recall": 0.6851918294124842,
      "leakage": 0.3148081705875158,
      "overmask": 0.017169098167442262,
      "boundary_f5": 0.18696034604215656,
      "iou_ignore_boundary5": 0.643403242943595,
      "predicted_fraction": 0.09696378200632724,
      "gt_fraction": 0.11944905481010488
    },
    "routing_mask_metrics": {
      "iou": 0.6857349874527269,
      "dice": 0.81357389216786,
      "precision": 0.8374446962339485,
      "recall": 0.7910262160068496,
      "safety_recall": 0.7910262160068496,
      "leakage": 0.20897378399315042,
      "overmask": 0.020828793511769982,
      "boundary_f5": 0.3946239744743433,
      "iou_ignore_boundary5": 0.699450611476952,
      "predicted_fraction": 0.11282814764598625,
      "gt_fraction": 0.11944905481010488
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.8608839238390346,
      "area_precision": 0.6528934308374947,
      "box_fraction": 0.15750161687471753
    },
    "frame_detected": true,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.00784313678741455,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_c0cd1b9b9c0e",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0019.JPG",
    "source_split": "train",
    "source_frame_id": "0019",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0019.JPG",
    "context_ids": [
      "b_bc84988e0e0a",
      "b_42d3a74baab7",
      "b_c0a10b3aef20",
      "b_17642d7e1164"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_c0cd1b9b9c0e",
      "target_sha256": "fcb5d7fcc27dc2c89475f6d2923b6f666a06af139f14a1713be5fe487be7896b",
      "context_ids": [
        "b_bc84988e0e0a",
        "b_42d3a74baab7",
        "b_c0a10b3aef20",
        "b_17642d7e1164"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_bc84988e0e0a",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_42d3a74baab7",
              "observation": "red umbrella is partially obscured"
            },
            {
              "context_id": "b_c0a10b3aef20",
              "observation": "red umbrella is fully obscured"
            },
            {
              "context_id": "b_17642d7e1164",
              "observation": "red umbrella is not visible"
            }
          ],
          "uncertainty": "red umbrella is only visible in one context",
          "fallback_box_norm_xyxy": [
            0.5,
            0.4,
            0.7,
            0.5
          ],
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "scroll painting",
          "grounding_queries": [
            "scroll painting"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_bc84988e0e0a",
              "observation": "scroll painting is visible"
            },
            {
              "context_id": "b_42d3a74baab7",
              "observation": "scroll painting is partially obscured"
            },
            {
              "context_id": "b_c0a10b3aef20",
              "observation": "scroll painting is fully obscured"
            },
            {
              "context_id": "b_17642d7e1164",
              "observation": "scroll painting is not visible"
            }
          ],
          "uncertainty": "scroll painting is only visible in one context",
          "fallback_box_norm_xyxy": [
            0.6,
            0.4,
            0.8,
            0.5
          ],
          "routing_eligible": true,
          "consistency_filter": "passed"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_bc84988e0e0a\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_42d3a74baab7\", \"observation\": \"red umbrella is partially obscured\"},\n      {\"context_id\": \"b_c0a10b3aef20\", \"observation\": \"red umbrella is fully obscured\"},\n      {\"context_id\": \"b_17642d7e1164\", \"observation\": \"red umbrella is not visible\"}\n    ],\n    \"uncertainty\": \"red umbrella is only visible in one context\",\n    \"fallback_box_norm_xyxy\": [0.5, 0.4, 0.7, 0.5]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"scroll painting\",\n    \"grounding_queries\": [\"scroll painting\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_bc84988e0e0a\", \"observation\": \"scroll painting is visible\"},\n      {\"context_id\": \"b_42d3a74baab7\", \"observation\": \"scroll painting is partially obscured\"},\n      {\"context_id\": \"b_c0a10b3aef20\", \"observation\": \"scroll painting is fully obscured\"},\n      {\"context_id\": \"b_17642d7e1164\", \"observation\": \"scroll painting is not visible\"}\n    ],\n    \"uncertainty\": \"scroll painting is only visible in one context\",\n    \"fallback_box_norm_xyxy\": [0.6, 0.4, 0.8, 0.5]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_c0cd1b9b9c0e",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "queries": [
            "red umbrella"
          ],
          "source": "grounding",
          "boxes_xyxy": [
            [
              974.6791381835938,
              756.0464477539062,
              1211.8345947265625,
              1101.6326904296875
            ]
          ],
          "box_scores": [
            0.27580294013023376
          ],
          "box_queries": [
            "red umbrella"
          ],
          "sam_predicted_iou": [
            0.98046875
          ],
          "mask_path": "masks/b_c0cd1b9b9c0e_r1.png"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "scroll painting",
          "queries": [
            "scroll painting"
          ],
          "source": "vlm_fallback",
          "boxes_xyxy": [
            [
              1054.8,
              467.20000000000005,
              1406.4,
              584.0
            ]
          ],
          "box_scores": [
            0.9
          ],
          "box_queries": [
            "vlm_fallback"
          ],
          "sam_predicted_iou": [
            0.8828125
          ],
          "mask_path": "masks/b_c0cd1b9b9c0e_r2.png"
        }
      ],
      "final_mask_path": "masks/b_c0cd1b9b9c0e.png",
      "routing_mask_path": "masks/b_c0cd1b9b9c0e_routing.png",
      "predicted_fraction": 0.04364003303878941
    },
    "raw_mask_metrics": {
      "iou": 0.1663498400744051,
      "dice": 0.28524861813980806,
      "precision": 0.8183644317471654,
      "recall": 0.1727271442502014,
      "safety_recall": 0.1727271442502014,
      "leakage": 0.8272728557497986,
      "overmask": 0.00999269396300321,
      "boundary_f5": 0.06648116416003601,
      "iou_ignore_boundary5": 0.21833857589573732,
      "predicted_fraction": 0.04364003303878941,
      "gt_fraction": 0.20676223759876572
    },
    "routing_mask_metrics": {
      "iou": 0.1860315780667422,
      "dice": 0.31370425797596013,
      "precision": 0.8228340550649559,
      "recall": 0.19379395789463766,
      "safety_recall": 0.19379395789463766,
      "leakage": 0.8062060421053623,
      "overmask": 0.010876171882194759,
      "boundary_f5": 0.06240564545849471,
      "iou_ignore_boundary5": 0.23662842039513968,
      "predicted_fraction": 0.048696662614739664,
      "gt_fraction": 0.20676223759876572
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.22794509061273713,
      "area_precision": 0.7863474961200627,
      "box_fraction": 0.05993588994342886
    },
    "frame_detected": true,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_42d3a74baab7",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0020.JPG",
    "source_split": "train",
    "source_frame_id": "0020",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0020.JPG",
    "context_ids": [
      "b_17642d7e1164",
      "b_c0cd1b9b9c0e",
      "b_d95203c68daa",
      "b_bc84988e0e0a"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_42d3a74baab7",
      "target_sha256": "beecb1ddf9871d8ad3aff7649bc170553336bbfad9551fe94b82dfc99715c5ab",
      "context_ids": [
        "b_17642d7e1164",
        "b_c0cd1b9b9c0e",
        "b_d95203c68daa",
        "b_bc84988e0e0a"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "person holding mirror",
          "grounding_queries": [
            "person holding mirror"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_17642d7e1164",
              "observation": "mirror is visible"
            },
            {
              "context_id": "b_c0cd1b9b9c0e",
              "observation": "mirror is visible"
            },
            {
              "context_id": "b_d95203c68daa",
              "observation": "mirror is visible"
            },
            {
              "context_id": "b_bc84988e0e0a",
              "observation": "mirror is visible"
            }
          ],
          "uncertainty": "mirror is consistently visible across all contexts",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_17642d7e1164",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_c0cd1b9b9c0e",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_d95203c68daa",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_bc84988e0e0a",
              "observation": "red umbrella is visible"
            }
          ],
          "uncertainty": "red umbrella is consistently visible across all contexts",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "color chart",
          "grounding_queries": [
            "color chart"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_17642d7e1164",
              "observation": "color chart is visible"
            },
            {
              "context_id": "b_c0cd1b9b9c0e",
              "observation": "color chart is visible"
            },
            {
              "context_id": "b_d95203c68daa",
              "observation": "color chart is visible"
            },
            {
              "context_id": "b_bc84988e0e0a",
              "observation": "color chart is visible"
            }
          ],
          "uncertainty": "color chart is consistently visible across all contexts",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"person holding mirror\",\n    \"grounding_queries\": [\"person holding mirror\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_17642d7e1164\", \"observation\": \"mirror is visible\"},\n      {\"context_id\": \"b_c0cd1b9b9c0e\", \"observation\": \"mirror is visible\"},\n      {\"context_id\": \"b_d95203c68daa\", \"observation\": \"mirror is visible\"},\n      {\"context_id\": \"b_bc84988e0e0a\", \"observation\": \"mirror is visible\"}\n    ],\n    \"uncertainty\": \"mirror is consistently visible across all contexts\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_17642d7e1164\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_c0cd1b9b9c0e\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_d95203c68daa\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_bc84988e0e0a\", \"observation\": \"red umbrella is visible\"}\n    ],\n    \"uncertainty\": \"red umbrella is consistently visible across all contexts\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"color chart\",\n    \"grounding_queries\": [\"color chart\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_17642d7e1164\", \"observation\": \"color chart is visible\"},\n      {\"context_id\": \"b_c0cd1b9b9c0e\", \"observation\": \"color chart is visible\"},\n      {\"context_id\": \"b_d95203c68daa\", \"observation\": \"color chart is visible\"},\n      {\"context_id\": \"b_bc84988e0e0a\", \"observation\": \"color chart is visible\"}\n    ],\n    \"uncertainty\": \"color chart is consistently visible across all contexts\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_42d3a74baab7",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_42d3a74baab7.png",
