diff --git "a/qa/xbd_train.jsonl" "b/qa/xbd_train.jsonl"
new file mode 100644--- /dev/null
+++ "b/qa/xbd_train.jsonl"
@@ -0,0 +1,1212 @@
+{"query_id": "XBD_Q1_santa-rosa-wildfire_00000244_0001", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000244", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a fire event (santa-rosa-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 47, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 47, "classified_buildings": 47, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 125585.814, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_santa-rosa-wildfire_00000244_0002", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000244", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "For a post-disaster fire scene (santa-rosa-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 22, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 54908.6606, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 25, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 70677.1534, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 22, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 54908.6606, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 25, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 70677.1534, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_santa-rosa-wildfire_00000244_0003", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000244", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 9, "severe_count": 0, "class_histogram": {"no_damage": 9, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 25249.7702, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 27554.6709, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 31512.7333, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_santa-rosa-wildfire_00000244_0004", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000244", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (santa-rosa-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "santa-rosa-wildfire", "disaster_type": "fire", "capture_date": "2017-10-11T19:19:41.000Z", "gsd": 1.8769937, "damage_summary": {"total_buildings": 47, "classified_buildings": 47, "counts": {"no-damage": 47, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 125585.814, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 1010226, "1": 38350, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00001058_0005", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001058", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 25, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 25, "classified_buildings": 25, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 57102.0001, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00001058_0006", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001058", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 930.9724, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 21, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 45588.6829, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 930.9724, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 21, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 45588.6829, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 10582.3448, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00001058_0007", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001058", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00001058_0008", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001058", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 25, "classified_buildings": 25, "counts": {"no-damage": 25, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 57102.0001, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_sunda-tsunami_00000008_0009", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000008", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a tsunami event (sunda-tsunami) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 18, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 18, "classified_buildings": 18, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 73678.38, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_sunda-tsunami_00000008_0010", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000008", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "no-low-damage", "question": "For a post-disaster tsunami scene (sunda-tsunami), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "none", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6846.982, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6846.982, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 17, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 66831.398, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_sunda-tsunami_00000008_0011", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000008", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster tsunami satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_sunda-tsunami_00000008_0012", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000008", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster tsunami scene (sunda-tsunami). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "sunda-tsunami", "disaster_type": "tsunami", "capture_date": "2019-01-02T03:17:07.000Z", "gsd": 1.7746705, "damage_summary": {"total_buildings": 18, "classified_buildings": 18, "counts": {"no-damage": 18, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 73678.38, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00001588_0017", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001588", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 24, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "total_buildings": 25, "classified_buildings": 24, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 102357.5785, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00001588_0018", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001588", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "none", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 13, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 62601.1079, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 13, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 62601.1079, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 39756.4706, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00001588_0019", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001588", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00001588_0020", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001588", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 25, "classified_buildings": 24, "counts": {"no-damage": 24, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 102357.5785, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_midwest-flooding_00000024_0025", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000024", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a flooding event (midwest-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 277, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 27}, "total_buildings": 304, "classified_buildings": 277, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 832475.6407, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_midwest-flooding_00000024_0026", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000024", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "For a post-disaster flooding scene (midwest-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 82, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 177744.1891, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 65, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 246039.7726, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 80, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 219433.3192, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 82, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 177744.1891, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 65, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 246039.7726, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 80, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 219433.3192, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 77, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 189258.3598, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_midwest-flooding_00000024_0027", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000024", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 23, "severe_count": 0, "class_histogram": {"no_damage": 23, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 23, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 45949.0765, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 19, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 54509.9842, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 43853.3682, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_midwest-flooding_00000024_0028", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000024", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (midwest-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "midwest-flooding", "disaster_type": "flooding", "capture_date": "2019-05-31T16:54:31.000Z", "gsd": 1.7392077, "damage_summary": {"total_buildings": 304, "classified_buildings": 277, "counts": {"no-damage": 277, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 27}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 832475.6407, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 770341, "1": 278235, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_tuscaloosa-tornado_00000000_0029", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000000", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a wind event (tuscaloosa-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 9, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 4}, "total_buildings": 13, "classified_buildings": 9, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 59561.3682, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_tuscaloosa-tornado_00000000_0030", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000000", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "For a post-disaster wind scene (tuscaloosa-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 5691.4625, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 52217.0505, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 5691.4625, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 52217.0505, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1652.8552, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_tuscaloosa-tornado_00000000_0031", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000000", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 3, "severe_count": 0, "class_histogram": {"no_damage": 3, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 4997.3557, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_tuscaloosa-tornado_00000000_0032", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000000", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (tuscaloosa-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "tuscaloosa-tornado", "disaster_type": "wind", "capture_date": "2011-05-19T16:48:01.085Z", "gsd": 1.85690224170685, "damage_summary": {"total_buildings": 13, "classified_buildings": 9, "counts": {"no-damage": 9, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 4}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 59561.3682, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-florence_00000398_0033", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000398", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a flooding event (hurricane-florence) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 28, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 28, "classified_buildings": 28, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 153712.211, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-florence_00000398_0034", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000398", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "For a post-disaster flooding scene (hurricane-florence), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 33028.0691, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 68394.9616, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 15599.4592, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 33028.0691, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 68394.9616, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 15599.4592, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 36689.7211, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-florence_00000398_0035", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000398", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 1, "severe_count": 0, "class_histogram": {"no_damage": 1, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6749.3391, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 19307.5014, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 11262.6597, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-florence_00000398_0036", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000398", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-florence). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-florence", "disaster_type": "flooding", "capture_date": "2018-09-20T16:04:41.000Z", "gsd": 2.0916247, "damage_summary": {"total_buildings": 28, "classified_buildings": 28, "counts": {"no-damage": 28, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 153712.211, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 1011740, "1": 36836, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_moore-tornado_00000076_0037", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "moore-tornado_00000076", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a wind event (moore-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 240, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 240, "classified_buildings": 240, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 464307.9513, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_moore-tornado_00000076_0038", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "moore-tornado_00000076", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "For a post-disaster wind scene (moore-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 94, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 157403.5979, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 96, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 161544.1954, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 46, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 131516.7388, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 94, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 157403.5979, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 96, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 161544.1954, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 46, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 131516.7388, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 13843.4192, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_moore-tornado_00000076_0039", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "moore-tornado_00000076", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 27, "severe_count": 0, "class_histogram": {"no_damage": 27, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 27, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 45065.1239, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 28, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 47621.2088, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 22, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 43012.1931, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_moore-tornado_00000076_0040", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "moore-tornado_00000076", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (moore-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "moore-tornado", "disaster_type": "wind", "capture_date": "2013-05-22T17:26:30.085Z", "gsd": 1.72984850406647, "damage_summary": {"total_buildings": 240, "classified_buildings": 240, "counts": {"no-damage": 240, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 464307.9513, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_mexico-earthquake_00000021_0041", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000021", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a earthquake event (mexico-earthquake) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 283, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 283, "classified_buildings": 283, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 2290176.9368, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_mexico-earthquake_00000021_0042", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000021", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "no-low-damage", "question": "For a post-disaster earthquake scene (mexico-earthquake), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 78, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 616889.045, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 84, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 536072.0581, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 59, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 542706.5564, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 78, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 616889.045, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 84, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 536072.0581, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 59, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 542706.5564, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 62, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 594509.2773, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_mexico-earthquake_00000021_0043", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000021", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster earthquake satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 27, "severe_count": 0, "class_histogram": {"no_damage": 27, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 27, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 138105.9233, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 17, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 190165.7398, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 112487.3402, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_mexico-earthquake_00000021_0044", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000021", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster earthquake scene (mexico-earthquake). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "mexico-earthquake", "disaster_type": "earthquake", "capture_date": "2017-09-20T17:46:11.000Z", "gsd": 2.6503215, "damage_summary": {"total_buildings": 283, "classified_buildings": 283, "counts": {"no-damage": 283, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 2290176.9368, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 705361, "1": 343215, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-harvey_00000016_0045", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000016", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a flooding event (hurricane-harvey) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 41, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 41, "classified_buildings": 41, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 1474194.2331, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-harvey_00000016_0046", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000016", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "For a post-disaster flooding scene (hurricane-harvey), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1327951.623, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 29, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 146242.6101, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1327951.623, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 29, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 146242.6101, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-harvey_00000016_0047", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000016", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 53629.1629, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 22, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 113923.0484, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-harvey_00000016_0048", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000016", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-harvey). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-harvey", "disaster_type": "flooding", "capture_date": "2017-08-31T17:39:25.085Z", "gsd": 3.1466618, "damage_summary": {"total_buildings": 41, "classified_buildings": 41, "counts": {"no-damage": 41, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 1474194.2331, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 897253, "1": 151323, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-harvey_00000292_0053", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000292", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a flooding event (hurricane-harvey) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 82, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 82, "classified_buildings": 82, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 1182037.7664, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-harvey_00000292_0054", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000292", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "For a post-disaster flooding scene (hurricane-harvey), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 95514.7018, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 46, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 678540.3641, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 95514.7018, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 46, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 678540.3641, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 27, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 407982.7005, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-harvey_00000292_0055", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000292", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 1, "severe_count": 0, "class_histogram": {"no_damage": 1, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1740.1183, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 93774.5835, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 28150.8978, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-harvey_00000292_0056", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000292", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-harvey). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-harvey", "disaster_type": "flooding", "capture_date": "2017-08-31T17:38:50.685Z", "gsd": 3.024613, "damage_summary": {"total_buildings": 82, "classified_buildings": 82, "counts": {"no-damage": 82, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 1182037.7664, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 914070, "1": 134506, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_moore-tornado_00000062_0057", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "moore-tornado_00000062", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a wind event (moore-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 21, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 21, "classified_buildings": 21, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 90305.5114, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_moore-tornado_00000062_0058", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "moore-tornado_00000062", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "For a post-disaster wind scene (moore-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 796.2495, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 36720.9694, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 28154.2064, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 796.2495, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 36720.9694, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 28154.2064, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 24634.0861, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_moore-tornado_00000062_0059", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "moore-tornado_00000062", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 18904.846, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_moore-tornado_00000062_0060", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "moore-tornado_00000062", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (moore-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "moore-tornado", "disaster_type": "wind", "capture_date": "2013-05-22T17:26:30.085Z", "gsd": 1.72984850406647, "damage_summary": {"total_buildings": 21, "classified_buildings": 21, "counts": {"no-damage": 21, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 90305.5114, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-florence_00000297_0061", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000297", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a flooding event (hurricane-florence) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 15, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 15, "classified_buildings": 15, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 40451.1306, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-florence_00000297_0062", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000297", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "For a post-disaster flooding scene (hurricane-florence), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 18948.1147, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 21503.0159, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 18948.1147, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 21503.0159, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-florence_00000297_0063", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000297", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 1, "severe_count": 0, "class_histogram": {"no_damage": 1, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 3814.5123, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-florence_00000297_0064", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000297", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-florence). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-florence", "disaster_type": "flooding", "capture_date": "2018-09-20T16:04:41.000Z", "gsd": 2.0916247, "damage_summary": {"total_buildings": 15, "classified_buildings": 15, "counts": {"no-damage": 15, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 40451.1306, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 1038702, "1": 9874, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_pinery-bushfire_00000897_0065", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000897", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a fire event (pinery-bushfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 16, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 4}, "total_buildings": 20, "classified_buildings": 16, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 52891.8415, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_pinery-bushfire_00000897_0066", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000897", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "For a post-disaster fire scene (pinery-bushfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 19, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 52724.1852, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 167.6563, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 19, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 52724.1852, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 167.6563, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_pinery-bushfire_00000897_0067", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000897", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 7, "severe_count": 0, "class_histogram": {"no_damage": 6, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 1}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6709.815, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1899.4205, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_pinery-bushfire_00000897_0068", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000897", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (pinery-bushfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "pinery-bushfire", "disaster_type": "fire", "capture_date": "2015-11-29T01:09:22.998Z", "gsd": 1.65928483009338, "damage_summary": {"total_buildings": 20, "classified_buildings": 16, "counts": {"no-damage": 16, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 4}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 52891.8415, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_socal-fire_00001341_0069", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "socal-fire_00001341", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a fire event (socal-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 74, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 74, "classified_buildings": 74, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 524780.1939, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_socal-fire_00001341_0070", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "socal-fire_00001341", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "For a post-disaster fire scene (socal-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 14, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 103505.7948, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 108273.0135, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 22, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 145614.5314, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 14, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 103505.7948, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 108273.0135, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 22, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 145614.5314, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 27, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 167386.8542, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_socal-fire_00001341_0071", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "socal-fire_00001341", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 2, "severe_count": 0, "class_histogram": {"no_damage": 2, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 17092.97, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 8610.1597, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_socal-fire_00001341_0072", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "socal-fire_00001341", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (socal-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "socal-fire", "disaster_type": "fire", "capture_date": "2018-11-14T18:42:58.000Z", "gsd": 2.5700748, "damage_summary": {"total_buildings": 74, "classified_buildings": 74, "counts": {"no-damage": 74, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 524780.1939, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 964022, "1": 84554, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00001399_0073", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001399", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 100, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 2}, "total_buildings": 102, "classified_buildings": 100, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 350500.009, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00001399_0074", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001399", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 46, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 145678.1298, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 21, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 56003.4165, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 29, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 79006.8116, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 46, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 145678.1298, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 21, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 56003.4165, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 29, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 79006.8116, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 69811.6511, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00001399_0075", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001399", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 15, "severe_count": 0, "class_histogram": {"no_damage": 15, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 52846.6229, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 19, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 53331.129, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 18777.4498, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00001399_0076", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001399", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 102, "classified_buildings": 100, "counts": {"no-damage": 100, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 2}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 350500.009, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_moore-tornado_00000037_0077", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "moore-tornado_00000037", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a wind event (moore-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 10, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 10, "classified_buildings": 10, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 40764.4436, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_moore-tornado_00000037_0078", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "moore-tornado_00000037", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "For a post-disaster wind scene (moore-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "none", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 563.1137, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 25513.9816, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 563.1137, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 25513.9816, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 14687.3483, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_moore-tornado_00000037_0079", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "moore-tornado_00000037", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 563.1137, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_moore-tornado_00000037_0080", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "moore-tornado_00000037", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (moore-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "moore-tornado", "disaster_type": "wind", "capture_date": "2013-05-22T17:26:30.085Z", "gsd": 1.72984850406647, "damage_summary": {"total_buildings": 10, "classified_buildings": 10, "counts": {"no-damage": 10, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 40764.4436, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_tuscaloosa-tornado_00000326_0081", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000326", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a wind event (tuscaloosa-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 77, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "total_buildings": 78, "classified_buildings": 77, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 287044.307, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_tuscaloosa-tornado_00000326_0082", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000326", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "For a post-disaster wind scene (tuscaloosa-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 21, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 85114.5628, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 42540.4757, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 23, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 87879.8552, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 21, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 85114.5628, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 42540.4757, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 23, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 87879.8552, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 19, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 71509.4133, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_tuscaloosa-tornado_00000326_0083", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000326", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 2, "severe_count": 0, "class_histogram": {"no_damage": 2, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 8705.3342, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 11208.9452, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 17750.9952, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_tuscaloosa-tornado_00000326_0084", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000326", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (tuscaloosa-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "tuscaloosa-tornado", "disaster_type": "wind", "capture_date": "2011-05-19T16:48:01.085Z", "gsd": 1.85690224170685, "damage_summary": {"total_buildings": 78, "classified_buildings": 77, "counts": {"no-damage": 77, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 287044.307, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_midwest-flooding_00000285_0085", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000285", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a flooding event (midwest-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 11, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 4}, "total_buildings": 15, "classified_buildings": 11, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 11056.2728, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_midwest-flooding_00000285_0086", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000285", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "For a post-disaster flooding scene (midwest-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1772.9214, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 9283.3514, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1772.9214, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 9283.3514, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_midwest-flooding_00000285_0087", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000285", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 1, "severe_count": 0, "class_histogram": {"no_damage": 1, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1478.7128, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_midwest-flooding_00000285_0088", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000285", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (midwest-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "midwest-flooding", "disaster_type": "flooding", "capture_date": "2019-05-30T17:35:04.000Z", "gsd": 1.2524601, "damage_summary": {"total_buildings": 15, "classified_buildings": 11, "counts": {"no-damage": 11, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 4}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 11056.2728, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 1041343, "1": 7233, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_lower-puna-volcano_00000226_0097", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000226", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a volcano event (lower-puna-volcano) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 34, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 2}, "total_buildings": 36, "classified_buildings": 34, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 140185.9777, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_lower-puna-volcano_00000226_0098", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000226", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "no-low-damage", "question": "For a post-disaster volcano scene (lower-puna-volcano), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 23469.9365, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 45012.3134, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 22550.4043, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 23469.9365, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 45012.3134, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 22550.4043, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 49153.3235, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_lower-puna-volcano_00000226_0099", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000226", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster volcano satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 4, "severe_count": 0, "class_histogram": {"no_damage": 4, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 15607.42, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1549.6402, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6889.0767, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_lower-puna-volcano_00000226_0100", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000226", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster volcano scene (lower-puna-volcano). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "lower-puna-volcano", "disaster_type": "volcano", "capture_date": "2018-05-23T20:59:21.000Z", "gsd": 2.2423599, "damage_summary": {"total_buildings": 36, "classified_buildings": 34, "counts": {"no-damage": 34, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 2}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 140185.9777, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_pinery-bushfire_00001282_0101", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001282", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a fire event (pinery-bushfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 11, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "total_buildings": 12, "classified_buildings": 11, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 28943.0636, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_pinery-bushfire_00001282_0102", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001282", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "For a post-disaster fire scene (pinery-bushfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "none", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 28943.0636, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 28943.0636, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_pinery-bushfire_00001282_0103", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001282", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_pinery-bushfire_00001282_0104", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001282", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (pinery-bushfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "pinery-bushfire", "disaster_type": "fire", "capture_date": "2015-11-27T00:39:41.429Z", "gsd": 2.03157305717468, "damage_summary": {"total_buildings": 12, "classified_buildings": 11, "counts": {"no-damage": 11, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 28943.0636, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_mexico-earthquake_00000191_0109", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000191", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a earthquake event (mexico-earthquake) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 359, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 359, "classified_buildings": 359, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 1676639.5153, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_mexico-earthquake_00000191_0110", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000191", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "no-low-damage", "question": "For a post-disaster earthquake scene (mexico-earthquake), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 173, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 841286.0286, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 118, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 510523.7784, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 45, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 197161.7248, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 173, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 841286.0286, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 118, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 510523.7784, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 45, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 197161.7248, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 23, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 127667.9835, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_mexico-earthquake_00000191_0111", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000191", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster earthquake satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 37, "severe_count": 0, "class_histogram": {"no_damage": 37, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 37, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 141893.38, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 37, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 172464.3033, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 34, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 180040.9236, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_mexico-earthquake_00000191_0112", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000191", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster earthquake scene (mexico-earthquake). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "mexico-earthquake", "disaster_type": "earthquake", "capture_date": "2017-09-20T17:46:11.000Z", "gsd": 2.6503215, "damage_summary": {"total_buildings": 359, "classified_buildings": 359, "counts": {"no-damage": 359, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 1676639.5153, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 791803, "1": 256773, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_guatemala-volcano_00000023_0113", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "guatemala-volcano_00000023", "split": "train", "qa_split": "train", "disaster": "guatemala-volcano", "disaster_type": "volcano", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a volcano event (guatemala-volcano) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 167, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 167, "classified_buildings": 167, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 106828.9179, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_guatemala-volcano_00000023_0114", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "guatemala-volcano_00000023", "split": "train", "qa_split": "train", "disaster": "guatemala-volcano", "disaster_type": "volcano", "damage_bucket": "no-low-damage", "question": "For a post-disaster volcano scene (guatemala-volcano), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "none", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 96, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 64689.8277, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 96, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 64689.8277, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 71, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 42139.0902, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_guatemala-volcano_00000023_0115", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "guatemala-volcano_00000023", "split": "train", "qa_split": "train", "disaster": "guatemala-volcano", "disaster_type": "volcano", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster volcano satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 14, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 8702.2676, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_guatemala-volcano_00000023_0116", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "guatemala-volcano_00000023", "split": "train", "qa_split": "train", "disaster": "guatemala-volcano", "disaster_type": "volcano", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster volcano scene (guatemala-volcano). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "guatemala-volcano", "disaster_type": "volcano", "capture_date": "2018-06-22T16:55:40.000Z", "gsd": 1.4085245, "damage_summary": {"total_buildings": 167, "classified_buildings": 167, "counts": {"no-damage": 167, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 106828.9179, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 989329, "1": 59247, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00000060_0117", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000060", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 13, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "total_buildings": 14, "classified_buildings": 13, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 49772.4589, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00000060_0118", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000060", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "none", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 13472.0434, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 31833.5085, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 13472.0434, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 31833.5085, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 4466.907, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00000060_0119", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000060", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 514.2801, "severity_score": 0.0, "dominant_damage": "unclassified"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00000060_0120", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000060", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 14, "classified_buildings": 13, "counts": {"no-damage": 13, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 49772.4589, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_moore-tornado_00000098_0125", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "moore-tornado_00000098", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a wind event (moore-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 135, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 135, "classified_buildings": 135, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 372563.0616, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_moore-tornado_00000098_0126", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "moore-tornado_00000098", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "For a post-disaster wind scene (moore-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 40, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 109685.4282, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 50, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 165281.7081, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 39, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 79997.567, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 40, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 109685.4282, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 50, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 165281.7081, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 39, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 79997.567, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 17598.3583, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_moore-tornado_00000098_0127", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "moore-tornado_00000098", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 9, "severe_count": 0, "class_histogram": {"no_damage": 9, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 19298.7698, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 13, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 45628.4403, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 21, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 65283.0351, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_moore-tornado_00000098_0128", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "moore-tornado_00000098", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (moore-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "moore-tornado", "disaster_type": "wind", "capture_date": "2013-05-22T17:26:30.085Z", "gsd": 1.72984850406647, "damage_summary": {"total_buildings": 135, "classified_buildings": 135, "counts": {"no-damage": 135, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 372563.0616, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00000619_0129", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000619", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 43, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 43, "classified_buildings": 43, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 103774.1026, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00000619_0130", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000619", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "none", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 43, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 103774.1026, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 43, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 103774.1026, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00000619_0131", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000619", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00000619_0132", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000619", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 43, "classified_buildings": 43, "counts": {"no-damage": 43, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 103774.1026, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_lower-puna-volcano_00000024_0133", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000024", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a volcano event (lower-puna-volcano) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 20, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 4}, "total_buildings": 24, "classified_buildings": 20, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 27489.6811, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_lower-puna-volcano_00000024_0134", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000024", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "no-low-damage", "question": "For a post-disaster volcano scene (lower-puna-volcano), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 9180.0131, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 14, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 18309.668, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 9180.0131, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 14, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 18309.668, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_lower-puna-volcano_00000024_0135", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000024", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster volcano satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_lower-puna-volcano_00000024_0136", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000024", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster volcano scene (lower-puna-volcano). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "lower-puna-volcano", "disaster_type": "volcano", "capture_date": "2018-05-23T20:59:21.000Z", "gsd": 2.2423599, "damage_summary": {"total_buildings": 24, "classified_buildings": 20, "counts": {"no-damage": 20, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 4}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 27489.6811, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00000535_0137", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000535", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 82, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 3}, "total_buildings": 85, "classified_buildings": 82, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 292726.1414, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00000535_0138", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000535", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 19, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 78976.9132, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 24981.5371, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 52, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 164885.0389, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 19, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 78976.9132, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 24981.5371, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 52, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 164885.0389, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 23882.6522, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00000535_0139", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000535", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 5, "severe_count": 0, "class_histogram": {"no_damage": 5, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 24897.2273, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 38444.7554, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00000535_0140", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000535", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 85, "classified_buildings": 82, "counts": {"no-damage": 82, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 3}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 292726.1414, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-florence_00000226_0149", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000226", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a flooding event (hurricane-florence) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 17, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "total_buildings": 18, "classified_buildings": 17, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 51216.7991, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-florence_00000226_0150", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000226", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "For a post-disaster flooding scene (hurricane-florence), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "none", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1350.3301, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 13451.9512, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1350.3301, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 13451.9512, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 36414.5178, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-florence_00000226_0151", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000226", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-florence_00000226_0152", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000226", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-florence). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-florence", "disaster_type": "flooding", "capture_date": "2018-09-20T16:04:41.000Z", "gsd": 2.0916247, "damage_summary": {"total_buildings": 18, "classified_buildings": 17, "counts": {"no-damage": 17, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 51216.7991, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 1036020, "1": 12556, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00000627_0153", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000627", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 22, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "total_buildings": 23, "classified_buildings": 22, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 33222.5409, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00000627_0154", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000627", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 22, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 32376.8286, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 845.7123, "severity_score": 0.0, "dominant_damage": "unclassified"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 22, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 32376.8286, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 845.7123, "severity_score": 0.0, "dominant_damage": "unclassified"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00000627_0155", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000627", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 20, "severe_count": 0, "class_histogram": {"no_damage": 20, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 20, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 28836.3259, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 821.1735, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00000627_0156", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000627", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 23, "classified_buildings": 22, "counts": {"no-damage": 22, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 33222.5409, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000392_0157", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000392", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 155, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 155, "classified_buildings": 155, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 346246.0038, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000392_0158", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000392", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 649.8755, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 44, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 128158.3081, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6947.6816, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 649.8755, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 44, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 128158.3081, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6947.6816, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 101, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 210490.1386, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000392_0159", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000392", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 649.8755, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 19, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 43808.4571, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000392_0160", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000392", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 155, "classified_buildings": 155, "counts": {"no-damage": 155, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 346246.0038, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_sunda-tsunami_00000119_0165", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000119", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a tsunami event (sunda-tsunami) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 10, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 5}, "total_buildings": 15, "classified_buildings": 10, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 26895.17, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_sunda-tsunami_00000119_0166", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000119", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "no-low-damage", "question": "For a post-disaster tsunami scene (sunda-tsunami), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 7561.7865, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 19333.3835, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 7561.7865, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 19333.3835, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_sunda-tsunami_00000119_0167", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000119", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster tsunami satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_sunda-tsunami_00000119_0168", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000119", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster tsunami scene (sunda-tsunami). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "sunda-tsunami", "disaster_type": "tsunami", "capture_date": "2019-01-02T03:17:07.000Z", "gsd": 1.7746705, "damage_summary": {"total_buildings": 15, "classified_buildings": 10, "counts": {"no-damage": 10, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 5}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 26895.17, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000527_0169", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000527", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 230, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 230, "classified_buildings": 230, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 925819.0688, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000527_0170", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000527", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 28, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 140339.9974, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 37, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 121659.5121, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 61, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 226580.2357, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 28, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 140339.9974, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 37, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 121659.5121, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 61, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 226580.2357, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 104, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 437239.3236, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000527_0171", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000527", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 1, "severe_count": 0, "class_histogram": {"no_damage": 1, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6343.8971, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 31908.0916, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000527_0172", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000527", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 230, "classified_buildings": 230, "counts": {"no-damage": 230, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 925819.0688, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 819626, "1": 228950, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_socal-fire_00000046_0173", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "socal-fire_00000046", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a fire event (socal-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 8, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 16}, "total_buildings": 24, "classified_buildings": 8, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 56360.3587, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_socal-fire_00000046_0174", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "socal-fire_00000046", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "For a post-disaster fire scene (socal-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 12542.2511, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 17635.8811, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 25164.5885, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 12542.2511, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 17635.8811, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 25164.5885, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1017.638, "severity_score": 0.0, "dominant_damage": "unclassified"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_socal-fire_00000046_0175", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "socal-fire_00000046", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_socal-fire_00000046_0176", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "socal-fire_00000046", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (socal-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "socal-fire", "disaster_type": "fire", "capture_date": "2018-11-14T18:42:58.000Z", "gsd": 2.5700748, "damage_summary": {"total_buildings": 24, "classified_buildings": 8, "counts": {"no-damage": 8, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 16}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 56360.3587, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 1042191, "1": 6385, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_palu-tsunami_00000128_0177", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000128", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a tsunami event (palu-tsunami) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 178, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 178, "classified_buildings": 178, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 640093.8642, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_palu-tsunami_00000128_0178", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000128", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "no-low-damage", "question": "For a post-disaster tsunami scene (palu-tsunami), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 61, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 267104.4728, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 37, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 118155.0824, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 27, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 93997.6423, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 61, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 267104.4728, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 37, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 118155.0824, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 27, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 93997.6423, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 53, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 160836.6667, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_palu-tsunami_00000128_0179", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000128", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster tsunami satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 2, "severe_count": 0, "class_histogram": {"no_damage": 2, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 14679.8466, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 27, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 131236.9025, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 14, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 35156.8604, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_palu-tsunami_00000128_0180", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000128", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster tsunami scene (palu-tsunami). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "palu-tsunami", "disaster_type": "tsunami", "capture_date": "2018-10-01T02:26:02.000Z", "gsd": 2.8247435, "damage_summary": {"total_buildings": 178, "classified_buildings": 178, "counts": {"no-damage": 178, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 640093.8642, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 961568, "1": 87008, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_woolsey-fire_00000482_0181", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000482", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a fire event (woolsey-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 22, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 22, "classified_buildings": 22, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 63021.2463, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_woolsey-fire_00000482_0182", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000482", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "For a post-disaster fire scene (woolsey-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "none", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 22273.6067, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 22273.6067, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 14, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 40747.6396, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_woolsey-fire_00000482_0183", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000482", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_woolsey-fire_00000482_0184", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000482", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (woolsey-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "woolsey-fire", "disaster_type": "fire", "capture_date": "2018-11-18T18:44:07.000Z", "gsd": 1.7726423000000002, "damage_summary": {"total_buildings": 22, "classified_buildings": 22, "counts": {"no-damage": 22, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 63021.2463, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_santa-rosa-wildfire_00000166_0185", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000166", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a fire event (santa-rosa-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 208, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 208, "classified_buildings": 208, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 705933.8809, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_santa-rosa-wildfire_00000166_0186", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000166", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "For a post-disaster fire scene (santa-rosa-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 57, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 180146.8703, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 80, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 289724.6929, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 57, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 180146.8703, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 80, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 289724.6929, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 71, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 236062.3177, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_santa-rosa-wildfire_00000166_0187", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000166", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 23, "severe_count": 0, "class_histogram": {"no_damage": 23, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 23, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 68833.5298, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 18283.8435, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_santa-rosa-wildfire_00000166_0188", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000166", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (santa-rosa-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "santa-rosa-wildfire", "disaster_type": "fire", "capture_date": "2017-10-11T19:19:41.000Z", "gsd": 1.8769937, "damage_summary": {"total_buildings": 208, "classified_buildings": 208, "counts": {"no-damage": 208, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 705933.8809, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 835953, "1": 212623, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-harvey_00000366_0189", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000366", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a flooding event (hurricane-harvey) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 156, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 156, "classified_buildings": 156, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 2420761.2124, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-harvey_00000366_0190", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000366", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "For a post-disaster flooding scene (hurricane-harvey), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 54, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 695180.4874, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 55, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 663738.9433, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 20, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 571380.1033, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 54, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 695180.4874, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 55, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 663738.9433, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 20, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 571380.1033, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 27, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 490461.6784, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-harvey_00000366_0191", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000366", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 15, "severe_count": 0, "class_histogram": {"no_damage": 15, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 143737.3385, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 16, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 149358.3388, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 13, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 169889.5495, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-harvey_00000366_0192", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000366", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-harvey). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-harvey", "disaster_type": "flooding", "capture_date": "2017-08-31T17:38:50.685Z", "gsd": 3.024613, "damage_summary": {"total_buildings": 156, "classified_buildings": 156, "counts": {"no-damage": 156, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 2420761.2124, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 770493, "1": 278083, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00000791_0193", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000791", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 10, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 10, "classified_buildings": 10, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 33690.3389, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00000791_0194", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000791", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 28624.4625, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 5065.8764, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 28624.4625, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 5065.8764, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00000791_0195", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000791", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 0, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "none", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 10321.3154, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00000791_0196", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000791", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 10, "classified_buildings": 10, "counts": {"no-damage": 10, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 33690.3389, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_mexico-earthquake_00000043_0197", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000043", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "no-low-damage", "question": "A post-disaster satellite scene from a earthquake event (mexico-earthquake) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 839, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "total_buildings": 840, "classified_buildings": 839, "damaged_count": 0, "severe_count": 0, "damaged_pct": 0.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 3407248.8045, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_mexico-earthquake_00000043_0198", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000043", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "no-low-damage", "question": "For a post-disaster earthquake scene (mexico-earthquake), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 190, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 824747.2998, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 211, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 895243.8944, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 241, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 860471.0306, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 190, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 824747.2998, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 211, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 895243.8944, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 241, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 860471.0306, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 198, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 826786.5797, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_mexico-earthquake_00000043_0199", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000043", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "no-low-damage", "question": "A 4\u00d74 grid has been overlaid on a post-disaster earthquake satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0, "building_count": 55, "severe_count": 0, "class_histogram": {"no_damage": 55, "minor_damage": 0, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 55, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 207715.2297, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 44, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 192434.9845, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 49, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 224350.3383, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_mexico-earthquake_00000043_0200", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000043", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "no-low-damage", "question": "Create a comprehensive scene-level damage summary for this post-disaster earthquake scene (mexico-earthquake). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "mexico-earthquake", "disaster_type": "earthquake", "capture_date": "2017-09-20T17:46:11.000Z", "gsd": 2.6503215, "damage_summary": {"total_buildings": 840, "classified_buildings": 839, "counts": {"no-damage": 839, "minor-damage": 0, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 0.0}, "area_summary": {"total_footprint_m2": 3407248.8045, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 526098, "1": 522478, "2": 0, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 100.0, "2": 0.0, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_joplin-tornado_00000042_0201", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000042", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a wind event (joplin-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 139, "minor-damage": 5, "major-damage": 0, "destroyed": 0, "unclassified": 6}, "total_buildings": 150, "classified_buildings": 144, "damaged_count": 5, "severe_count": 0, "damaged_pct": 3.47, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 670922.4591, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_joplin-tornado_00000042_0202", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000042", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "For a post-disaster wind scene (joplin-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 37, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 144897.4403, "severity_score": 0.0089, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 35, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 127430.691, "severity_score": 0.0189, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 46, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 233053.6407, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 37, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 144897.4403, "severity_score": 0.0089, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 35, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 127430.691, "severity_score": 0.0189, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 46, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 233053.6407, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 32, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 165540.6871, "severity_score": 0.0206, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_joplin-tornado_00000042_0203", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000042", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G22", "severity_score": 0.0388, "building_count": 17, "severe_count": 0, "class_histogram": {"no_damage": 15, "minor_damage": 2, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G22", "row": 2, "col": 2, "building_count": 17, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 70564.3902, "severity_score": 0.0388, "dominant_damage": "no-damage"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 29987.226, "severity_score": 0.0367, "dominant_damage": "no-damage"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 13, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 74333.4766, "severity_score": 0.0254, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_joplin-tornado_00000042_0204", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000042", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (joplin-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "joplin-tornado", "disaster_type": "wind", "capture_date": "2011-05-29T17:29:25.433Z", "gsd": 2.35255861282349, "damage_summary": {"total_buildings": 150, "classified_buildings": 144, "counts": {"no-damage": 139, "minor-damage": 5, "major-damage": 0, "destroyed": 0, "unclassified": 6}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 3.47}, "area_summary": {"total_footprint_m2": 670922.4591, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_lower-puna-volcano_00000059_0209", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000059", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a volcano event (lower-puna-volcano) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 25, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 3}, "total_buildings": 29, "classified_buildings": 26, "damaged_count": 1, "severe_count": 0, "damaged_pct": 3.85, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 32027.7292, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_lower-puna-volcano_00000059_0210", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000059", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "minor-moderate", "question": "For a post-disaster volcano scene (lower-puna-volcano), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6912.139, "severity_score": 0.055, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 23, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 25115.5902, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6912.139, "severity_score": 0.055, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 23, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 25115.5902, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_lower-puna-volcano_00000059_0211", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000059", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster volcano satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G10", "severity_score": 0.33, "building_count": 1, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G10", "row": 1, "col": 0, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1349.6198, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 549.0651, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 4378.133, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_lower-puna-volcano_00000059_0212", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000059", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster volcano scene (lower-puna-volcano). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "lower-puna-volcano", "disaster_type": "volcano", "capture_date": "2018-05-23T20:59:21.000Z", "gsd": 2.2423599, "damage_summary": {"total_buildings": 29, "classified_buildings": 26, "counts": {"no-damage": 25, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 3}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 3.85}, "area_summary": {"total_footprint_m2": 32027.7292, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000000_0213", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000000", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 43, "minor-damage": 10, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "total_buildings": 54, "classified_buildings": 53, "damaged_count": 10, "severe_count": 0, "damaged_pct": 18.87, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 255616.4127, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000000_0214", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000000", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 18, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 78821.7875, "severity_score": 0.0733, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 18, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 81773.1973, "severity_score": 0.055, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 18, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 78821.7875, "severity_score": 0.0733, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 18, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 81773.1973, "severity_score": 0.055, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 18, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 95021.4279, "severity_score": 0.055, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000000_0215", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000000", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G10", "severity_score": 0.11, "building_count": 12, "severe_count": 0, "class_histogram": {"no_damage": 8, "minor_damage": 4, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G10", "row": 1, "col": 0, "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 50792.3235, "severity_score": 0.11, "dominant_damage": "no-damage"}, {"cell_id": "G21", "row": 2, "col": 1, "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 28693.9708, "severity_score": 0.11, "dominant_damage": "no-damage"}, {"cell_id": "G32", "row": 3, "col": 2, "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 65209.5507, "severity_score": 0.09, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000000_0216", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000000", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 54, "classified_buildings": 53, "counts": {"no-damage": 43, "minor-damage": 10, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 18.87}, "area_summary": {"total_footprint_m2": 255616.4127, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 985485, "1": 49292, "2": 13799, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 78.1284, "2": 21.8716, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_tuscaloosa-tornado_00000307_0221", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000307", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a wind event (tuscaloosa-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 32, "minor-damage": 2, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "total_buildings": 35, "classified_buildings": 34, "damaged_count": 2, "severe_count": 0, "damaged_pct": 5.88, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 186487.5602, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_tuscaloosa-tornado_00000307_0222", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000307", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "For a post-disaster wind scene (tuscaloosa-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 83227.9313, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 31146.4226, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1273.9482, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 83227.9313, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 31146.4226, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1273.9482, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 28, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 70839.2581, "severity_score": 0.0236, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_tuscaloosa-tornado_00000307_0223", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000307", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G32", "severity_score": 0.055, "building_count": 6, "severe_count": 0, "class_histogram": {"no_damage": 5, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G32", "row": 3, "col": 2, "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 14913.1279, "severity_score": 0.055, "dominant_damage": "no-damage"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 20933.2119, "severity_score": 0.0413, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_tuscaloosa-tornado_00000307_0224", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000307", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (tuscaloosa-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "tuscaloosa-tornado", "disaster_type": "wind", "capture_date": "2011-05-19T16:48:01.085Z", "gsd": 1.85690224170685, "damage_summary": {"total_buildings": 35, "classified_buildings": 34, "counts": {"no-damage": 32, "minor-damage": 2, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 5.88}, "area_summary": {"total_footprint_m2": 186487.5602, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000267_0225", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000267", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 176, "minor-damage": 4, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 180, "classified_buildings": 180, "damaged_count": 4, "severe_count": 0, "damaged_pct": 2.22, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 691032.8487, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000267_0226", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000267", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "none", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 43262.3704, "severity_score": 0.132, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 35, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 131586.9346, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 43262.3704, "severity_score": 0.132, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 35, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 131586.9346, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 135, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 516183.5437, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000267_0227", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000267", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G12", "severity_score": 0.264, "building_count": 5, "severe_count": 0, "class_histogram": {"no_damage": 1, "minor_damage": 4, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G12", "row": 1, "col": 2, "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 12710.6606, "severity_score": 0.264, "dominant_damage": "minor-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000267_0228", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000267", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:17:02.085Z", "gsd": 2.9729018, "damage_summary": {"total_buildings": 180, "classified_buildings": 180, "counts": {"no-damage": 176, "minor-damage": 4, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 2.22}, "area_summary": {"total_footprint_m2": 691032.8487, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000411_0233", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000411", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 139, "minor-damage": 2, "major-damage": 0, "destroyed": 0, "unclassified": 2}, "total_buildings": 143, "classified_buildings": 141, "damaged_count": 2, "severe_count": 0, "damaged_pct": 1.42, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 300639.3247, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000411_0234", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000411", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 26, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 31329.2164, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 28701.73, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 16, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 30921.3066, "severity_score": 0.0413, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 26, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 31329.2164, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 28701.73, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 16, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 30921.3066, "severity_score": 0.0413, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 90, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 209687.0717, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000411_0235", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000411", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G20", "severity_score": 0.066, "building_count": 10, "severe_count": 0, "class_histogram": {"no_damage": 8, "minor_damage": 2, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G20", "row": 2, "col": 0, "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 16468.8474, "severity_score": 0.066, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1935.3735, "severity_score": 0.0, "dominant_damage": "unclassified"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000411_0236", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000411", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 143, "classified_buildings": 141, "counts": {"no-damage": 139, "minor-damage": 2, "major-damage": 0, "destroyed": 0, "unclassified": 2}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 1.42}, "area_summary": {"total_footprint_m2": 300639.3247, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-matthew_00000083_0237", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000083", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a wind event (hurricane-matthew) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 59, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 11}, "total_buildings": 71, "classified_buildings": 60, "damaged_count": 1, "severe_count": 0, "damaged_pct": 1.67, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 214885.0568, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-matthew_00000083_0238", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000083", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "For a post-disaster wind scene (hurricane-matthew), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 33412.9347, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 29, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 91164.9567, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 14, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 42897.8403, "severity_score": 0.0236, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 33412.9347, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 29, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 91164.9567, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 14, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 42897.8403, "severity_score": 0.0236, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 13, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 47409.3251, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-matthew_00000083_0239", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000083", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G21", "severity_score": 0.0471, "building_count": 7, "severe_count": 0, "class_histogram": {"no_damage": 6, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G21", "row": 2, "col": 1, "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 13705.923, "severity_score": 0.0471, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 18848.1169, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-matthew_00000083_0240", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000083", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-matthew). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-matthew", "disaster_type": "wind", "capture_date": "2016-10-01T15:29:45.000Z", "gsd": 2.37887692, "damage_summary": {"total_buildings": 71, "classified_buildings": 60, "counts": {"no-damage": 59, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 11}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 1.67}, "area_summary": {"total_footprint_m2": 214885.0568, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 1012715, "1": 34995, "2": 866, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 97.5851, "2": 2.4149, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_tuscaloosa-tornado_00000031_0241", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000031", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a wind event (tuscaloosa-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 130, "minor-damage": 6, "major-damage": 0, "destroyed": 0, "unclassified": 4}, "total_buildings": 140, "classified_buildings": 136, "damaged_count": 6, "severe_count": 0, "damaged_pct": 4.41, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 447881.8975, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_tuscaloosa-tornado_00000031_0242", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000031", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "For a post-disaster wind scene (tuscaloosa-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 30, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 93163.4395, "severity_score": 0.033, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 38, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 122945.7416, "severity_score": 0.0087, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 31, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 92153.5573, "severity_score": 0.0213, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 30, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 93163.4395, "severity_score": 0.033, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 38, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 122945.7416, "severity_score": 0.0087, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 31, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 92153.5573, "severity_score": 0.0213, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 41, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 139619.1591, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_tuscaloosa-tornado_00000031_0243", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000031", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0733, "building_count": 9, "severe_count": 0, "class_histogram": {"no_damage": 7, "minor_damage": 2, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 24847.7271, "severity_score": 0.0733, "dominant_damage": "no-damage"}, {"cell_id": "G30", "row": 3, "col": 0, "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 22172.8387, "severity_score": 0.0471, "dominant_damage": "no-damage"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 37514.4336, "severity_score": 0.03, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_tuscaloosa-tornado_00000031_0244", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000031", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (tuscaloosa-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "tuscaloosa-tornado", "disaster_type": "wind", "capture_date": "2011-05-19T16:48:01.085Z", "gsd": 1.85690224170685, "damage_summary": {"total_buildings": 140, "classified_buildings": 136, "counts": {"no-damage": 130, "minor-damage": 6, "major-damage": 0, "destroyed": 0, "unclassified": 4}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 4.41}, "area_summary": {"total_footprint_m2": 447881.8975, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_moore-tornado_00000013_0249", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "moore-tornado_00000013", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a wind event (moore-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 159, "minor-damage": 4, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 163, "classified_buildings": 163, "damaged_count": 4, "severe_count": 0, "damaged_pct": 2.45, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 431584.0903, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_moore-tornado_00000013_0250", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "moore-tornado_00000013", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "For a post-disaster wind scene (moore-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 38115.3392, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 66, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 147488.8838, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 66709.6071, "severity_score": 0.022, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 38115.3392, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 66, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 147488.8838, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 66709.6071, "severity_score": 0.022, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 67, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 179270.2602, "severity_score": 0.0148, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_moore-tornado_00000013_0251", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "moore-tornado_00000013", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G30", "severity_score": 0.066, "building_count": 5, "severe_count": 0, "class_histogram": {"no_damage": 4, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G30", "row": 3, "col": 0, "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 35914.5224, "severity_score": 0.066, "dominant_damage": "no-damage"}, {"cell_id": "G32", "row": 3, "col": 2, "building_count": 24, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 59918.335, "severity_score": 0.0275, "dominant_damage": "no-damage"}, {"cell_id": "G33", "row": 3, "col": 3, "building_count": 17, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 39513.6944, "severity_score": 0.0194, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_moore-tornado_00000013_0252", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "moore-tornado_00000013", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (moore-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "moore-tornado", "disaster_type": "wind", "capture_date": "2013-05-22T17:26:30.085Z", "gsd": 1.72984850406647, "damage_summary": {"total_buildings": 163, "classified_buildings": 163, "counts": {"no-damage": 159, "minor-damage": 4, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 2.45}, "area_summary": {"total_footprint_m2": 431584.0903, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_moore-tornado_00000023_0253", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "moore-tornado_00000023", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a wind event (moore-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 279, "minor-damage": 3, "major-damage": 0, "destroyed": 0, "unclassified": 8}, "total_buildings": 290, "classified_buildings": 282, "damaged_count": 3, "severe_count": 0, "damaged_pct": 1.06, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 800383.8553, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_moore-tornado_00000023_0254", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "moore-tornado_00000023", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "For a post-disaster wind scene (moore-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 83, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 211312.9455, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 90, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 240752.6727, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 59, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 174790.5194, "severity_score": 0.0056, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 83, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 211312.9455, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 90, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 240752.6727, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 59, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 174790.5194, "severity_score": 0.0056, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 58, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 173527.7177, "severity_score": 0.0114, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_moore-tornado_00000023_0255", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "moore-tornado_00000023", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G23", "severity_score": 0.0264, "building_count": 25, "severe_count": 0, "class_histogram": {"no_damage": 23, "minor_damage": 2, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G23", "row": 2, "col": 3, "building_count": 25, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 65207.132, "severity_score": 0.0264, "dominant_damage": "no-damage"}, {"cell_id": "G21", "row": 2, "col": 1, "building_count": 25, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 74252.125, "severity_score": 0.0132, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 19, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 37561.5076, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_moore-tornado_00000023_0256", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "moore-tornado_00000023", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (moore-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "moore-tornado", "disaster_type": "wind", "capture_date": "2013-05-22T17:26:30.085Z", "gsd": 1.72984850406647, "damage_summary": {"total_buildings": 290, "classified_buildings": 282, "counts": {"no-damage": 279, "minor-damage": 3, "major-damage": 0, "destroyed": 0, "unclassified": 8}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 1.06}, "area_summary": {"total_footprint_m2": 800383.8553, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_tuscaloosa-tornado_00000094_0273", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000094", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a wind event (tuscaloosa-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 38, "minor-damage": 2, "major-damage": 0, "destroyed": 0, "unclassified": 5}, "total_buildings": 45, "classified_buildings": 40, "damaged_count": 2, "severe_count": 0, "damaged_pct": 5.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 117262.929, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_tuscaloosa-tornado_00000094_0274", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000094", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "For a post-disaster wind scene (tuscaloosa-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 19, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 46771.0828, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 26, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 70491.8462, "severity_score": 0.0254, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 19, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 46771.0828, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 26, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 70491.8462, "severity_score": 0.0254, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_tuscaloosa-tornado_00000094_0275", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000094", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G20", "severity_score": 0.0825, "building_count": 4, "severe_count": 0, "class_histogram": {"no_damage": 3, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G20", "row": 2, "col": 0, "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 11527.5957, "severity_score": 0.0825, "dominant_damage": "no-damage"}, {"cell_id": "G30", "row": 3, "col": 0, "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 23822.6932, "severity_score": 0.0367, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 22042.4682, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_tuscaloosa-tornado_00000094_0276", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000094", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (tuscaloosa-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "tuscaloosa-tornado", "disaster_type": "wind", "capture_date": "2011-05-19T16:48:01.085Z", "gsd": 1.85690224170685, "damage_summary": {"total_buildings": 45, "classified_buildings": 40, "counts": {"no-damage": 38, "minor-damage": 2, "major-damage": 0, "destroyed": 0, "unclassified": 5}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 5.0}, "area_summary": {"total_footprint_m2": 117262.929, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00000039_0281", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000039", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 46, "minor-damage": 2, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "total_buildings": 49, "classified_buildings": 48, "damaged_count": 2, "severe_count": 0, "damaged_pct": 4.17, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 326261.6312, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00000039_0282", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000039", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "minor-moderate", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 27, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 81168.065, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 189511.9171, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 22981.3065, "severity_score": 0.0943, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 27, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 81168.065, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 189511.9171, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 22981.3065, "severity_score": 0.0943, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 32600.3426, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00000039_0283", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000039", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G21", "severity_score": 0.165, "building_count": 4, "severe_count": 0, "class_histogram": {"no_damage": 2, "minor_damage": 2, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G21", "row": 2, "col": 1, "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 17534.3086, "severity_score": 0.165, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 23, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 63752.3709, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00000039_0284", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000039", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 49, "classified_buildings": 48, "counts": {"no-damage": 46, "minor-damage": 2, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 4.17}, "area_summary": {"total_footprint_m2": 326261.6312, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00000028_0285", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000028", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 25, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "total_buildings": 27, "classified_buildings": 26, "damaged_count": 1, "severe_count": 0, "damaged_pct": 3.85, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 90640.6807, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00000028_0286", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000028", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "minor-moderate", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 17992.9767, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 47455.442, "severity_score": 0.0275, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 17992.9767, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 47455.442, "severity_score": 0.0275, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 25192.262, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00000028_0287", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000028", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G20", "severity_score": 0.33, "building_count": 1, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G20", "row": 2, "col": 0, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6503.1372, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00000028_0288", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000028", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 27, "classified_buildings": 26, "counts": {"no-damage": 25, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 3.85}, "area_summary": {"total_footprint_m2": 90640.6807, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_woolsey-fire_00000704_0289", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000704", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a fire event (woolsey-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 9, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 10, "classified_buildings": 10, "damaged_count": 1, "severe_count": 0, "damaged_pct": 10.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 15523.5438, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_woolsey-fire_00000704_0290", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000704", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "minor-moderate", "question": "For a post-disaster fire scene (woolsey-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6647.4032, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1942.729, "severity_score": 0.11, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6933.4116, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6647.4032, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1942.729, "severity_score": 0.11, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6933.4116, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_woolsey-fire_00000704_0291", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000704", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 0.11, "building_count": 3, "severe_count": 0, "class_histogram": {"no_damage": 2, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1942.729, "severity_score": 0.11, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_woolsey-fire_00000704_0292", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000704", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (woolsey-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "woolsey-fire", "disaster_type": "fire", "capture_date": "2018-11-18T18:44:07.000Z", "gsd": 1.7726423000000002, "damage_summary": {"total_buildings": 10, "classified_buildings": 10, "counts": {"no-damage": 9, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 10.0}, "area_summary": {"total_footprint_m2": 15523.5438, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000501_0293", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000501", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 40, "minor-damage": 67, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 107, "classified_buildings": 107, "damaged_count": 67, "severe_count": 0, "damaged_pct": 62.62, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 252852.4725, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000501_0294", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000501", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "none", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 60, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 148271.8284, "severity_score": 0.209, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 60, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 148271.8284, "severity_score": 0.209, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 47, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 104580.6441, "severity_score": 0.2036, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000501_0295", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000501", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G31", "severity_score": 0.255, "building_count": 22, "severe_count": 0, "class_histogram": {"no_damage": 5, "minor_damage": 17, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G31", "row": 3, "col": 1, "building_count": 22, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 50560.9642, "severity_score": 0.255, "dominant_damage": "minor-damage"}, {"cell_id": "G32", "row": 3, "col": 2, "building_count": 25, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 43256.4911, "severity_score": 0.2508, "dominant_damage": "minor-damage"}, {"cell_id": "G30", "row": 3, "col": 0, "building_count": 22, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 54530.1757, "severity_score": 0.24, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000501_0296", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000501", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 107, "classified_buildings": 107, "counts": {"no-damage": 40, "minor-damage": 67, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 62.62}, "area_summary": {"total_footprint_m2": 252852.4725, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 985099, "1": 29424, "2": 34053, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 46.3538, "2": 53.6462, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000217_0301", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000217", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 67, "minor-damage": 12, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 79, "classified_buildings": 79, "damaged_count": 12, "severe_count": 0, "damaged_pct": 15.19, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 588862.848, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000217_0302", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000217", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 75896.0561, "severity_score": 0.06, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 30, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 222171.4454, "severity_score": 0.022, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 19, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 144126.9836, "severity_score": 0.0695, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 75896.0561, "severity_score": 0.06, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 30, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 222171.4454, "severity_score": 0.022, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 19, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 144126.9836, "severity_score": 0.0695, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 19, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 146668.3629, "severity_score": 0.0695, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000217_0303", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000217", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G30", "severity_score": 0.22, "building_count": 6, "severe_count": 0, "class_histogram": {"no_damage": 2, "minor_damage": 4, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G30", "row": 3, "col": 0, "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 37345.2531, "severity_score": 0.22, "dominant_damage": "minor-damage"}, {"cell_id": "G33", "row": 3, "col": 3, "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 43427.5933, "severity_score": 0.132, "dominant_damage": "no-damage"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 21718.2804, "severity_score": 0.11, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000217_0304", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000217", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 79, "classified_buildings": 79, "counts": {"no-damage": 67, "minor-damage": 12, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 15.19}, "area_summary": {"total_footprint_m2": 588862.848, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 905322, "1": 122330, "2": 20924, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 85.3938, "2": 14.6062, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_joplin-tornado_00000132_0309", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000132", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a wind event (joplin-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 161, "minor-damage": 2, "major-damage": 0, "destroyed": 0, "unclassified": 11}, "total_buildings": 174, "classified_buildings": 163, "damaged_count": 2, "severe_count": 0, "damaged_pct": 1.23, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 634653.1943, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_joplin-tornado_00000132_0310", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000132", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "For a post-disaster wind scene (joplin-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 56, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 175600.8598, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 16496.3429, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 85, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 267122.8449, "severity_score": 0.0078, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 56, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 175600.8598, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 16496.3429, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 85, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 267122.8449, "severity_score": 0.0078, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 29, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 175433.1467, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_joplin-tornado_00000132_0311", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000132", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G30", "severity_score": 0.0213, "building_count": 31, "severe_count": 0, "class_histogram": {"no_damage": 27, "minor_damage": 2, "major_damage": 0, "destroyed": 0, "unclassified": 2}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G30", "row": 3, "col": 0, "building_count": 31, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 77176.9635, "severity_score": 0.0213, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 26072.9173, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 48398.4563, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_joplin-tornado_00000132_0312", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000132", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (joplin-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "joplin-tornado", "disaster_type": "wind", "capture_date": "2011-05-29T17:29:25.433Z", "gsd": 2.35255861282349, "damage_summary": {"total_buildings": 174, "classified_buildings": 163, "counts": {"no-damage": 161, "minor-damage": 2, "major-damage": 0, "destroyed": 0, "unclassified": 11}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 1.23}, "area_summary": {"total_footprint_m2": 634653.1943, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000282_0313", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000282", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 62, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 63, "classified_buildings": 63, "damaged_count": 1, "severe_count": 0, "damaged_pct": 1.59, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 216289.0807, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000282_0314", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000282", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 16408.9525, "severity_score": 0.11, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 50, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 175193.1975, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 546.9013, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 16408.9525, "severity_score": 0.11, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 50, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 175193.1975, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 546.9013, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 24140.0294, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000282_0315", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000282", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.33, "building_count": 1, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 10220.2291, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 9420.6756, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000282_0316", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000282", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:17:02.085Z", "gsd": 2.9729018, "damage_summary": {"total_buildings": 63, "classified_buildings": 63, "counts": {"no-damage": 62, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 1.59}, "area_summary": {"total_footprint_m2": 216289.0807, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_tuscaloosa-tornado_00000250_0329", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000250", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a wind event (tuscaloosa-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 100, "minor-damage": 8, "major-damage": 0, "destroyed": 0, "unclassified": 5}, "total_buildings": 113, "classified_buildings": 108, "damaged_count": 8, "severe_count": 0, "damaged_pct": 7.41, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 320560.0288, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_tuscaloosa-tornado_00000250_0330", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000250", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "For a post-disaster wind scene (tuscaloosa-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 62309.6347, "severity_score": 0.022, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 33, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 82874.7777, "severity_score": 