      "routing_mask_path": "masks/b_42d3a74baab7_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.22075599607274768
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.22075599607274768
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_bc84988e0e0a",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0021.JPG",
    "source_split": "train",
    "source_frame_id": "0021",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0021.JPG",
    "context_ids": [
      "b_c0cd1b9b9c0e",
      "b_45dcf0b2cf9e",
      "b_42d3a74baab7",
      "b_c0a10b3aef20"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_bc84988e0e0a",
      "target_sha256": "b2a22676fb275f1471a6a57d62769047f357b1dc24ac5d355f90ef4037ea737b",
      "context_ids": [
        "b_c0cd1b9b9c0e",
        "b_45dcf0b2cf9e",
        "b_42d3a74baab7",
        "b_c0a10b3aef20"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "scroll with cranes",
          "grounding_queries": [
            "scroll with cranes"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_c0cd1b9b9c0e",
              "observation": "scroll with cranes is visible"
            },
            {
              "context_id": "b_45dcf0b2cf9e",
              "observation": "scroll with cranes is visible"
            },
            {
              "context_id": "b_42d3a74baab7",
              "observation": "scroll with cranes is visible"
            },
            {
              "context_id": "b_c0a10b3aef20",
              "observation": "scroll with cranes is visible"
            }
          ],
          "uncertainty": "scroll with cranes is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_c0cd1b9b9c0e",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_45dcf0b2cf9e",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_42d3a74baab7",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_c0a10b3aef20",
              "observation": "red umbrella is visible"
            }
          ],
          "uncertainty": "red umbrella is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "color chart",
          "grounding_queries": [
            "color chart"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_c0cd1b9b9c0e",
              "observation": "color chart is visible"
            },
            {
              "context_id": "b_45dcf0b2cf9e",
              "observation": "color chart is visible"
            },
            {
              "context_id": "b_42d3a74baab7",
              "observation": "color chart is visible"
            },
            {
              "context_id": "b_c0a10b3aef20",
              "observation": "color chart is visible"
            }
          ],
          "uncertainty": "color chart is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"scroll with cranes\",\n    \"grounding_queries\": [\"scroll with cranes\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_c0cd1b9b9c0e\", \"observation\": \"scroll with cranes is visible\"},\n      {\"context_id\": \"b_45dcf0b2cf9e\", \"observation\": \"scroll with cranes is visible\"},\n      {\"context_id\": \"b_42d3a74baab7\", \"observation\": \"scroll with cranes is visible\"},\n      {\"context_id\": \"b_c0a10b3aef20\", \"observation\": \"scroll with cranes is visible\"}\n    ],\n    \"uncertainty\": \"scroll with cranes is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_c0cd1b9b9c0e\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_45dcf0b2cf9e\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_42d3a74baab7\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_c0a10b3aef20\", \"observation\": \"red umbrella is visible\"}\n    ],\n    \"uncertainty\": \"red umbrella is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"color chart\",\n    \"grounding_queries\": [\"color chart\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_c0cd1b9b9c0e\", \"observation\": \"color chart is visible\"},\n      {\"context_id\": \"b_45dcf0b2cf9e\", \"observation\": \"color chart is visible\"},\n      {\"context_id\": \"b_42d3a74baab7\", \"observation\": \"color chart is visible\"},\n      {\"context_id\": \"b_c0a10b3aef20\", \"observation\": \"color chart is visible\"}\n    ],\n    \"uncertainty\": \"color chart is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_bc84988e0e0a",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_bc84988e0e0a.png",
      "routing_mask_path": "masks/b_bc84988e0e0a_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.24241140305764644
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.24241140305764644
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_45dcf0b2cf9e",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0022.JPG",
    "source_split": "train",
    "source_frame_id": "0022",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0022.JPG",
    "context_ids": [
      "b_bc84988e0e0a",
      "b_c0cd1b9b9c0e",
      "b_2f68fc214982",
      "b_c0a10b3aef20"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_45dcf0b2cf9e",
      "target_sha256": "6094aaf918724214a99944de0f905327768498a637303147495f79f41fb064e1",
      "context_ids": [
        "b_bc84988e0e0a",
        "b_c0cd1b9b9c0e",
        "b_2f68fc214982",
        "b_c0a10b3aef20"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "black curtain",
          "grounding_queries": [
            "black curtain"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_c0a10b3aef20",
              "observation": "black curtain partially obscures the view"
            }
          ],
          "uncertainty": "curtain may be a reflection or specular effect",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_bc84988e0e0a",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_c0cd1b9b9c0e",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_2f68fc214982",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_c0a10b3aef20",
              "observation": "red umbrella is not visible"
            }
          ],
          "uncertainty": "umbrella may be a reflection or specular effect",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "white sign with red text",
          "grounding_queries": [
            "white sign with red text"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_bc84988e0e0a",
              "observation": "white sign with red text is visible"
            },
            {
              "context_id": "b_c0cd1b9b9c0e",
              "observation": "white sign with red text is visible"
            },
            {
              "context_id": "b_2f68fc214982",
              "observation": "white sign with red text is visible"
            },
            {
              "context_id": "b_c0a10b3aef20",
              "observation": "white sign with red text is not visible"
            }
          ],
          "uncertainty": "sign may be a reflection or specular effect",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"black curtain\",\n    \"grounding_queries\": [\"black curtain\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_c0a10b3aef20\", \"observation\": \"black curtain partially obscures the view\"}\n    ],\n    \"uncertainty\": \"curtain may be a reflection or specular effect\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_bc84988e0e0a\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_c0cd1b9b9c0e\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_2f68fc214982\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_c0a10b3aef20\", \"observation\": \"red umbrella is not visible\"}\n    ],\n    \"uncertainty\": \"umbrella may be a reflection or specular effect\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"white sign with red text\",\n    \"grounding_queries\": [\"white sign with red text\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_bc84988e0e0a\", \"observation\": \"white sign with red text is visible\"},\n      {\"context_id\": \"b_c0cd1b9b9c0e\", \"observation\": \"white sign with red text is visible\"},\n      {\"context_id\": \"b_2f68fc214982\", \"observation\": \"white sign with red text is visible\"},\n      {\"context_id\": \"b_c0a10b3aef20\", \"observation\": \"white sign with red text is not visible\"}\n    ],\n    \"uncertainty\": \"sign may be a reflection or specular effect\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_45dcf0b2cf9e",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "black curtain",
          "queries": [
            "black curtain"
          ],
          "source": "grounding",
          "boxes_xyxy": [],
          "box_scores": [],
          "box_queries": [],
          "sam_predicted_iou": [],
          "mask_path": "masks/b_45dcf0b2cf9e_r1.png"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "queries": [
            "red umbrella"
          ],
          "source": "grounding",
          "boxes_xyxy": [
            [
              281.3736572265625,
              890.8401489257812,
              416.388916015625,
              1016.050048828125
            ],
            [
              1122.8323974609375,
              582.1143188476562,
              1406.656494140625,
              736.9505615234375
            ]
          ],
          "box_scores": [
            0.2556205093860626,
            0.2505408525466919
          ],
          "box_queries": [
            "red umbrella",
            "red umbrella"
          ],
          "sam_predicted_iou": [
            0.97265625,
            0.97265625
          ],
          "mask_path": "masks/b_45dcf0b2cf9e_r2.png"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "white sign with red text",
          "queries": [
            "white sign with red text"
          ],
          "source": "grounding",
          "boxes_xyxy": [
            [
              1122.9132080078125,
              581.886474609375,
              1406.521484375,
              736.3436889648438
            ]
          ],
          "box_scores": [
            0.32850027084350586
          ],
          "box_queries": [
            "white sign with red text"
          ],
          "sam_predicted_iou": [
            0.97265625
          ],
          "mask_path": "masks/b_45dcf0b2cf9e_r3.png"
        }
      ],
      "final_mask_path": "masks/b_45dcf0b2cf9e.png",
      "routing_mask_path": "masks/b_45dcf0b2cf9e_routing.png",
      "predicted_fraction": 0.017133027880374647
    },
    "raw_mask_metrics": {
      "iou": 0.004627504683937668,
      "dice": 0.009212379040714222,
      "precision": 0.05244457077885162,
      "recall": 0.005049703312824331,
      "safety_recall": 0.005049703312824331,
      "leakage": 0.9949502966871757,
      "overmask": 0.01974850353322559,
      "boundary_f5": 0.0751433096427025,
      "iou_ignore_boundary5": 0.002886513071102152,
      "predicted_fraction": 0.017133027880374647,
      "gt_fraction": 0.17793803668552371
    },
    "routing_mask_metrics": {
      "iou": 0.0105129527512199,
      "dice": 0.020807160804020102,
      "precision": 0.10051204248056135,
      "recall": 0.011604738236517703,
      "safety_recall": 0.011604738236517703,
      "leakage": 0.9883952617634822,
      "overmask": 0.022478992592311738,
      "boundary_f5": 0.01760604361304794,
      "iou_ignore_boundary5": 0.00862226301161281,
      "predicted_fraction": 0.020544049121822745,
      "gt_fraction": 0.17793803668552371
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.032419368964988725,
      "area_precision": 0.19451514902701372,
      "box_fraction": 0.029656501784406316
    },
    "frame_detected": true,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_2f68fc214982",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0023.JPG",
    "source_split": "train",
    "source_frame_id": "0023",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0023.JPG",
    "context_ids": [
      "b_45dcf0b2cf9e",
      "b_c0a10b3aef20",
      "b_bc84988e0e0a",
      "b_c0cd1b9b9c0e"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_2f68fc214982",
      "target_sha256": "9aeb3f64365e0a88f38e6c4fe971124b99abdc8889daa561865c2178a8458f76",
      "context_ids": [
        "b_45dcf0b2cf9e",
        "b_c0a10b3aef20",
        "b_bc84988e0e0a",
        "b_c0cd1b9b9c0e"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "bear on floor",
          "grounding_queries": [
            "bear on floor"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_45dcf0b2cf9e",
              "observation": "bear partially obscured"
            },
            {
              "context_id": "b_c0a10b3aef20",
              "observation": "bear fully visible"
            },
            {
              "context_id": "b_bc84988e0e0a",
              "observation": "bear partially obscured"
            },
            {
              "context_id": "b_c0cd1b9b9c0e",
              "observation": "bear fully visible"
            }
          ],
          "uncertainty": "bear's position varies slightly across views",
          "fallback_box_norm_xyxy": [
            0.2,
            0.5,
            0.6,
            0.8
          ],
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "umbrella on floor",
          "grounding_queries": [
            "umbrella on floor"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_45dcf0b2cf9e",
              "observation": "umbrella partially obscured"
            },
            {
              "context_id": "b_c0a10b3aef20",
              "observation": "umbrella fully visible"
            },
            {
              "context_id": "b_bc84988e0e0a",
              "observation": "umbrella partially obscured"
            },
            {
              "context_id": "b_c0cd1b9b9c0e",
              "observation": "umbrella fully visible"
            }
          ],
          "uncertainty": "umbrella's position varies slightly across views",
          "fallback_box_norm_xyxy": [
            0.5,
            0.5,
            0.7,
            0.8
          ],
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "scroll on floor",
          "grounding_queries": [
            "scroll on floor"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_45dcf0b2cf9e",
              "observation": "scroll partially obscured"
            },
            {
              "context_id": "b_c0a10b3aef20",
              "observation": "scroll fully visible"
            },
            {
              "context_id": "b_bc84988e0e0a",
              "observation": "scroll partially obscured"
            },
            {
              "context_id": "b_c0cd1b9b9c0e",
              "observation": "scroll fully visible"
            }
          ],
          "uncertainty": "scroll's position varies slightly across views",
          "fallback_box_norm_xyxy": [
            0.7,
            0.5,
            0.9,
            0.8
          ],
          "routing_eligible": true,
          "consistency_filter": "passed"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"bear on floor\",\n    \"grounding_queries\": [\"bear on floor\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_45dcf0b2cf9e\", \"observation\": \"bear partially obscured\"},\n      {\"context_id\": \"b_c0a10b3aef20\", \"observation\": \"bear fully visible\"},\n      {\"context_id\": \"b_bc84988e0e0a\", \"observation\": \"bear partially obscured\"},\n      {\"context_id\": \"b_c0cd1b9b9c0e\", \"observation\": \"bear fully visible\"}\n    ],\n    \"uncertainty\": \"bear's position varies slightly across views\",\n    \"fallback_box_norm_xyxy\": [0.2, 0.5, 0.6, 0.8]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"umbrella on floor\",\n    \"grounding_queries\": [\"umbrella on floor\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_45dcf0b2cf9e\", \"observation\": \"umbrella partially obscured\"},\n      {\"context_id\": \"b_c0a10b3aef20\", \"observation\": \"umbrella fully visible\"},\n      {\"context_id\": \"b_bc84988e0e0a\", \"observation\": \"umbrella partially obscured\"},\n      {\"context_id\": \"b_c0cd1b9b9c0e\", \"observation\": \"umbrella fully visible\"}\n    ],\n    \"uncertainty\": \"umbrella's position varies slightly across views\",\n    \"fallback_box_norm_xyxy\": [0.5, 0.5, 0.7, 0.8]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"scroll on floor\",\n    \"grounding_queries\": [\"scroll on floor\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_45dcf0b2cf9e\", \"observation\": \"scroll partially obscured\"},\n      {\"context_id\": \"b_c0a10b3aef20\", \"observation\": \"scroll fully visible\"},\n      {\"context_id\": \"b_bc84988e0e0a\", \"observation\": \"scroll partially obscured\"},\n      {\"context_id\": \"b_c0cd1b9b9c0e\", \"observation\": \"scroll fully visible\"}\n    ],\n    \"uncertainty\": \"scroll's position varies slightly across views\",\n    \"fallback_box_norm_xyxy\": [0.7, 0.5, 0.9, 0.8]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_2f68fc214982",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "bear on floor",
          "queries": [
            "bear on floor"
          ],
          "source": "vlm_fallback",
          "boxes_xyxy": [
            [
              351.6,
              584.0,
              1054.8,
              934.4000000000001
            ]
          ],
          "box_scores": [
            0.9
          ],
          "box_queries": [
            "vlm_fallback"
          ],
          "sam_predicted_iou": [
            0.478515625
          ],
          "mask_path": "masks/b_2f68fc214982_r1.png"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "umbrella on floor",
          "queries": [
            "umbrella on floor"
          ],
          "source": "grounding",
          "boxes_xyxy": [
            [
              138.65228271484375,
              864.3546752929688,
              303.8629150390625,
              1014.9258422851562
            ]
          ],
          "box_scores": [
            0.2737511098384857
          ],
          "box_queries": [
            "umbrella on floor"
          ],
          "sam_predicted_iou": [
            0.97265625
          ],
          "mask_path": "masks/b_2f68fc214982_r2.png"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "scroll on floor",
          "queries": [
            "scroll on floor"
          ],
          "source": "vlm_fallback",
          "boxes_xyxy": [
            [
              1230.6,
              584.0,
              1582.2,
              934.4000000000001
            ]
          ],
          "box_scores": [
            0.9
          ],
          "box_queries": [
            "vlm_fallback"
          ],
          "sam_predicted_iou": [
            0.83203125
          ],
          "mask_path": "masks/b_2f68fc214982_r3.png"
        }
      ],
      "final_mask_path": "masks/b_2f68fc214982.png",
      "routing_mask_path": "masks/b_2f68fc214982_routing.png",
      "predicted_fraction": 0.015903813486683186
    },
    "raw_mask_metrics": {
      "iou": 0.03125341260692835,
      "dice": 0.06061247841676888,
      "precision": 0.3155009799118079,
      "recall": 0.03352673385246675,
      "safety_recall": 0.03352673385246675,
      "leakage": 0.9664732661475333,
      "overmask": 0.012802134204487076,
      "boundary_f5": 0.09002814402788738,
      "iou_ignore_boundary5": 0.033187962332306586,
      "predicted_fraction": 0.015903813486683186,
      "gt_fraction": 0.1496617225365063
    },
    "routing_mask_metrics": {
      "iou": 0.05218082441094991,
      "dice": 0.09918603951020051,
      "precision": 0.31205331771918643,
      "recall": 0.05896383746546613,
      "safety_recall": 0.05896383746546613,
      "leakage": 0.9410361625345338,
      "overmask": 0.02287866751964592,
      "boundary_f5": 0.1333250484638646,
      "iou_ignore_boundary5": 0.04658328395496693,
      "predicted_fraction": 0.02827923621175994,
      "gt_fraction": 0.1496617225365063
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.259701210841276,
      "area_precision": 0.20265353021088583,
      "box_fraction": 0.19179202315832125
    },
    "frame_detected": true,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_c0a10b3aef20",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0024.JPG",
    "source_split": "train",
    "source_frame_id": "0024",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0024.JPG",
    "context_ids": [
      "b_c0cd1b9b9c0e",
      "b_42d3a74baab7",
      "b_bc84988e0e0a",
      "b_2f68fc214982"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_c0a10b3aef20",
      "target_sha256": "8d91c5d290a5064fe0a6fccead5e94c7d4909c3820c3252916712b856df52c85",
      "context_ids": [
        "b_c0cd1b9b9c0e",
        "b_42d3a74baab7",
        "b_bc84988e0e0a",
        "b_2f68fc214982"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "black curtain",
          "grounding_queries": [
            "black curtain"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_c0cd1b9b9c0e",
              "observation": "black curtain partially visible"
            },
            {
              "context_id": "b_42d3a74baab7",
              "observation": "black curtain fully visible"
            },
            {
              "context_id": "b_bc84988e0e0a",
              "observation": "black curtain partially visible"
            },
            {
              "context_id": "b_2f68fc214982",
              "observation": "black curtain partially visible"
            }
          ],
          "uncertainty": "black curtain is consistently visible across all contexts.",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_c0cd1b9b9c0e",
              "observation": "red umbrella partially visible"
            },
            {
              "context_id": "b_42d3a74baab7",
              "observation": "red umbrella fully visible"
            },
            {
              "context_id": "b_bc84988e0e0a",
              "observation": "red umbrella partially visible"
            },
            {
              "context_id": "b_2f68fc214982",
              "observation": "red umbrella partially visible"
            }
          ],
          "uncertainty": "red umbrella is consistently visible across all contexts.",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "white signboard",
          "grounding_queries": [
            "white signboard"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_c0cd1b9b9c0e",
              "observation": "white signboard partially visible"
            },
            {
              "context_id": "b_42d3a74baab7",
              "observation": "white signboard fully visible"
            },
            {
              "context_id": "b_bc84988e0e0a",
              "observation": "white signboard partially visible"
            },
            {
              "context_id": "b_2f68fc214982",
              "observation": "white signboard partially visible"
            }
          ],