0.01, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 23, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 58107.3762, "severity_score": 0.0143, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 62309.6347, "severity_score": 0.022, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 33, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 82874.7777, "severity_score": 0.01, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 23, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 58107.3762, "severity_score": 0.0143, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 42, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 117268.2402, "severity_score": 0.0393, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_tuscaloosa-tornado_00000250_0331", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000250", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G32", "severity_score": 0.09, "building_count": 11, "severe_count": 0, "class_histogram": {"no_damage": 8, "minor_damage": 3, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G32", "row": 3, "col": 2, "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 29431.7264, "severity_score": 0.09, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 8882.7941, "severity_score": 0.0825, "dominant_damage": "no-damage"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 11254.9901, "severity_score": 0.066, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_tuscaloosa-tornado_00000250_0332", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000250", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (tuscaloosa-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "tuscaloosa-tornado", "disaster_type": "wind", "capture_date": "2011-05-19T16:48:01.085Z", "gsd": 1.85690224170685, "damage_summary": {"total_buildings": 113, "classified_buildings": 108, "counts": {"no-damage": 100, "minor-damage": 8, "major-damage": 0, "destroyed": 0, "unclassified": 5}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 7.41}, "area_summary": {"total_footprint_m2": 320560.0288, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000002_0333", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000002", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 53, "minor-damage": 8, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "total_buildings": 62, "classified_buildings": 61, "damaged_count": 8, "severe_count": 0, "damaged_pct": 13.11, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 115423.376, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000002_0334", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000002", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 41, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 72812.9349, "severity_score": 0.0322, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 18125.7262, "severity_score": 0.099, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 5453.3489, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 41, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 72812.9349, "severity_score": 0.0322, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 18125.7262, "severity_score": 0.099, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 5453.3489, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 19031.366, "severity_score": 0.0413, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000002_0335", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000002", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 0.099, "building_count": 10, "severe_count": 0, "class_histogram": {"no_damage": 7, "minor_damage": 3, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 16584.7226, "severity_score": 0.099, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 18125.7262, "severity_score": 0.099, "dominant_damage": "no-damage"}, {"cell_id": "G22", "row": 2, "col": 2, "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 19031.366, "severity_score": 0.0413, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000002_0336", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000002", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 62, "classified_buildings": 61, "counts": {"no-damage": 53, "minor-damage": 8, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 13.11}, "area_summary": {"total_footprint_m2": 115423.376, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_joplin-tornado_00000054_0337", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000054", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a wind event (joplin-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 21, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "total_buildings": 23, "classified_buildings": 22, "damaged_count": 1, "severe_count": 0, "damaged_pct": 4.55, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 473322.911, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_joplin-tornado_00000054_0338", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000054", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "For a post-disaster wind scene (joplin-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 167153.6807, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 165172.6368, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 105996.5854, "severity_score": 0.0367, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 167153.6807, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 165172.6368, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 105996.5854, "severity_score": 0.0367, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 35000.0081, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_joplin-tornado_00000054_0339", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000054", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G30", "severity_score": 0.055, "building_count": 6, "severe_count": 0, "class_histogram": {"no_damage": 5, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G30", "row": 3, "col": 0, "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 51968.3886, "severity_score": 0.055, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 56800.3925, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 95317.1855, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_joplin-tornado_00000054_0340", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000054", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (joplin-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "joplin-tornado", "disaster_type": "wind", "capture_date": "2011-05-29T17:29:25.433Z", "gsd": 2.35255861282349, "damage_summary": {"total_buildings": 23, "classified_buildings": 22, "counts": {"no-damage": 21, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 1}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 4.55}, "area_summary": {"total_footprint_m2": 473322.911, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000608_0341", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000608", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 460, "minor-damage": 7, "major-damage": 0, "destroyed": 0, "unclassified": 8}, "total_buildings": 475, "classified_buildings": 467, "damaged_count": 7, "severe_count": 0, "damaged_pct": 1.5, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 1733807.9851, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000608_0342", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000608", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 112, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 384958.4101, "severity_score": 0.0059, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 116, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 379768.0139, "severity_score": 0.0142, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 117, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 452874.4067, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 112, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 384958.4101, "severity_score": 0.0059, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 116, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 379768.0139, "severity_score": 0.0142, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 117, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 452874.4067, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 130, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 516207.1544, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000608_0343", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000608", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 0.0516, "building_count": 32, "severe_count": 0, "class_histogram": {"no_damage": 25, "minor_damage": 5, "major_damage": 0, "destroyed": 0, "unclassified": 2}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 32, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 99536.2016, "severity_score": 0.0516, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 35, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 108997.2677, "severity_score": 0.0189, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 27, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 96213.9502, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000608_0344", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000608", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 475, "classified_buildings": 467, "counts": {"no-damage": 460, "minor-damage": 7, "major-damage": 0, "destroyed": 0, "unclassified": 8}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 1.5}, "area_summary": {"total_footprint_m2": 1733807.9851, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_mexico-earthquake_00000119_0349", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000119", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a earthquake event (mexico-earthquake) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 53, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 54, "classified_buildings": 54, "damaged_count": 1, "severe_count": 0, "damaged_pct": 1.85, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 78639.1776, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_mexico-earthquake_00000119_0350", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000119", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "minor-moderate", "question": "For a post-disaster earthquake scene (mexico-earthquake), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 54, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 78639.1776, "severity_score": 0.0061, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 54, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 78639.1776, "severity_score": 0.0061, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_mexico-earthquake_00000119_0351", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000119", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster earthquake satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.0077, "building_count": 43, "severe_count": 0, "class_histogram": {"no_damage": 42, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 43, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 64967.1409, "severity_score": 0.0077, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 13672.0367, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_mexico-earthquake_00000119_0352", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000119", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster earthquake scene (mexico-earthquake). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "mexico-earthquake", "disaster_type": "earthquake", "capture_date": "2017-09-20T17:46:11.000Z", "gsd": 2.6503215, "damage_summary": {"total_buildings": 54, "classified_buildings": 54, "counts": {"no-damage": 53, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 1.85}, "area_summary": {"total_footprint_m2": 78639.1776, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 1035969, "1": 12265, "2": 342, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 97.2872, "2": 2.7128, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-florence_00000477_0353", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000477", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a flooding event (hurricane-florence) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 11, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 12, "classified_buildings": 12, "damaged_count": 1, "severe_count": 0, "damaged_pct": 8.33, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 34281.9746, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-florence_00000477_0354", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000477", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "For a post-disaster flooding scene (hurricane-florence), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 443.0109, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 27495.0462, "severity_score": 0.0367, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 443.0109, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 27495.0462, "severity_score": 0.0367, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6343.9175, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-florence_00000477_0355", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000477", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G21", "severity_score": 0.11, "building_count": 3, "severe_count": 0, "class_histogram": {"no_damage": 2, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G21", "row": 2, "col": 1, "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 10141.4425, "severity_score": 0.11, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-florence_00000477_0356", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000477", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-florence). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-florence", "disaster_type": "flooding", "capture_date": "2018-09-20T16:04:41.000Z", "gsd": 2.0916247, "damage_summary": {"total_buildings": 12, "classified_buildings": 12, "counts": {"no-damage": 11, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 8.33}, "area_summary": {"total_footprint_m2": 34281.9746, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 1040120, "1": 7170, "2": 1286, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 84.7919, "2": 15.2081, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_tuscaloosa-tornado_00000059_0357", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000059", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a wind event (tuscaloosa-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 59, "minor-damage": 9, "major-damage": 0, "destroyed": 0, "unclassified": 2}, "total_buildings": 70, "classified_buildings": 68, "damaged_count": 9, "severe_count": 0, "damaged_pct": 13.24, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 247452.7259, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_tuscaloosa-tornado_00000059_0358", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000059", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "For a post-disaster wind scene (tuscaloosa-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 23, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 51672.2834, "severity_score": 0.0861, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 28628.4151, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 25, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 125338.2482, "severity_score": 0.0396, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 23, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 51672.2834, "severity_score": 0.0861, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 28628.4151, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 25, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 125338.2482, "severity_score": 0.0396, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 41813.7792, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_tuscaloosa-tornado_00000059_0359", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000059", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G21", "severity_score": 0.33, "building_count": 3, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 3, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G21", "row": 2, "col": 1, "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 37658.8128, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 26172.6022, "severity_score": 0.18, "dominant_damage": "minor-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 6306.0749, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_tuscaloosa-tornado_00000059_0360", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000059", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (tuscaloosa-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "tuscaloosa-tornado", "disaster_type": "wind", "capture_date": "2011-05-19T16:48:01.085Z", "gsd": 1.85690224170685, "damage_summary": {"total_buildings": 70, "classified_buildings": 68, "counts": {"no-damage": 59, "minor-damage": 9, "major-damage": 0, "destroyed": 0, "unclassified": 2}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 13.24}, "area_summary": {"total_footprint_m2": 247452.7259, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000090_0361", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000090", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 12, "minor-damage": 3, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 15, "classified_buildings": 15, "damaged_count": 3, "severe_count": 0, "damaged_pct": 20.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 38922.3847, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000090_0362", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000090", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "none", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 14, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 34996.4008, "severity_score": 0.0707, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 14, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 34996.4008, "severity_score": 0.0707, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 3925.9839, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000090_0363", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000090", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G31", "severity_score": 0.33, "building_count": 1, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G31", "row": 3, "col": 1, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 13410.0494, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "G21", "row": 2, "col": 1, "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 16297.0322, "severity_score": 0.066, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000090_0364", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000090", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 15, "classified_buildings": 15, "counts": {"no-damage": 12, "minor-damage": 3, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 20.0}, "area_summary": {"total_footprint_m2": 38922.3847, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000277_0365", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000277", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 154, "minor-damage": 6, "major-damage": 0, "destroyed": 0, "unclassified": 4}, "total_buildings": 164, "classified_buildings": 160, "damaged_count": 6, "severe_count": 0, "damaged_pct": 3.75, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 349121.129, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000277_0366", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000277", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 82, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 173560.6849, "severity_score": 0.0121, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 41, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 88662.5363, "severity_score": 0.0241, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 41, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 86897.9078, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 82, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 173560.6849, "severity_score": 0.0121, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 41, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 88662.5363, "severity_score": 0.0241, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 41, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 86897.9078, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000277_0367", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000277", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G13", "severity_score": 0.0381, "building_count": 26, "severe_count": 0, "class_histogram": {"no_damage": 22, "minor_damage": 3, "major_damage": 0, "destroyed": 0, "unclassified": 1}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G13", "row": 1, "col": 3, "building_count": 26, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 58986.9013, "severity_score": 0.0381, "dominant_damage": "no-damage"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 29, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 66997.92, "severity_score": 0.0341, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 13, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 24261.3325, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000277_0368", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000277", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:17:02.085Z", "gsd": 2.9729018, "damage_summary": {"total_buildings": 164, "classified_buildings": 160, "counts": {"no-damage": 154, "minor-damage": 6, "major-damage": 0, "destroyed": 0, "unclassified": 4}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 3.75}, "area_summary": {"total_footprint_m2": 349121.129, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_pinery-bushfire_00000134_0373", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000134", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a fire event (pinery-bushfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 4, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 9}, "total_buildings": 14, "classified_buildings": 5, "damaged_count": 1, "severe_count": 0, "damaged_pct": 20.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 5192.9301, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_pinery-bushfire_00000134_0374", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000134", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "minor-moderate", "question": "For a post-disaster fire scene (pinery-bushfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "none", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 14, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 5192.9301, "severity_score": 0.0236, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_pinery-bushfire_00000134_0375", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000134", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G32", "severity_score": 0.33, "building_count": 1, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G32", "row": 3, "col": 2, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 565.514, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_pinery-bushfire_00000134_0376", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000134", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (pinery-bushfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "pinery-bushfire", "disaster_type": "fire", "capture_date": "2015-11-29T01:09:22.998Z", "gsd": 1.65928483009338, "damage_summary": {"total_buildings": 14, "classified_buildings": 5, "counts": {"no-damage": 4, "minor-damage": 1, "major-damage": 0, "destroyed": 0, "unclassified": 9}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 20.0}, "area_summary": {"total_footprint_m2": 5192.9301, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 0}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-matthew_00000086_0377", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000086", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a wind event (hurricane-matthew) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 12, "minor-damage": 38, "major-damage": 0, "destroyed": 0, "unclassified": 6}, "total_buildings": 56, "classified_buildings": 50, "damaged_count": 38, "severe_count": 0, "damaged_pct": 76.0, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 75732.5592, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-matthew_00000086_0378", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000086", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "For a post-disaster wind scene (hurricane-matthew), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "minor-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 19, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 19210.2093, "severity_score": 0.1737, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 21383.8253, "severity_score": 0.22, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 25, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 31466.3132, "severity_score": 0.2508, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 19, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 19210.2093, "severity_score": 0.1737, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 21383.8253, "severity_score": 0.22, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 25, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 31466.3132, "severity_score": 0.2508, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 3672.2114, "severity_score": 0.33, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-matthew_00000086_0379", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000086", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.33, "building_count": 4, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 4, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 3813.3901, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "G03", "row": 0, "col": 3, "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 13822.0272, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 793.7771, "severity_score": 0.33, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-matthew_00000086_0380", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000086", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-matthew). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-matthew", "disaster_type": "wind", "capture_date": "2016-10-01T15:29:45.000Z", "gsd": 2.37887692, "damage_summary": {"total_buildings": 56, "classified_buildings": 50, "counts": {"no-damage": 12, "minor-damage": 38, "major-damage": 0, "destroyed": 0, "unclassified": 6}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 76.0}, "area_summary": {"total_footprint_m2": 75732.5592, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 1034567, "1": 3419, "2": 10590, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 24.4057, "2": 75.5943, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_mexico-earthquake_00000075_0381", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000075", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a earthquake event (mexico-earthquake) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 42, "minor-damage": 2, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 44, "classified_buildings": 44, "damaged_count": 2, "severe_count": 0, "damaged_pct": 4.55, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 1525708.6349, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_mexico-earthquake_00000075_0382", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000075", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "minor-moderate", "question": "For a post-disaster earthquake scene (mexico-earthquake), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "none", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 20, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 862625.0364, "severity_score": 0.033, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 381689.2128, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 20, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 862625.0364, "severity_score": 0.033, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 381689.2128, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 16, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 281394.3857, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_mexico-earthquake_00000075_0383", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000075", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster earthquake satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G03", "severity_score": 0.0388, "building_count": 17, "severe_count": 0, "class_histogram": {"no_damage": 15, "minor_damage": 2, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G03", "row": 0, "col": 3, "building_count": 17, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 49744.7671, "severity_score": 0.0388, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_mexico-earthquake_00000075_0384", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000075", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster earthquake scene (mexico-earthquake). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "mexico-earthquake", "disaster_type": "earthquake", "capture_date": "2017-09-20T17:46:11.000Z", "gsd": 2.6503215, "damage_summary": {"total_buildings": 44, "classified_buildings": 44, "counts": {"no-damage": 42, "minor-damage": 2, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 4.55}, "area_summary": {"total_footprint_m2": 1525708.6349, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 827426, "1": 219145, "2": 2005, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 99.0934, "2": 0.9066, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_mexico-earthquake_00000074_0393", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000074", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "minor-moderate", "question": "A post-disaster satellite scene from a earthquake event (mexico-earthquake) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 289, "minor-damage": 7, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "total_buildings": 296, "classified_buildings": 296, "damaged_count": 7, "severe_count": 0, "damaged_pct": 2.36, "severe_pct": 0.0, "severe_footprint_m2": 0.0, "total_footprint_m2": 2695479.6213, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "no severe damage"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_mexico-earthquake_00000074_0394", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000074", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "minor-moderate", "question": "For a post-disaster earthquake scene (mexico-earthquake), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 0, "severe_footprint_m2": 0.0, "severe_pct_in_quadrant": 0.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 69, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 695195.0712, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 66, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 645584.7075, "severity_score": 0.01, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 80, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 734125.4189, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 69, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 695195.0712, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 66, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 645584.7075, "severity_score": 0.01, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 80, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 734125.4189, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 81, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 620574.4237, "severity_score": 0.0204, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_mexico-earthquake_00000074_0395", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000074", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "minor-moderate", "question": "A 4\u00d74 grid has been overlaid on a post-disaster earthquake satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G33", "severity_score": 0.0717, "building_count": 23, "severe_count": 0, "class_histogram": {"no_damage": 18, "minor_damage": 5, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G33", "row": 3, "col": 3, "building_count": 23, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 172024.0244, "severity_score": 0.0717, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 17, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 169664.2894, "severity_score": 0.0194, "dominant_damage": "no-damage"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 20, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 177571.4407, "severity_score": 0.0165, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_mexico-earthquake_00000074_0396", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000074", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "minor-moderate", "question": "Create a comprehensive scene-level damage summary for this post-disaster earthquake scene (mexico-earthquake). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "mexico-earthquake", "disaster_type": "earthquake", "capture_date": "2017-09-20T17:46:11.000Z", "gsd": 2.6503215, "damage_summary": {"total_buildings": 296, "classified_buildings": 296, "counts": {"no-damage": 289, "minor-damage": 7, "major-damage": 0, "destroyed": 0, "unclassified": 0}, "severe_count": 0, "severe_pct": 0.0, "damaged_pct": 2.36}, "area_summary": {"total_footprint_m2": 2695479.6213, "severe_footprint_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "no severe damage", "concentration_score": null, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 0, "polygon_severe_count": 0, "pixel_histogram": {"0": 646777, "1": 393739, "2": 8060, "3": 0, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 97.994, "2": 2.006, "3": 0.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-florence_00000181_0401", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000181", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (hurricane-florence) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 20, "minor-damage": 0, "major-damage": 11, "destroyed": 0, "unclassified": 1}, "total_buildings": 32, "classified_buildings": 31, "damaged_count": 11, "severe_count": 11, "damaged_pct": 35.48, "severe_pct": 35.48, "severe_footprint_m2": 35991.7055, "total_footprint_m2": 102060.3104, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-florence_00000181_0402", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000181", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (hurricane-florence), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 7, "severe_footprint_m2": 21031.8543, "severe_pct_in_quadrant": 70.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 10, "severe_count": 7, "severe_pct": 70.0, "severe_area_m2": 21031.8543, "total_area_m2": 31487.6371, "severity_score": 0.469, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 4, "severe_pct": 80.0, "severe_area_m2": 14959.8512, "total_area_m2": 15465.6946, "severity_score": 0.536, "dominant_damage": "major-damage"}, {"cell_id": "Q0", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 19599.232, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 19599.232, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 4, "severe_pct": 80.0, "severe_area_m2": 14959.8512, "total_area_m2": 15465.6946, "severity_score": 0.536, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 35507.7467, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 10, "severe_count": 7, "severe_pct": 70.0, "severe_area_m2": 21031.8543, "total_area_m2": 31487.6371, "severity_score": 0.469, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-florence_00000181_0403", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000181", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G03", "severity_score": 0.67, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 2, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G03", "row": 0, "col": 3, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 5408.3975, "total_area_m2": 5408.3975, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 5549.722, "total_area_m2": 5549.722, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 4001.7317, "total_area_m2": 4001.7317, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-florence_00000181_0404", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000181", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (hurricane-florence), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 11, "severe_footprint_m2": 35991.7055, "major_damage_area_m2": 35991.7055, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-florence_00000181_0405", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000181", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 11, "max_severe_in_cell": 7, "concentration_score": 0.6364, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 4, "Q2": 0, "Q3": 7}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-florence_00000181_0406", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000181", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-florence). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-florence", "disaster_type": "flooding", "capture_date": "2018-09-20T16:04:41.000Z", "gsd": 2.0916247, "damage_summary": {"total_buildings": 32, "classified_buildings": 31, "counts": {"no-damage": 20, "minor-damage": 0, "major-damage": 11, "destroyed": 0, "unclassified": 1}, "severe_count": 11, "severe_pct": 35.48, "damaged_pct": 35.48}, "area_summary": {"total_footprint_m2": 102060.3104, "severe_footprint_m2": 35991.7055, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6364, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 8733, "polygon_severe_count": 11, "pixel_histogram": {"0": 1023911, "1": 15932, "2": 0, "3": 8733, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 64.5936, "2": 0.0, "3": 35.4064, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-harvey_00000455_0413", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000455", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (hurricane-harvey) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 29, "minor-damage": 7, "major-damage": 11, "destroyed": 0, "unclassified": 0}, "total_buildings": 47, "classified_buildings": 47, "damaged_count": 18, "severe_count": 11, "damaged_pct": 38.3, "severe_pct": 23.4, "severe_footprint_m2": 726663.5474, "total_footprint_m2": 1629529.6297, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-harvey_00000455_0414", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000455", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (hurricane-harvey), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 8, "severe_footprint_m2": 303974.1506, "severe_pct_in_quadrant": 100.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 303974.1506, "total_area_m2": 303974.1506, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 23, "severe_count": 1, "severe_pct": 4.35, "severe_area_m2": 279077.3033, "total_area_m2": 868770.1014, "severity_score": 0.1009, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 143612.0935, "total_area_m2": 203881.2969, "severity_score": 0.5567, "dominant_damage": "major-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 143612.0935, "total_area_m2": 203881.2969, "severity_score": 0.5567, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 303974.1506, "total_area_m2": 303974.1506, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 13, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 252904.0808, "severity_score": 0.0254, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 23, "severe_count": 1, "severe_pct": 4.35, "severe_area_m2": 279077.3033, "total_area_m2": 868770.1014, "severity_score": 0.1009, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-harvey_00000455_0415", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000455", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 0.67, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 53388.9106, "total_area_m2": 53388.9106, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 112771.6081, "total_area_m2": 112771.6081, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G03", "row": 0, "col": 3, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 3871.9972, "total_area_m2": 3871.9972, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-harvey_00000455_0416", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000455", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (hurricane-harvey), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 11, "severe_footprint_m2": 726663.5474, "major_damage_area_m2": 726663.5474, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-harvey_00000455_0417", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000455", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 11, "max_severe_in_cell": 8, "concentration_score": 0.7273, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 2, "Q1": 8, "Q2": 0, "Q3": 1}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-harvey_00000455_0418", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000455", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-harvey). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-harvey", "disaster_type": "flooding", "capture_date": "2017-08-31T17:38:50.685Z", "gsd": 3.024613, "damage_summary": {"total_buildings": 47, "classified_buildings": 47, "counts": {"no-damage": 29, "minor-damage": 7, "major-damage": 11, "destroyed": 0, "unclassified": 0}, "severe_count": 11, "severe_pct": 23.4, "damaged_pct": 38.3}, "area_summary": {"total_footprint_m2": 1629529.6297, "severe_footprint_m2": 726663.5474, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.7273, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 81382, "polygon_severe_count": 11, "pixel_histogram": {"0": 864327, "1": 73864, "2": 29003, "3": 81382, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 40.0892, "2": 15.7412, "3": 44.1696, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_midwest-flooding_00000353_0425", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000353", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (midwest-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 5, "minor-damage": 4, "major-damage": 3, "destroyed": 0, "unclassified": 0}, "total_buildings": 12, "classified_buildings": 12, "damaged_count": 7, "severe_count": 3, "damaged_pct": 58.33, "severe_pct": 25.0, "severe_footprint_m2": 5330.4961, "total_footprint_m2": 16834.21, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_midwest-flooding_00000353_0426", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000353", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (midwest-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 2, "severe_footprint_m2": 5076.8624, "severe_pct_in_quadrant": 100.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 5076.8624, "total_area_m2": 5076.8624, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 1, "severe_pct": 11.11, "severe_area_m2": 253.6337, "total_area_m2": 11150.2631, "severity_score": 0.2211, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 5076.8624, "total_area_m2": 5076.8624, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 607.0845, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 1, "severe_pct": 11.11, "severe_area_m2": 253.6337, "total_area_m2": 11150.2631, "severity_score": 0.2211, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_midwest-flooding_00000353_0427", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000353", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 0.67, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 627.1437, "total_area_m2": 627.1437, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 4449.7187, "total_area_m2": 4449.7187, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G33", "row": 3, "col": 3, "building_count": 6, "severe_count": 1, "severe_pct": 16.67, "severe_area_m2": 253.6337, "total_area_m2": 7367.0566, "severity_score": 0.2767, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_midwest-flooding_00000353_0428", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000353", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (midwest-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 3, "severe_footprint_m2": 5330.4961, "major_damage_area_m2": 5330.4961, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_midwest-flooding_00000353_0429", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000353", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 3, "max_severe_in_cell": 2, "concentration_score": 0.6667, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 2, "Q1": 0, "Q2": 0, "Q3": 1}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_midwest-flooding_00000353_0430", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000353", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (midwest-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "midwest-flooding", "disaster_type": "flooding", "capture_date": "2019-05-30T17:35:04.000Z", "gsd": 1.2524601, "damage_summary": {"total_buildings": 12, "classified_buildings": 12, "counts": {"no-damage": 5, "minor-damage": 4, "major-damage": 3, "destroyed": 0, "unclassified": 0}, "severe_count": 3, "severe_pct": 25.0, "damaged_pct": 58.33}, "area_summary": {"total_footprint_m2": 16834.21, "severe_footprint_m2": 5330.4961, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6667, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 3594, "polygon_severe_count": 3, "pixel_histogram": {"0": 1037190, "1": 4361, "2": 3431, "3": 3594, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 38.3014, "2": 30.1335, "3": 31.5651, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_santa-rosa-wildfire_00000180_0431", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000180", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a fire event (santa-rosa-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 41, "minor-damage": 0, "major-damage": 2, "destroyed": 0, "unclassified": 0}, "total_buildings": 43, "classified_buildings": 43, "damaged_count": 2, "severe_count": 2, "damaged_pct": 4.65, "severe_pct": 4.65, "severe_footprint_m2": 893.6604, "total_footprint_m2": 350790.5367, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_santa-rosa-wildfire_00000180_0432", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000180", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "For a post-disaster fire scene (santa-rosa-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 2, "severe_footprint_m2": 893.6604, "severe_pct_in_quadrant": 20.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 10, "severe_count": 2, "severe_pct": 20.0, "severe_area_m2": 893.6604, "total_area_m2": 46508.5104, "severity_score": 0.134, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 104252.8632, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 91020.2334, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 104252.8632, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 91020.2334, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 10, "severe_count": 2, "severe_pct": 20.0, "severe_area_m2": 893.6604, "total_area_m2": 46508.5104, "severity_score": 0.134, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 27, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 109008.9297, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_santa-rosa-wildfire_00000180_0433", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000180", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G30", "severity_score": 0.1914, "building_count": 7, "severe_count": 2, "class_histogram": {"no_damage": 5, "minor_damage": 0, "major_damage": 2, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G30", "row": 3, "col": 0, "building_count": 7, "severe_count": 2, "severe_pct": 28.57, "severe_area_m2": 893.6604, "total_area_m2": 15132.2494, "severity_score": 0.1914, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 38311.766, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_santa-rosa-wildfire_00000180_0434", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000180", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "From the post-disaster fire scene (santa-rosa-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 2, "severe_footprint_m2": 893.6604, "major_damage_area_m2": 893.6604, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_santa-rosa-wildfire_00000180_0435", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000180", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 2, "max_severe_in_cell": 2, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 2, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_santa-rosa-wildfire_00000180_0436", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000180", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (santa-rosa-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "santa-rosa-wildfire", "disaster_type": "fire", "capture_date": "2017-10-11T19:19:41.000Z", "gsd": 1.8769937, "damage_summary": {"total_buildings": 43, "classified_buildings": 43, "counts": {"no-damage": 41, "minor-damage": 0, "major-damage": 2, "destroyed": 0, "unclassified": 0}, "severe_count": 2, "severe_pct": 4.65, "damaged_pct": 4.65}, "area_summary": {"total_footprint_m2": 350790.5367, "severe_footprint_m2": 893.6604, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 291, "polygon_severe_count": 2, "pixel_histogram": {"0": 945626, "1": 102659, "2": 0, "3": 291, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 99.7173, "2": 0.0, "3": 0.2827, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000444_0437", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000444", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 0, "major-damage": 11, "destroyed": 0, "unclassified": 0}, "total_buildings": 11, "classified_buildings": 11, "damaged_count": 11, "severe_count": 11, "damaged_pct": 100.0, "severe_pct": 100.0, "severe_footprint_m2": 6629.2814, "total_footprint_m2": 6629.2814, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000444_0438", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000444", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 11, "severe_footprint_m2": 6629.2814, "severe_pct_in_quadrant": 100.