          "uncertainty": "white signboard is consistently visible across all contexts.",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"black curtain\",\n    \"grounding_queries\": [\"black curtain\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_c0cd1b9b9c0e\", \"observation\": \"black curtain partially visible\"},\n      {\"context_id\": \"b_42d3a74baab7\", \"observation\": \"black curtain fully visible\"},\n      {\"context_id\": \"b_bc84988e0e0a\", \"observation\": \"black curtain partially visible\"},\n      {\"context_id\": \"b_2f68fc214982\", \"observation\": \"black curtain partially visible\"}\n    ],\n    \"uncertainty\": \"black curtain is consistently visible across all contexts.\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_c0cd1b9b9c0e\", \"observation\": \"red umbrella partially visible\"},\n      {\"context_id\": \"b_42d3a74baab7\", \"observation\": \"red umbrella fully visible\"},\n      {\"context_id\": \"b_bc84988e0e0a\", \"observation\": \"red umbrella partially visible\"},\n      {\"context_id\": \"b_2f68fc214982\", \"observation\": \"red umbrella partially visible\"}\n    ],\n    \"uncertainty\": \"red umbrella is consistently visible across all contexts.\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"white signboard\",\n    \"grounding_queries\": [\"white signboard\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_c0cd1b9b9c0e\", \"observation\": \"white signboard partially visible\"},\n      {\"context_id\": \"b_42d3a74baab7\", \"observation\": \"white signboard fully visible\"},\n      {\"context_id\": \"b_bc84988e0e0a\", \"observation\": \"white signboard partially visible\"},\n      {\"context_id\": \"b_2f68fc214982\", \"observation\": \"white signboard partially visible\"}\n    ],\n    \"uncertainty\": \"white signboard is consistently visible across all contexts.\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_c0a10b3aef20",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_c0a10b3aef20.png",
      "routing_mask_path": "masks/b_c0a10b3aef20_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.2129696728848162
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.2129696728848162
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_8ffaf4db41bf",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0025.JPG",
    "source_split": "train",
    "source_frame_id": "0025",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0025.JPG",
    "context_ids": [
      "b_d95203c68daa",
      "b_4dd28c1b3b60",
      "b_17642d7e1164",
      "b_42d3a74baab7"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_8ffaf4db41bf",
      "target_sha256": "215efed0325c4c5bfc4cab03cf298ffef068a601d17f729ebdc3803b1505db7d",
      "context_ids": [
        "b_d95203c68daa",
        "b_4dd28c1b3b60",
        "b_17642d7e1164",
        "b_42d3a74baab7"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "mirror",
          "grounding_queries": [
            "mirror"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_d95203c68daa",
              "observation": "mirror is visible"
            },
            {
              "context_id": "b_4dd28c1b3b60",
              "observation": "mirror is visible"
            },
            {
              "context_id": "b_17642d7e1164",
              "observation": "mirror is visible"
            },
            {
              "context_id": "b_42d3a74baab7",
              "observation": "mirror is partially visible"
            }
          ],
          "uncertainty": "mirror is partially visible in some views",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "person holding mirror",
          "grounding_queries": [
            "person holding mirror"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_d95203c68daa",
              "observation": "person holding mirror"
            },
            {
              "context_id": "b_4dd28c1b3b60",
              "observation": "person holding mirror"
            },
            {
              "context_id": "b_17642d7e1164",
              "observation": "person holding mirror"
            },
            {
              "context_id": "b_42d3a74baab7",
              "observation": "person holding mirror"
            }
          ],
          "uncertainty": "person holding mirror is partially visible",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"mirror\",\n    \"grounding_queries\": [\"mirror\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_d95203c68daa\", \"observation\": \"mirror is visible\"},\n      {\"context_id\": \"b_4dd28c1b3b60\", \"observation\": \"mirror is visible\"},\n      {\"context_id\": \"b_17642d7e1164\", \"observation\": \"mirror is visible\"},\n      {\"context_id\": \"b_42d3a74baab7\", \"observation\": \"mirror is partially visible\"}\n    ],\n    \"uncertainty\": \"mirror is partially visible in some views\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"person holding mirror\",\n    \"grounding_queries\": [\"person holding mirror\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_d95203c68daa\", \"observation\": \"person holding mirror\"},\n      {\"context_id\": \"b_4dd28c1b3b60\", \"observation\": \"person holding mirror\"},\n      {\"context_id\": \"b_17642d7e1164\", \"observation\": \"person holding mirror\"},\n      {\"context_id\": \"b_42d3a74baab7\", \"observation\": \"person holding mirror\"}\n    ],\n    \"uncertainty\": \"person holding mirror is partially visible\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_8ffaf4db41bf",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "mirror",
          "queries": [
            "mirror"
          ],
          "source": "grounding",
          "boxes_xyxy": [
            [
              1185.1329345703125,
              284.9571533203125,
              1387.48193359375,
              921.6734619140625
            ]
          ],
          "box_scores": [
            0.2762335240840912
          ],
          "box_queries": [
            "mirror"
          ],
          "sam_predicted_iou": [
            0.93359375
          ],
          "mask_path": "masks/b_8ffaf4db41bf_r1.png"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "person holding mirror",
          "queries": [
            "person holding mirror"
          ],
          "source": "grounding",
          "boxes_xyxy": [
            [
              1185.721435546875,
              284.9627380371094,
              1387.1141357421875,
              921.4432983398438
            ]
          ],
          "box_scores": [
            0.2745721936225891
          ],
          "box_queries": [
            "person holding mirror"
          ],
          "sam_predicted_iou": [
            0.93359375
          ],
          "mask_path": "masks/b_8ffaf4db41bf_r2.png"
        }
      ],
      "final_mask_path": "masks/b_8ffaf4db41bf.png",
      "routing_mask_path": "masks/b_8ffaf4db41bf_routing.png",
      "predicted_fraction": 0.03442384714884598
    },
    "raw_mask_metrics": {
      "iou": 0.13937883928024547,
      "dice": 0.24465758793325,
      "precision": 0.9979910587968989,
      "recall": 0.13941795543258065,
      "safety_recall": 0.13941795543258065,
      "leakage": 0.8605820445674194,
      "overmask": 9.176867314777535e-05,
      "boundary_f5": 0.08186245031215608,
      "iou_ignore_boundary5": 0.14958321922497603,
      "predicted_fraction": 0.03442384714884598,
      "gt_fraction": 0.2464151160253713
    },
    "routing_mask_metrics": {
      "iou": 0.17070194491065258,
      "dice": 0.2916232362177898,
      "precision": 0.9945356448514271,
      "recall": 0.17086219674885122,
      "safety_recall": 0.17086219674885122,
      "leakage": 0.8291378032511487,
      "overmask": 0.0003069726742619246,
      "boundary_f5": 0.15888570258090945,
      "iou_ignore_boundary5": 0.1763979700123237,
      "predicted_fraction": 0.04233435800333505,
      "gt_fraction": 0.2464151160253713
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.2290152675527447,
      "area_precision": 0.9005393474983291,
      "box_fraction": 0.0626655835554101
    },
    "frame_detected": true,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_42d9140f0598",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0026.JPG",
    "source_split": "train",
    "source_frame_id": "0026",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0026.JPG",
    "context_ids": [
      "b_66a730d6ef90",
      "b_3d50f98d854f",
      "b_2d3a471eb3b2",
      "b_4dd28c1b3b60"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_42d9140f0598",
      "target_sha256": "267fc28ff59ccf3e906e9f6236c98d8f7ab103a5101b3228b4fed69b043b83f1",
      "context_ids": [
        "b_66a730d6ef90",
        "b_3d50f98d854f",
        "b_2d3a471eb3b2",
        "b_4dd28c1b3b60"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "mannequin torso",
          "grounding_queries": [
            "mannequin torso"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_66a730d6ef90",
              "observation": "mannequin torso is visible"
            },
            {
              "context_id": "b_3d50f98d854f",
              "observation": "mannequin torso is visible"
            },
            {
              "context_id": "b_2d3a471eb3b2",
              "observation": "mannequin torso is visible"
            },
            {
              "context_id": "b_4dd28c1b3b60",
              "observation": "mannequin torso is partially visible"
            }
          ],
          "uncertainty": "mannequin torso is partially visible in some views",
          "fallback_box_norm_xyxy": [
            0.5,
            0.0,
            1.0,
            0.5
          ],
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_66a730d6ef90",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_3d50f98d854f",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_2d3a471eb3b2",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_4dd28c1b3b60",
              "observation": "red umbrella is partially visible"
            }
          ],
          "uncertainty": "red umbrella is partially visible in some views",
          "fallback_box_norm_xyxy": [
            0.5,
            0.5,
            0.7,
            0.7
          ],
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "color chart",
          "grounding_queries": [
            "color chart"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_66a730d6ef90",
              "observation": "color chart is visible"
            },
            {
              "context_id": "b_3d50f98d854f",
              "observation": "color chart is visible"
            },
            {
              "context_id": "b_2d3a471eb3b2",
              "observation": "color chart is visible"
            },
            {
              "context_id": "b_4dd28c1b3b60",
              "observation": "color chart is partially visible"
            }
          ],
          "uncertainty": "color chart is partially visible in some views",
          "fallback_box_norm_xyxy": [
            0.0,
            0.5,
            0.1,
            0.6
          ],
          "routing_eligible": true,
          "consistency_filter": "passed"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"mannequin torso\",\n    \"grounding_queries\": [\"mannequin torso\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_66a730d6ef90\", \"observation\": \"mannequin torso is visible\"},\n      {\"context_id\": \"b_3d50f98d854f\", \"observation\": \"mannequin torso is visible\"},\n      {\"context_id\": \"b_2d3a471eb3b2\", \"observation\": \"mannequin torso is visible\"},\n      {\"context_id\": \"b_4dd28c1b3b60\", \"observation\": \"mannequin torso is partially visible\"}\n    ],\n    \"uncertainty\": \"mannequin torso is partially visible in some views\",\n    \"fallback_box_norm_xyxy\": [0.5, 0.0, 1.0, 0.5]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_66a730d6ef90\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_3d50f98d854f\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_2d3a471eb3b2\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_4dd28c1b3b60\", \"observation\": \"red umbrella is partially visible\"}\n    ],\n    \"uncertainty\": \"red umbrella is partially visible in some views\",\n    \"fallback_box_norm_xyxy\": [0.5, 0.5, 0.7, 0.7]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"color chart\",\n    \"grounding_queries\": [\"color chart\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_66a730d6ef90\", \"observation\": \"color chart is visible\"},\n      {\"context_id\": \"b_3d50f98d854f\", \"observation\": \"color chart is visible\"},\n      {\"context_id\": \"b_2d3a471eb3b2\", \"observation\": \"color chart is visible\"},\n      {\"context_id\": \"b_4dd28c1b3b60\", \"observation\": \"color chart is partially visible\"}\n    ],\n    \"uncertainty\": \"color chart is partially visible in some views\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.5, 0.1, 0.6]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_42d9140f0598",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "mannequin torso",
          "queries": [
            "mannequin torso"
          ],
          "source": "vlm_fallback",
          "boxes_xyxy": [
            [
              879.0,
              0.0,
              1758.0,
              584.0
            ]
          ],
          "box_scores": [
            0.9
          ],
          "box_queries": [
            "vlm_fallback"
          ],
          "sam_predicted_iou": [
            0.890625
          ],
          "mask_path": "masks/b_42d9140f0598_r1.png"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "queries": [
            "red umbrella"
          ],
          "source": "vlm_fallback",
          "boxes_xyxy": [
            [
              879.0,
              584.0,
              1230.6,
              817.5999999999999
            ]
          ],
          "box_scores": [
            0.9
          ],
          "box_queries": [
            "vlm_fallback"
          ],
          "sam_predicted_iou": [
            0.80078125
          ],
          "mask_path": "masks/b_42d9140f0598_r2.png"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "color chart",
          "queries": [
            "color chart"
          ],
          "source": "vlm_fallback",
          "boxes_xyxy": [
            [
              0.0,
              584.0,
              175.8,
              700.8
            ]
          ],
          "box_scores": [
            0.9
          ],
          "box_queries": [
            "vlm_fallback"
          ],
          "sam_predicted_iou": [
            0.86328125
          ],
          "mask_path": "masks/b_42d9140f0598_r3.png"
        }
      ],
      "final_mask_path": "masks/b_42d9140f0598.png",
      "routing_mask_path": "masks/b_42d9140f0598_routing.png",
      "predicted_fraction": 0.05020201193760032
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.06332833457943811,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.05020201193760032,
      "gt_fraction": 0.20727408558916577
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.08112848558644845,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.06431265292128352,
      "gt_fraction": 0.20727408558916577
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.27558181882261723,
      "area_precision": 0.19031277178498643,
      "box_fraction": 0.3001425966618355
    },
    "frame_detected": true,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_48b647235f33",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0027.JPG",
    "source_split": "train",
    "source_frame_id": "0027",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0027.JPG",
    "context_ids": [
      "b_0a52973d1b8d",
      "b_3fad5029fb0d",
      "b_6125c5000375",
      "b_2962b6747c5c"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_48b647235f33",
      "target_sha256": "4478eb05fc8017524c6130aaf321ae48dcba6afa4f0b8a6f6bfcc8a7608b3d88",
      "context_ids": [
        "b_0a52973d1b8d",
        "b_3fad5029fb0d",
        "b_6125c5000375",
        "b_2962b6747c5c"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "black and white umbrella",
          "grounding_queries": [
            "black and white umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_2962b6747c5c",
              "observation": "appears in one view"
            }
          ],
          "uncertainty": "insufficient evidence for persistence",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "red cloth",
          "grounding_queries": [
            "red cloth"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_6125c5000375",
              "observation": "appears in one view"
            }
          ],
          "uncertainty": "insufficient evidence for persistence",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "green curtain",
          "grounding_queries": [
            "green curtain"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_3fad5029fb0d",
              "observation": "appears in one view"
            }
          ],
          "uncertainty": "insufficient evidence for persistence",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"black and white umbrella\",\n    \"grounding_queries\": [\"black and white umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_2962b6747c5c\", \"observation\": \"appears in one view\"}\n    ],\n    \"uncertainty\": \"insufficient evidence for persistence\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red cloth\",\n    \"grounding_queries\": [\"red cloth\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_6125c5000375\", \"observation\": \"appears in one view\"}\n    ],\n    \"uncertainty\": \"insufficient evidence for persistence\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"green curtain\",\n    \"grounding_queries\": [\"green curtain\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_3fad5029fb0d\", \"observation\": \"appears in one view\"}\n    ],\n    \"uncertainty\": \"insufficient evidence for persistence\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_48b647235f33",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "black and white umbrella",
          "queries": [
            "black and white umbrella"
          ],
          "source": "grounding",
          "boxes_xyxy": [],
          "box_scores": [],
          "box_queries": [],
          "sam_predicted_iou": [],
          "mask_path": "masks/b_48b647235f33_r1.png"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "red cloth",
          "queries": [
            "red cloth"
          ],
          "source": "grounding",
          "boxes_xyxy": [],
          "box_scores": [],
          "box_queries": [],
          "sam_predicted_iou": [],
          "mask_path": "masks/b_48b647235f33_r2.png"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "green curtain",
          "queries": [
            "green curtain"
          ],
          "source": "grounding",
          "boxes_xyxy": [],
          "box_scores": [],
          "box_queries": [],
          "sam_predicted_iou": [],
          "mask_path": "masks/b_48b647235f33_r3.png"
        }
      ],
      "final_mask_path": "masks/b_48b647235f33.png",
      "routing_mask_path": "masks/b_48b647235f33_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.02949919740676672
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.02949919740676672
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.00784313678741455,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_57cf54a3bbc0",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0028.JPG",
    "source_split": "train",
    "source_frame_id": "0028",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0028.JPG",
    "context_ids": [
      "b_0ed8205374dd",
      "b_c7dd5015d023",
      "b_31faa8f0af47",
      "b_3fad5029fb0d"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_57cf54a3bbc0",
      "target_sha256": "afcf569e721329b5cb0850ec96c8f3bced66556ebadc4e3191fdb396f9f7cdcd",
      "context_ids": [
        "b_0ed8205374dd",
        "b_c7dd5015d023",
        "b_31faa8f0af47",
        "b_3fad5029fb0d"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "black curtain",
          "grounding_queries": [
            "black curtain"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_31faa8f0af47",
              "observation": "curtain partially visible"
            },
            {
              "context_id": "b_3fad5029fb0d",
              "observation": "curtain fully visible"
            }
          ],
          "uncertainty": "curtain may be partially obscured in some views",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_0ed8205374dd",
              "observation": "umbrella present"
            },
            {
              "context_id": "b_c7dd5015d023",
              "observation": "umbrella present"
            },
            {
              "context_id": "b_31faa8f0af47",
              "observation": "umbrella present"
            },
            {
              "context_id": "b_3fad5029fb0d",
              "observation": "umbrella present"
            }
          ],
          "uncertainty": "umbrella consistently visible",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"black curtain\",\n    \"grounding_queries\": [\"black curtain\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_31faa8f0af47\", \"observation\": \"curtain partially visible\"},\n      {\"context_id\": \"b_3fad5029fb0d\", \"observation\": \"curtain fully visible\"}\n    ],\n    \"uncertainty\": \"curtain may be partially obscured in some views\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_0ed8205374dd\", \"observation\": \"umbrella present\"},\n      {\"context_id\": \"b_c7dd5015d023\", \"observation\": \"umbrella present\"},\n      {\"context_id\": \"b_31faa8f0af47\", \"observation\": \"umbrella present\"},\n      {\"context_id\": \"b_3fad5029fb0d\", \"observation\": \"umbrella present\"}\n    ],\n    \"uncertainty\": \"umbrella consistently visible\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_57cf54a3bbc0",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "black curtain",
          "queries": [
            "black curtain"
          ],
          "source": "grounding",
          "boxes_xyxy": [],
          "box_scores": [],
          "box_queries": [],
          "sam_predicted_iou": [],
          "mask_path": "masks/b_57cf54a3bbc0_r1.png"
        }
      ],
      "final_mask_path": "masks/b_57cf54a3bbc0.png",
      "routing_mask_path": "masks/b_57cf54a3bbc0_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.14584502158430346
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.14584502158430346
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_39a6a990391f",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0029.JPG",
    "source_split": "train",
    "source_frame_id": "0029",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0029.JPG",
    "context_ids": [
      "b_1ea940c8ac4e",
      "b_31faa8f0af47",
      "b_f9646cc9c954",
      "b_0ed8205374dd"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_39a6a990391f",
      "target_sha256": "f219da94ff53ac815426911c9fd887e919db470e608cea05b3894efd77048001",
      "context_ids": [
        "b_1ea940c8ac4e",
        "b_31faa8f0af47",
        "b_f9646cc9c954",
        "b_0ed8205374dd"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "mirror",
          "grounding_queries": [
            "mirror"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_1ea940c8ac4e",
              "observation": "mirror is visible"
            },
            {
              "context_id": "b_31faa8f0af47",
              "observation": "mirror is visible"
            },
            {
              "context_id": "b_f9646cc9c954",
              "observation": "mirror is visible"
            },
            {
              "context_id": "b_0ed8205374dd",
              "observation": "mirror is visible"
            }
          ],
          "uncertainty": "mirror is consistently visible across all views",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "umbrella",
          "grounding_queries": [
            "umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_1ea940c8ac4e",
              "observation": "umbrella is visible"
            },
            {
              "context_id": "b_31faa8f0af47",
              "observation": "umbrella is visible"
            },
            {
              "context_id": "b_f9646cc9c954",
              "observation": "umbrella is visible"
            },
            {
              "context_id": "b_0ed8205374dd",
              "observation": "umbrella is visible"
            }
          ],
          "uncertainty": "umbrella is consistently visible across all views",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "pillow",
          "grounding_queries": [
            "pillow"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_1ea940c8ac4e",
              "observation": "pillow is visible"
            },
            {
              "context_id": "b_31faa8f0af47",
              "observation": "pillow is visible"
            },
            {
              "context_id": "b_f9646cc9c954",
              "observation": "pillow is visible"
            },
            {
              "context_id": "b_0ed8205374dd",
              "observation": "pillow is visible"
            }
          ],
          "uncertainty": "pillow is consistently visible across all views",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"mirror\",\n    \"grounding_queries\": [\"mirror\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_1ea940c8ac4e\", \"observation\": \"mirror is visible\"},\n      {\"context_id\": \"b_31faa8f0af47\", \"observation\": \"mirror is visible\"},\n      {\"context_id\": \"b_f9646cc9c954\", \"observation\": \"mirror is visible\"},\n      {\"context_id\": \"b_0ed8205374dd\", \"observation\": \"mirror is visible\"}\n    ],\n    \"uncertainty\": \"mirror is consistently visible across all views\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"umbrella\",\n    \"grounding_queries\": [\"umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_1ea940c8ac4e\", \"observation\": \"umbrella is visible\"},\n      {\"context_id\": \"b_31faa8f0af47\", \"observation\": \"umbrella is visible\"},\n      {\"context_id\": \"b_f9646cc9c954\", \"observation\": \"umbrella is visible\"},\n      {\"context_id\": \"b_0ed8205374dd\", \"observation\": \"umbrella is visible\"}\n    ],\n    \"uncertainty\": \"umbrella is consistently visible across all views\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"pillow\",\n    \"grounding_queries\": [\"pillow\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_1ea940c8ac4e\", \"observation\": \"pillow is visible\"},\n      {\"context_id\": \"b_31faa8f0af47\", \"observation\": \"pillow is visible\"},\n      {\"context_id\": \"b_f9646cc9c954\", \"observation\": \"pillow is visible\"},\n      {\"context_id\": \"b_0ed8205374dd\", \"observation\": \"pillow is visible\"}\n    ],\n    \"uncertainty\": \"pillow is consistently visible across all views\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_39a6a990391f",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_39a6a990391f.png",
      "routing_mask_path": "masks/b_39a6a990391f_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.15133022036249164
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.15133022036249164
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_e21902361190",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/train/0030.JPG",
    "source_split": "train",
    "source_frame_id": "0030",
    "clean_reference_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/val/0030.JPG",
    "context_ids": [
      "b_6c74696a698a",
      "b_c7168d6511cd",
      "b_442bffc55440",
      "b_f2d05c389498"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_e21902361190",
      "target_sha256": "7aad75333082d3e970f7577b36d73f84a6996115d0084e245bc54f56316fa30a",
      "context_ids": [
        "b_6c74696a698a",
        "b_c7168d6511cd",
        "b_442bffc55440",
        "b_f2d05c389498"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_6c74696a698a",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_c7168d6511cd",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_442bffc55440",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_f2d05c389498",
              "observation": "red umbrella is visible"
            }
          ],
          "uncertainty": "red umbrella is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "white flower bouquet",
          "grounding_queries": [
            "white flower bouquet"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_6c74696a698a",
              "observation": "white flower bouquet is visible"
            },
            {
              "context_id": "b_c7168d6511cd",
              "observation": "white flower bouquet is visible"
            },
            {
              "context_id": "b_442bffc55440",
              "observation": "white flower bouquet is visible"
            },
            {
              "context_id": "b_f2d05c389498",
              "observation": "white flower bouquet is visible"
            }
          ],
          "uncertainty": "white flower bouquet is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "green wall",
          "grounding_queries": [
            "green wall"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_6c74696a698a",
              "observation": "green wall is visible"
            },
            {
              "context_id": "b_c7168d6511cd",
              "observation": "green wall is visible"
            },
            {
              "context_id": "b_442bffc55440",
              "observation": "green wall is visible"
            },
            {
              "context_id": "b_f2d05c389498",
              "observation": "green wall is visible"
            }
          ],
          "uncertainty": "green wall is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_6c74696a698a\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_c7168d6511cd\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_442bffc55440\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_f2d05c389498\", \"observation\": \"red umbrella is visible\"}\n    ],\n    \"uncertainty\": \"red umbrella is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [300, 270, 440, 380]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"white flower bouquet\",\n    \"grounding_queries\": [\"white flower bouquet\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_6c74696a698a\", \"observation\": \"white flower bouquet is visible\"},\n      {\"context_id\": \"b_c7168d6511cd\", \"observation\": \"white flower bouquet is visible\"},\n      {\"context_id\": \"b_442bffc55440\", \"observation\": \"white flower bouquet is visible\"},\n      {\"context_id\": \"b_f2d05c389498\", \"observation\": \"white flower bouquet is visible\"}\n    ],\n    \"uncertainty\": \"white flower bouquet is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [250, 100, 350, 200]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"green wall\",\n    \"grounding_queries\": [\"green wall\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_6c74696a698a\", \"observation\": \"green wall is visible\"},\n      {\"context_id\": \"b_c7168d6511cd\", \"observation\": \"green wall is visible\"},\n      {\"context_id\": \"b_442bffc55440\", \"observation\": \"green wall is visible\"},\n      {\"context_id\": \"b_f2d05c389498\", \"observation\": \"green wall is visible\"}\n    ],\n    \"uncertainty\": \"green wall is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [0, 0, 300, 300]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_e21902361190",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_e21902361190.png",
      "routing_mask_path": "masks/b_e21902361190_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.21333736577991802
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 0.0,
      "safety_recall": 0.0,
      "leakage": 1.0,
      "overmask": 0.0,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.21333736577991802
    },
    "box_metrics": {
      "gt_pixel_coverage": 0.0,
      "area_precision": 0.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.011764705181121826,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_442bffc55440",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/test/0031.JPG",
    "source_split": "test",
    "source_frame_id": "0031",
    "clean_reference_path": null,
    "context_ids": [
      "b_63ebbf87a7c1",
      "b_a38e3ce498d5",
      "b_e21902361190",
      "b_6c74696a698a"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_442bffc55440",
      "target_sha256": "ea5e203e121b9e94bef7b63ba762139d4a7bfd98071db955ef86ed4c1923dbc0",
      "context_ids": [
        "b_63ebbf87a7c1",
        "b_a38e3ce498d5",
        "b_e21902361190",
        "b_6c74696a698a"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_63ebbf87a7c1",
              "observation": "umbrella is visible"
            },
            {
              "context_id": "b_a38e3ce498d5",
              "observation": "umbrella is visible"
            },
            {
              "context_id": "b_e21902361190",
              "observation": "umbrella is visible"
            },
            {
              "context_id": "b_6c74696a698a",
              "observation": "umbrella is visible"
            }
          ],
          "uncertainty": "umbrella is consistently visible across all contexts",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "sign with Chinese characters",
          "grounding_queries": [
            "sign with Chinese characters"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_63ebbf87a7c1",
              "observation": "sign is visible"
            },
            {
              "context_id": "b_a38e3ce498d5",
              "observation": "sign is visible"
            },
            {
              "context_id": "b_e21902361190",
              "observation": "sign is visible"
            },
            {
              "context_id": "b_6c74696a698a",
              "observation": "sign is visible"
            }
          ],
          "uncertainty": "sign is consistently visible across all contexts",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_63ebbf87a7c1\", \"observation\": \"umbrella is visible\"},\n      {\"context_id\": \"b_a38e3ce498d5\", \"observation\": \"umbrella is visible\"},\n      {\"context_id\": \"b_e21902361190\", \"observation\": \"umbrella is visible\"},\n      {\"context_id\": \"b_6c74696a698a\", \"observation\": \"umbrella is visible\"}\n    ],\n    \"uncertainty\": \"umbrella is consistently visible across all contexts\",\n    \"fallback_box_norm_xyxy\": [294, 267, 400, 370]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"sign with Chinese characters\",\n    \"grounding_queries\": [\"sign with Chinese characters\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_63ebbf87a7c1\", \"observation\": \"sign is visible\"},\n      {\"context_id\": \"b_a38e3ce498d5\", \"observation\": \"sign is visible\"},\n      {\"context_id\": \"b_e21902361190\", \"observation\": \"sign is visible\"},\n      {\"context_id\": \"b_6c74696a698a\", \"observation\": \"sign is visible\"}\n    ],\n    \"uncertainty\": \"sign is consistently visible across all contexts\",\n    \"fallback_box_norm_xyxy\": [220, 400, 290, 476]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_442bffc55440",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_442bffc55440.png",
      "routing_mask_path": "masks/b_442bffc55440_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 1.0,
      "dice": 1.0,
      "precision": 1.0,
      "recall": 1.0,
      "safety_recall": 1.0,
      "leakage": 0.0,
      "overmask": 0.0,
      "boundary_f5": 1.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.0
    },
    "routing_mask_metrics": {
      "iou": 1.0,
      "dice": 1.0,
      "precision": 1.0,
      "recall": 1.0,
      "safety_recall": 1.0,
      "leakage": 0.0,
      "overmask": 0.0,
      "boundary_f5": 1.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.0
    },
    "box_metrics": {
      "gt_pixel_coverage": 1.0,
      "area_precision": 1.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.0,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_f2d05c389498",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/test/0032.JPG",
    "source_split": "test",
    "source_frame_id": "0032",
    "clean_reference_path": null,
    "context_ids": [
      "b_f9646cc9c954",
      "b_e21902361190",
      "b_c7168d6511cd",
      "b_1ea940c8ac4e"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_f2d05c389498",
      "target_sha256": "ce4ed46f67474b370e54193ad6ad1e8f2e0bf59b35a2c5e4e38dfe5a707dd6b6",
      "context_ids": [
        "b_f9646cc9c954",
        "b_e21902361190",
        "b_c7168d6511cd",
        "b_1ea940c8ac4e"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_f9646cc9c954",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_e21902361190",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_c7168d6511cd",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_1ea940c8ac4e",
              "observation": "red umbrella is visible"
            }
          ],
          "uncertainty": "red umbrella is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "black and white umbrella",
          "grounding_queries": [
            "black and white umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_f9646cc9c954",
              "observation": "black and white umbrella is visible"
            },
            {
              "context_id": "b_e21902361190",
              "observation": "black and white umbrella is visible"
            },
            {
              "context_id": "b_c7168d6511cd",
              "observation": "black and white umbrella is visible"
            },
            {
              "context_id": "b_1ea940c8ac4e",
              "observation": "black and white umbrella is visible"
            }
          ],
          "uncertainty": "black and white umbrella is consistently visible across all context images",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_f9646cc9c954\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_e21902361190\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_c7168d6511cd\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_1ea940c8ac4e\", \"observation\": \"red umbrella is visible\"}\n    ],\n    \"uncertainty\": \"red umbrella is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [320, 224, 450, 345]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"black and white umbrella\",\n    \"grounding_queries\": [\"black and white umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_f9646cc9c954\", \"observation\": \"black and white umbrella is visible\"},\n      {\"context_id\": \"b_e21902361190\", \"observation\": \"black and white umbrella is visible\"},\n      {\"context_id\": \"b_c7168d6511cd\", \"observation\": \"black and white umbrella is visible\"},\n      {\"context_id\": \"b_1ea940c8ac4e\", \"observation\": \"black and white umbrella is visible\"}\n    ],\n    \"uncertainty\": \"black and white umbrella is consistently visible across all context images\",\n    \"fallback_box_norm_xyxy\": [140, 300, 260, 420]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_f2d05c389498",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_f2d05c389498.png",
      "routing_mask_path": "masks/b_f2d05c389498_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 1.0,
      "dice": 1.0,
      "precision": 1.0,
      "recall": 1.0,
      "safety_recall": 1.0,
      "leakage": 0.0,
      "overmask": 0.0,
      "boundary_f5": 1.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.0
    },
    "routing_mask_metrics": {
      "iou": 1.0,
      "dice": 1.0,
      "precision": 1.0,
      "recall": 1.0,
      "safety_recall": 1.0,
      "leakage": 0.0,
      "overmask": 0.0,
      "boundary_f5": 1.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.0
    },
    "box_metrics": {
      "gt_pixel_coverage": 1.0,
      "area_precision": 1.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.0,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_0a52973d1b8d",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/test/0033.JPG",
    "source_split": "test",
    "source_frame_id": "0033",
    "clean_reference_path": null,
    "context_ids": [
      "b_48b647235f33",
      "b_3fad5029fb0d",
      "b_6125c5000375",
      "b_c7dd5015d023"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_0a52973d1b8d",
      "target_sha256": "02d3286042e0c5392fa100b8c69e3a41868ea2bbd5a3cc9429cbb1b3789f65d7",
      "context_ids": [
        "b_48b647235f33",
        "b_3fad5029fb0d",
        "b_6125c5000375",
        "b_c7dd5015d023"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "grounding_queries": [
            "red umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_48b647235f33",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_3fad5029fb0d",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_6125c5000375",
              "observation": "red umbrella is visible"
            },
            {
              "context_id": "b_c7dd5015d023",
              "observation": "red umbrella is not visible"
            }
          ],
          "uncertainty": "red umbrella is not visible in the last context image",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "green curtain",
          "grounding_queries": [
            "green curtain"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_48b647235f33",
              "observation": "green curtain is visible"
            },
            {
              "context_id": "b_3fad5029fb0d",
              "observation": "green curtain is visible"
            },
            {
              "context_id": "b_6125c5000375",
              "observation": "green curtain is visible"
            },
            {
              "context_id": "b_c7dd5015d023",
              "observation": "green curtain is not visible"
            }
          ],
          "uncertainty": "green curtain is not visible in the last context image",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "black chair",
          "grounding_queries": [
            "black chair"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_48b647235f33",
              "observation": "black chair is visible"
            },
            {
              "context_id": "b_3fad5029fb0d",
              "observation": "black chair is visible"
            },
            {
              "context_id": "b_6125c5000375",
              "observation": "black chair is visible"
            },
            {
              "context_id": "b_c7dd5015d023",
              "observation": "black chair is not visible"
            }
          ],
          "uncertainty": "black chair is not visible in the last context image",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"red umbrella\",\n    \"grounding_queries\": [\"red umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_48b647235f33\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_3fad5029fb0d\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_6125c5000375\", \"observation\": \"red umbrella is visible\"},\n      {\"context_id\": \"b_c7dd5015d023\", \"observation\": \"red umbrella is not visible\"}\n    ],\n    \"uncertainty\": \"red umbrella is not visible in the last context image\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"green curtain\",\n    \"grounding_queries\": [\"green curtain\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_48b647235f33\", \"observation\": \"green curtain is visible\"},\n      {\"context_id\": \"b_3fad5029fb0d\", \"observation\": \"green curtain is visible\"},\n      {\"context_id\": \"b_6125c5000375\", \"observation\": \"green curtain is visible\"},\n      {\"context_id\": \"b_c7dd5015d023\", \"observation\": \"green curtain is not visible\"}\n    ],\n    \"uncertainty\": \"green curtain is not visible in the last context image\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r3\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"black chair\",\n    \"grounding_queries\": [\"black chair\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_48b647235f33\", \"observation\": \"black chair is visible\"},\n      {\"context_id\": \"b_3fad5029fb0d\", \"observation\": \"black chair is visible\"},\n      {\"context_id\": \"b_6125c5000375\", \"observation\": \"black chair is visible\"},\n      {\"context_id\": \"b_c7dd5015d023\", \"observation\": \"black chair is not visible\"}\n    ],\n    \"uncertainty\": \"black chair is not visible in the last context image\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_0a52973d1b8d",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "red umbrella",
          "queries": [
            "red umbrella"
          ],
          "source": "grounding",
          "boxes_xyxy": [],
          "box_scores": [],
          "box_queries": [],
          "sam_predicted_iou": [],
          "mask_path": "masks/b_0a52973d1b8d_r1.png"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "green curtain",
          "queries": [
            "green curtain"
          ],
          "source": "grounding",
          "boxes_xyxy": [],
          "box_scores": [],
          "box_queries": [],
          "sam_predicted_iou": [],
          "mask_path": "masks/b_0a52973d1b8d_r2.png"
        },
        {
          "region_id": "r3",
          "cause": "transient_occluder",
          "semantic_description": "black chair",
          "queries": [
            "black chair"
          ],
          "source": "grounding",
          "boxes_xyxy": [],
          "box_scores": [],
          "box_queries": [],
          "sam_predicted_iou": [],
          "mask_path": "masks/b_0a52973d1b8d_r3.png"
        }
      ],
      "final_mask_path": "masks/b_0a52973d1b8d.png",
      "routing_mask_path": "masks/b_0a52973d1b8d_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 1.0,
      "dice": 1.0,
      "precision": 1.0,
      "recall": 1.0,
      "safety_recall": 1.0,
      "leakage": 0.0,
      "overmask": 0.0,
      "boundary_f5": 1.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.0
    },
    "routing_mask_metrics": {
      "iou": 1.0,
      "dice": 1.0,
      "precision": 1.0,
      "recall": 1.0,
      "safety_recall": 1.0,
      "leakage": 0.0,
      "overmask": 0.0,
      "boundary_f5": 1.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.0
    },
    "box_metrics": {
      "gt_pixel_coverage": 1.0,
      "area_precision": 1.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.0,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_66a730d6ef90",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/test/0034.JPG",
    "source_split": "test",
    "source_frame_id": "0034",
    "clean_reference_path": null,
    "context_ids": [
      "b_42d9140f0598",
      "b_3d50f98d854f",
      "b_2d3a471eb3b2",
      "b_2962b6747c5c"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_66a730d6ef90",
      "target_sha256": "2caa6d5ab56862347537483d461fa637f435a1d2209c4e4106f78cab764e5592",
      "context_ids": [
        "b_42d9140f0598",
        "b_3d50f98d854f",
        "b_2d3a471eb3b2",
        "b_2962b6747c5c"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "mannequin",
          "grounding_queries": [
            "mannequin"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_42d9140f0598",
              "observation": "mannequin partially obscures the teddy bear"
            },
            {
              "context_id": "b_3d50f98d854f",
              "observation": "mannequin partially obscures the teddy bear"
            },
            {
              "context_id": "b_2d3a471eb3b2",
              "observation": "mannequin partially obscures the teddy bear"
            },
            {
              "context_id": "b_2962b6747c5c",
              "observation": "mannequin partially obscures the teddy bear"
            }
          ],
          "uncertainty": "mannequin is partially visible in all context images, indicating it is not a spatial nuisance.",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        },
        {
          "region_id": "r2",
          "cause": "transient_occluder",
          "semantic_description": "umbrella",
          "grounding_queries": [
            "umbrella"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_42d9140f0598",
              "observation": "umbrella partially obscures the teddy bear"
            },
            {
              "context_id": "b_3d50f98d854f",
              "observation": "umbrella partially obscures the teddy bear"
            },
            {
              "context_id": "b_2d3a471eb3b2",
              "observation": "umbrella partially obscures the teddy bear"
            },
            {
              "context_id": "b_2962b6747c5c",
              "observation": "umbrella partially obscures the teddy bear"
            }
          ],
          "uncertainty": "umbrella is partially visible in all context images, indicating it is not a spatial nuisance.",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": false,
          "consistency_filter": "rejected_explicitly_persistent"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"mannequin\",\n    \"grounding_queries\": [\"mannequin\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_42d9140f0598\", \"observation\": \"mannequin partially obscures the teddy bear\"},\n      {\"context_id\": \"b_3d50f98d854f\", \"observation\": \"mannequin partially obscures the teddy bear\"},\n      {\"context_id\": \"b_2d3a471eb3b2\", \"observation\": \"mannequin partially obscures the teddy bear\"},\n      {\"context_id\": \"b_2962b6747c5c\", \"observation\": \"mannequin partially obscures the teddy bear\"}\n    ],\n    \"uncertainty\": \"mannequin is partially visible in all context images, indicating it is not a spatial nuisance.\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  },\n  {\n    \"region_id\": \"r2\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"umbrella\",\n    \"grounding_queries\": [\"umbrella\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_42d9140f0598\", \"observation\": \"umbrella partially obscures the teddy bear\"},\n      {\"context_id\": \"b_3d50f98d854f\", \"observation\": \"umbrella partially obscures the teddy bear\"},\n      {\"context_id\": \"b_2d3a471eb3b2\", \"observation\": \"umbrella partially obscures the teddy bear\"},\n      {\"context_id\": \"b_2962b6747c5c\", \"observation\": \"umbrella partially obscures the teddy bear\"}\n    ],\n    \"uncertainty\": \"umbrella is partially visible in all context images, indicating it is not a spatial nuisance.\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_66a730d6ef90",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [],
      "final_mask_path": "masks/b_66a730d6ef90.png",
      "routing_mask_path": "masks/b_66a730d6ef90_routing.png",
      "predicted_fraction": 0.0
    },
    "raw_mask_metrics": {
      "iou": 1.0,
      "dice": 1.0,
      "precision": 1.0,
      "recall": 1.0,
      "safety_recall": 1.0,
      "leakage": 0.0,
      "overmask": 0.0,
      "boundary_f5": 1.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.0
    },
    "routing_mask_metrics": {
      "iou": 1.0,
      "dice": 1.0,
      "precision": 1.0,
      "recall": 1.0,
      "safety_recall": 1.0,
      "leakage": 0.0,
      "overmask": 0.0,
      "boundary_f5": 1.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.0,
      "gt_fraction": 0.0
    },
    "box_metrics": {
      "gt_pixel_coverage": 1.0,
      "area_precision": 1.0,
      "box_fraction": 0.0
    },
    "frame_detected": false,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.0,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  },
  {
    "sample_id": "b_17642d7e1164",
    "source_path": "/local/scratch-2/fy277/temp/diag3dgs/data/realx3d/data_4/dynamic/Ujikintoki/test/0035.JPG",
    "source_split": "test",
    "source_frame_id": "0035",
    "clean_reference_path": null,
    "context_ids": [
      "b_42d3a74baab7",
      "b_d95203c68daa",
      "b_c0cd1b9b9c0e",
      "b_8ffaf4db41bf"
    ],
    "vlm": {
      "schema_version": "diag.region.v1",
      "sample_id": "b_17642d7e1164",
      "target_sha256": "ee1fbf84bece876f54c8a63e2d4bbff0086d2ee90c0ed535b19e4040d6381f54",
      "context_ids": [
        "b_42d3a74baab7",
        "b_d95203c68daa",
        "b_c0cd1b9b9c0e",
        "b_8ffaf4db41bf"
      ],
      "model": {
        "provider": "local",
        "name": "qwen2.5-vl-7b-instruct",
        "version": "local_snapshot",
        "quantization": "nf4"
      },
      "prompt_sha256": "15497acb4f0a5266ff625a2745e66a38aa75c362c8f6e3fe7b4bfb2bc5003a36",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "mirror",
          "grounding_queries": [
            "mirror"
          ],
          "is_spatial": true,
          "confidence": 0.9,
          "context_evidence": [
            {
              "context_id": "b_42d3a74baab7",
              "observation": "mirror partially obscures the teddy bear"
            },
            {
              "context_id": "b_d95203c68daa",
              "observation": "mirror fully obscures the teddy bear"
            },
            {
              "context_id": "b_c0cd1b9b9c0e",
              "observation": "mirror partially obscures the teddy bear"
            },
            {
              "context_id": "b_8ffaf4db41bf",
              "observation": "mirror fully obscures the teddy bear"
            }
          ],
          "uncertainty": "mirror's position varies across views",
          "fallback_box_norm_xyxy": null,
          "routing_eligible": true,
          "consistency_filter": "passed"
        }
      ],
      "global_causes": [],
      "parse_error": null,
      "format_repair": "wrapped_top_level_region_list",
      "validation_errors": []
    },
    "vlm_raw": "```json\n[\n  {\n    \"region_id\": \"r1\",\n    \"cause\": \"transient_occluder\",\n    \"semantic_description\": \"mirror\",\n    \"grounding_queries\": [\"mirror\"],\n    \"is_spatial\": true,\n    \"confidence\": 0.9,\n    \"context_evidence\": [\n      {\"context_id\": \"b_42d3a74baab7\", \"observation\": \"mirror partially obscures the teddy bear\"},\n      {\"context_id\": \"b_d95203c68daa\", \"observation\": \"mirror fully obscures the teddy bear\"},\n      {\"context_id\": \"b_c0cd1b9b9c0e\", \"observation\": \"mirror partially obscures the teddy bear\"},\n      {\"context_id\": \"b_8ffaf4db41bf\", \"observation\": \"mirror fully obscures the teddy bear\"}\n    ],\n    \"uncertainty\": \"mirror's position varies across views\",\n    \"fallback_box_norm_xyxy\": [0.0, 0.0, 1.0, 1.0]\n  }\n]\n```",
    "grounding": {
      "schema_version": "diag.ground_sam.v1",
      "sample_id": "b_17642d7e1164",
      "thresholds": {
        "vlm_confidence": 0.35,
        "box": 0.25,
        "text": 0.2,
        "nms": 0.5
      },
      "grounding_model": "/local/scratch-2/fy277/temp/diag3dgs/models/grounding-dino-tiny",
      "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
      "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
      "regions": [
        {
          "region_id": "r1",
          "cause": "transient_occluder",
          "semantic_description": "mirror",
          "queries": [
            "mirror"
          ],
          "source": "grounding",
          "boxes_xyxy": [
            [
              592.1898193359375,
              452.15679931640625,
              883.3391723632812,
              680.7554931640625
            ]
          ],
          "box_scores": [
            0.3116818070411682
          ],
          "box_queries": [
            "mirror"
          ],
          "sam_predicted_iou": [
            0.9296875
          ],
          "mask_path": "masks/b_17642d7e1164_r1.png"
        }
      ],
      "final_mask_path": "masks/b_17642d7e1164.png",
      "routing_mask_path": "masks/b_17642d7e1164_routing.png",
      "predicted_fraction": 0.015087096950145713
    },
    "raw_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 1.0,
      "safety_recall": 1.0,
      "leakage": 0.0,
      "overmask": 0.015087096950145713,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.015087096950145713,
      "gt_fraction": 0.0
    },
    "routing_mask_metrics": {
      "iou": 0.0,
      "dice": 0.0,
      "precision": 0.0,
      "recall": 1.0,
      "safety_recall": 1.0,
      "leakage": 0.0,
      "overmask": 0.018559968519644053,
      "boundary_f5": 0.0,
      "iou_ignore_boundary5": 0.0,
      "predicted_fraction": 0.018559968519644053,
      "gt_fraction": 0.0
    },
    "box_metrics": {
      "gt_pixel_coverage": 1.0,
      "area_precision": 0.0,
      "box_fraction": 0.03245388984992286
    },
    "frame_detected": true,
    "gt_ambiguity": {
      "median_rgb_max_difference": 0.0,
      "note": "paired-reference pseudo-GT may include lighting, shadow, reflection, and registration differences"
    }
  }
]