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 11, "severe_count": 11, "severe_pct": 100.0, "severe_area_m2": 6629.2814, "total_area_m2": 6629.2814, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 11, "severe_count": 11, "severe_pct": 100.0, "severe_area_m2": 6629.2814, "total_area_m2": 6629.2814, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000444_0439", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000444", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 0.67, "building_count": 5, "severe_count": 5, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 5, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 5, "severe_count": 5, "severe_pct": 100.0, "severe_area_m2": 3984.1066, "total_area_m2": 3984.1066, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G03", "row": 0, "col": 3, "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 2645.1748, "total_area_m2": 2645.1748, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000444_0440", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000444", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 11, "severe_footprint_m2": 6629.2814, "major_damage_area_m2": 6629.2814, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000444_0441", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000444", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 11, "max_severe_in_cell": 11, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 11, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000444_0442", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000444", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 11, "classified_buildings": 11, "counts": {"no-damage": 0, "minor-damage": 0, "major-damage": 11, "destroyed": 0, "unclassified": 0}, "severe_count": 11, "severe_pct": 100.0, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 6629.2814, "severe_footprint_m2": 6629.2814, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 11}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000020_0443", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000020", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 119, "minor-damage": 63, "major-damage": 13, "destroyed": 0, "unclassified": 5}, "total_buildings": 200, "classified_buildings": 195, "damaged_count": 76, "severe_count": 13, "damaged_pct": 38.97, "severe_pct": 6.67, "severe_footprint_m2": 66753.4114, "total_footprint_m2": 583242.7404, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000020_0444", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000020", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 9, "severe_footprint_m2": 56931.3757, "severe_pct_in_quadrant": 13.64, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 66, "severe_count": 9, "severe_pct": 13.64, "severe_area_m2": 56931.3757, "total_area_m2": 232546.6052, "severity_score": 0.2064, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 48, "severe_count": 4, "severe_pct": 8.33, "severe_area_m2": 9822.0357, "total_area_m2": 115993.4464, "severity_score": 0.1727, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 20, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 84199.9508, "severity_score": 0.0495, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 20, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 84199.9508, "severity_score": 0.0495, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 66, "severe_count": 9, "severe_pct": 13.64, "severe_area_m2": 56931.3757, "total_area_m2": 232546.6052, "severity_score": 0.2064, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 66, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 150502.738, "severity_score": 0.1, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 48, "severe_count": 4, "severe_pct": 8.33, "severe_area_m2": 9822.0357, "total_area_m2": 115993.4464, "severity_score": 0.1727, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000020_0445", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000020", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G03", "severity_score": 0.468, "building_count": 5, "severe_count": 3, "class_histogram": {"no_damage": 0, "minor_damage": 1, "major_damage": 3, "destroyed": 0, "unclassified": 1}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G03", "row": 0, "col": 3, "building_count": 5, "severe_count": 3, "severe_pct": 60.0, "severe_area_m2": 10151.0028, "total_area_m2": 13890.2371, "severity_score": 0.468, "dominant_damage": "major-damage"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 11, "severe_count": 3, "severe_pct": 27.27, "severe_area_m2": 4647.7655, "total_area_m2": 20800.0177, "severity_score": 0.4227, "dominant_damage": "minor-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 12, "severe_count": 5, "severe_pct": 41.67, "severe_area_m2": 13049.3333, "total_area_m2": 31791.0478, "severity_score": 0.3892, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000020_0446", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000020", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 13, "severe_footprint_m2": 66753.4114, "major_damage_area_m2": 66753.4114, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000020_0447", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000020", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 13, "max_severe_in_cell": 9, "concentration_score": 0.6923, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 9, "Q2": 0, "Q3": 4}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000020_0448", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000020", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 200, "classified_buildings": 195, "counts": {"no-damage": 119, "minor-damage": 63, "major-damage": 13, "destroyed": 0, "unclassified": 5}, "severe_count": 13, "severe_pct": 6.67, "damaged_pct": 38.97}, "area_summary": {"total_footprint_m2": 583242.7404, "severe_footprint_m2": 66753.4114, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6923, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 13}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_lower-puna-volcano_00000121_0449", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000121", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a volcano event (lower-puna-volcano) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 13, "minor-damage": 8, "major-damage": 6, "destroyed": 1, "unclassified": 0}, "total_buildings": 28, "classified_buildings": 28, "damaged_count": 15, "severe_count": 7, "damaged_pct": 53.57, "severe_pct": 25.0, "severe_footprint_m2": 10480.591, "total_footprint_m2": 48916.9074, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_lower-puna-volcano_00000121_0450", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000121", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "major-heavy", "question": "For a post-disaster volcano scene (lower-puna-volcano), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 4, "severe_footprint_m2": 9247.3335, "severe_pct_in_quadrant": 30.77, "dominant_damage": "minor-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 13, "severe_count": 4, "severe_pct": 30.77, "severe_area_m2": 9247.3335, "total_area_m2": 28030.3904, "severity_score": 0.4092, "dominant_damage": "minor-damage"}, {"cell_id": "Q0", "building_count": 8, "severe_count": 3, "severe_pct": 37.5, "severe_area_m2": 1233.2575, "total_area_m2": 11918.6387, "severity_score": 0.2925, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 8967.8783, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 8, "severe_count": 3, "severe_pct": 37.5, "severe_area_m2": 1233.2575, "total_area_m2": 11918.6387, "severity_score": 0.2925, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 13, "severe_count": 4, "severe_pct": 30.77, "severe_area_m2": 9247.3335, "total_area_m2": 28030.3904, "severity_score": 0.4092, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 8967.8783, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_lower-puna-volcano_00000121_0451", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000121", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster volcano satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 0.67, "building_count": 3, "severe_count": 3, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 3, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 1233.2575, "total_area_m2": 1233.2575, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 1554.6133, "total_area_m2": 6155.7616, "severity_score": 0.5567, "dominant_damage": "major-damage"}, {"cell_id": "G03", "row": 0, "col": 3, "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 7492.2967, "total_area_m2": 7751.6282, "severity_score": 0.5, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_lower-puna-volcano_00000121_0452", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000121", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "major-heavy", "question": "From the post-disaster volcano scene (lower-puna-volcano), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 7, "severe_footprint_m2": 10480.591, "major_damage_area_m2": 10280.1675, "destroyed_area_m2": 200.4235, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_lower-puna-volcano_00000121_0453", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000121", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster volcano scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 7, "max_severe_in_cell": 4, "concentration_score": 0.5714, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 3, "Q1": 4, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_lower-puna-volcano_00000121_0454", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000121", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster volcano scene (lower-puna-volcano). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "lower-puna-volcano", "disaster_type": "volcano", "capture_date": "2018-05-23T20:59:21.000Z", "gsd": 2.2423599, "damage_summary": {"total_buildings": 28, "classified_buildings": 28, "counts": {"no-damage": 13, "minor-damage": 8, "major-damage": 6, "destroyed": 1, "unclassified": 0}, "severe_count": 7, "severe_pct": 25.0, "damaged_pct": 53.57}, "area_summary": {"total_footprint_m2": 48916.9074, "severe_footprint_m2": 10480.591, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5714, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 7}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000563_0461", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000563", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 46, "minor-damage": 78, "major-damage": 16, "destroyed": 0, "unclassified": 3}, "total_buildings": 143, "classified_buildings": 140, "damaged_count": 94, "severe_count": 16, "damaged_pct": 67.14, "severe_pct": 11.43, "severe_footprint_m2": 51390.0182, "total_footprint_m2": 477709.1048, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000563_0462", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000563", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 10, "severe_footprint_m2": 27107.4442, "severe_pct_in_quadrant": 37.04, "dominant_damage": "minor-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 27, "severe_count": 10, "severe_pct": 37.04, "severe_area_m2": 27107.4442, "total_area_m2": 84091.1155, "severity_score": 0.4559, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 76, "severe_count": 4, "severe_pct": 5.26, "severe_area_m2": 18890.1655, "total_area_m2": 273185.5911, "severity_score": 0.1699, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 22, "severe_count": 1, "severe_pct": 4.55, "severe_area_m2": 2992.5759, "total_area_m2": 50292.4707, "severity_score": 0.2705, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 27, "severe_count": 10, "severe_pct": 37.04, "severe_area_m2": 27107.4442, "total_area_m2": 84091.1155, "severity_score": 0.4559, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 22, "severe_count": 1, "severe_pct": 4.55, "severe_area_m2": 2992.5759, "total_area_m2": 50292.4707, "severity_score": 0.2705, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 76, "severe_count": 4, "severe_pct": 5.26, "severe_area_m2": 18890.1655, "total_area_m2": 273185.5911, "severity_score": 0.1699, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 18, "severe_count": 1, "severe_pct": 5.56, "severe_area_m2": 2399.8326, "total_area_m2": 70139.9275, "severity_score": 0.2939, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000563_0463", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000563", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.5, "building_count": 6, "severe_count": 3, "class_histogram": {"no_damage": 0, "minor_damage": 3, "major_damage": 3, "destroyed": 0, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 6, "severe_count": 3, "severe_pct": 50.0, "severe_area_m2": 3002.9665, "total_area_m2": 12631.271, "severity_score": 0.5, "dominant_damage": "minor-damage"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 8, "severe_count": 4, "severe_pct": 50.0, "severe_area_m2": 13827.6529, "total_area_m2": 29448.7894, "severity_score": 0.5, "dominant_damage": "minor-damage"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 13, "severe_count": 3, "severe_pct": 23.08, "severe_area_m2": 10276.8248, "total_area_m2": 42011.0551, "severity_score": 0.4085, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000563_0464", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000563", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 16, "severe_footprint_m2": 51390.0182, "major_damage_area_m2": 51390.0182, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000563_0465", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000563", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 16, "max_severe_in_cell": 10, "concentration_score": 0.625, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 10, "Q1": 1, "Q2": 4, "Q3": 1}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000563_0466", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000563", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 143, "classified_buildings": 140, "counts": {"no-damage": 46, "minor-damage": 78, "major-damage": 16, "destroyed": 0, "unclassified": 3}, "severe_count": 16, "severe_pct": 11.43, "damaged_pct": 67.14}, "area_summary": {"total_footprint_m2": 477709.1048, "severe_footprint_m2": 51390.0182, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.625, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 16}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_pinery-bushfire_00001507_0467", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001507", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a fire event (pinery-bushfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 4, "minor-damage": 1, "major-damage": 1, "destroyed": 0, "unclassified": 7}, "total_buildings": 13, "classified_buildings": 6, "damaged_count": 2, "severe_count": 1, "damaged_pct": 33.33, "severe_pct": 16.67, "severe_footprint_m2": 6435.7118, "total_footprint_m2": 23250.9949, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_pinery-bushfire_00001507_0468", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001507", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "For a post-disaster fire scene (pinery-bushfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 1, "severe_footprint_m2": 6435.7118, "severe_pct_in_quadrant": 33.33, "dominant_damage": "minor-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 6435.7118, "total_area_m2": 8893.7968, "severity_score": 0.3333, "dominant_damage": "minor-damage"}, {"cell_id": "Q0", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 403.2333, "severity_score": 0.0, "dominant_damage": "unclassified"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 403.2333, "severity_score": 0.0, "dominant_damage": "unclassified"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 6435.7118, "total_area_m2": 8893.7968, "severity_score": 0.3333, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 13953.9648, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_pinery-bushfire_00001507_0469", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001507", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G12", "severity_score": 0.3333, "building_count": 3, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 1, "major_damage": 1, "destroyed": 0, "unclassified": 1}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G12", "row": 1, "col": 2, "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 6435.7118, "total_area_m2": 8893.7968, "severity_score": 0.3333, "dominant_damage": "minor-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_pinery-bushfire_00001507_0470", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001507", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "From the post-disaster fire scene (pinery-bushfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 1, "severe_footprint_m2": 6435.7118, "major_damage_area_m2": 6435.7118, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_pinery-bushfire_00001507_0471", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001507", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 1, "max_severe_in_cell": 1, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 1, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_pinery-bushfire_00001507_0472", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001507", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (pinery-bushfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "pinery-bushfire", "disaster_type": "fire", "capture_date": "2015-11-27T00:39:41.429Z", "gsd": 2.03157305717468, "damage_summary": {"total_buildings": 13, "classified_buildings": 6, "counts": {"no-damage": 4, "minor-damage": 1, "major-damage": 1, "destroyed": 0, "unclassified": 7}, "severe_count": 1, "severe_pct": 16.67, "damaged_pct": 33.33}, "area_summary": {"total_footprint_m2": 23250.9949, "severe_footprint_m2": 6435.7118, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 1}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-harvey_00000081_0473", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000081", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (hurricane-harvey) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 4, "major-damage": 23, "destroyed": 0, "unclassified": 6}, "total_buildings": 33, "classified_buildings": 27, "damaged_count": 27, "severe_count": 23, "damaged_pct": 100.0, "severe_pct": 85.19, "severe_footprint_m2": 241740.9646, "total_footprint_m2": 324524.8312, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-harvey_00000081_0474", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000081", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (hurricane-harvey), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 8, "severe_footprint_m2": 90583.9424, "severe_pct_in_quadrant": 72.73, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 11, "severe_count": 8, "severe_pct": 72.73, "severe_area_m2": 90583.9424, "total_area_m2": 129432.732, "severity_score": 0.5773, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 13, "severe_count": 8, "severe_pct": 61.54, "severe_area_m2": 87614.3952, "total_area_m2": 117827.2107, "severity_score": 0.4377, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 7, "severe_count": 6, "severe_pct": 85.71, "severe_area_m2": 49942.6564, "total_area_m2": 56544.7488, "severity_score": 0.5743, "dominant_damage": "major-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 11, "severe_count": 8, "severe_pct": 72.73, "severe_area_m2": 90583.9424, "total_area_m2": 129432.732, "severity_score": 0.5773, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 13599.9706, "total_area_m2": 20720.1397, "severity_score": 0.335, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 13, "severe_count": 8, "severe_pct": 61.54, "severe_area_m2": 87614.3952, "total_area_m2": 117827.2107, "severity_score": 0.4377, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 7, "severe_count": 6, "severe_pct": 85.71, "severe_area_m2": 49942.6564, "total_area_m2": 56544.7488, "severity_score": 0.5743, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-harvey_00000081_0475", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000081", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 0.67, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 5251.9836, "total_area_m2": 5251.9836, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 15875.3046, "total_area_m2": 15875.3046, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 46914.9965, "total_area_m2": 46914.9965, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-harvey_00000081_0476", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000081", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (hurricane-harvey), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 23, "severe_footprint_m2": 241740.9646, "major_damage_area_m2": 241740.9646, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-harvey_00000081_0477", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000081", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 23, "max_severe_in_cell": 8, "concentration_score": 0.3478, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 8, "Q1": 1, "Q2": 8, "Q3": 6}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-harvey_00000081_0478", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000081", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-harvey). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-harvey", "disaster_type": "flooding", "capture_date": "2017-08-31T17:39:25.085Z", "gsd": 3.1466618, "damage_summary": {"total_buildings": 33, "classified_buildings": 27, "counts": {"no-damage": 0, "minor-damage": 4, "major-damage": 23, "destroyed": 0, "unclassified": 6}, "severe_count": 23, "severe_pct": 85.19, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 324524.8312, "severe_footprint_m2": 241740.9646, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.3478, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 25870, "polygon_severe_count": 23, "pixel_histogram": {"0": 1018274, "1": 0, "2": 4432, "3": 25870, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 14.6261, "3": 85.3739, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000494_0485", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000494", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 27, "minor-damage": 0, "major-damage": 3, "destroyed": 0, "unclassified": 0}, "total_buildings": 30, "classified_buildings": 30, "damaged_count": 3, "severe_count": 3, "damaged_pct": 10.0, "severe_pct": 10.0, "severe_footprint_m2": 5624.6425, "total_footprint_m2": 41012.7799, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000494_0486", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000494", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 3, "severe_footprint_m2": 5624.6425, "severe_pct_in_quadrant": 12.5, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 24, "severe_count": 3, "severe_pct": 12.5, "severe_area_m2": 5624.6425, "total_area_m2": 32823.0371, "severity_score": 0.0838, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1093.8411, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1093.8411, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 24, "severe_count": 3, "severe_pct": 12.5, "severe_area_m2": 5624.6425, "total_area_m2": 32823.0371, "severity_score": 0.0838, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 7095.9017, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000494_0487", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000494", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G03", "severity_score": 0.67, "building_count": 3, "severe_count": 3, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 3, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G03", "row": 0, "col": 3, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 5624.6425, "total_area_m2": 5624.6425, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1093.8411, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000494_0488", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000494", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 3, "severe_footprint_m2": 5624.6425, "major_damage_area_m2": 5624.6425, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000494_0489", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000494", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 3, "max_severe_in_cell": 3, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 3, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000494_0490", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000494", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 30, "classified_buildings": 30, "counts": {"no-damage": 27, "minor-damage": 0, "major-damage": 3, "destroyed": 0, "unclassified": 0}, "severe_count": 3, "severe_pct": 10.0, "damaged_pct": 10.0}, "area_summary": {"total_footprint_m2": 41012.7799, "severe_footprint_m2": 5624.6425, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 3}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000307_0509", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000307", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 37, "minor-damage": 40, "major-damage": 30, "destroyed": 7, "unclassified": 0}, "total_buildings": 114, "classified_buildings": 114, "damaged_count": 77, "severe_count": 37, "damaged_pct": 67.54, "severe_pct": 32.46, "severe_footprint_m2": 120489.2391, "total_footprint_m2": 353354.4885, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000307_0510", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000307", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 13, "severe_footprint_m2": 71304.1244, "severe_pct_in_quadrant": 40.62, "dominant_damage": "minor-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 32, "severe_count": 13, "severe_pct": 40.62, "severe_area_m2": 71304.1244, "total_area_m2": 143377.3897, "severity_score": 0.4166, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 47, "severe_count": 12, "severe_pct": 25.53, "severe_area_m2": 24417.7782, "total_area_m2": 108461.913, "severity_score": 0.3185, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 24, "severe_count": 9, "severe_pct": 37.5, "severe_area_m2": 19902.6687, "total_area_m2": 64489.0489, "severity_score": 0.3613, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 32, "severe_count": 13, "severe_pct": 40.62, "severe_area_m2": 71304.1244, "total_area_m2": 143377.3897, "severity_score": 0.4166, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 11, "severe_count": 3, "severe_pct": 27.27, "severe_area_m2": 4864.6678, "total_area_m2": 37026.1369, "severity_score": 0.3027, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 47, "severe_count": 12, "severe_pct": 25.53, "severe_area_m2": 24417.7782, "total_area_m2": 108461.913, "severity_score": 0.3185, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 24, "severe_count": 9, "severe_pct": 37.5, "severe_area_m2": 19902.6687, "total_area_m2": 64489.0489, "severity_score": 0.3613, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000307_0511", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000307", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G10", "severity_score": 0.5782, "building_count": 11, "severe_count": 9, "class_histogram": {"no_damage": 1, "minor_damage": 1, "major_damage": 9, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G10", "row": 1, "col": 0, "building_count": 11, "severe_count": 9, "severe_pct": 81.82, "severe_area_m2": 41151.6365, "total_area_m2": 46860.1924, "severity_score": 0.5782, "dominant_damage": "major-damage"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 9, "severe_count": 4, "severe_pct": 44.44, "severe_area_m2": 11305.8482, "total_area_m2": 24129.6206, "severity_score": 0.4811, "dominant_damage": "minor-damage"}, {"cell_id": "G30", "row": 3, "col": 0, "building_count": 13, "severe_count": 6, "severe_pct": 46.15, "severe_area_m2": 9667.1329, "total_area_m2": 30972.4281, "severity_score": 0.4362, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-michael_00000307_0512", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000307", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "From the post-disaster wind scene (hurricane-michael), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 37, "severe_footprint_m2": 120489.2391, "major_damage_area_m2": 108709.3646, "destroyed_area_m2": 11779.8745, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-michael_00000307_0513", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000307", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 37, "max_severe_in_cell": 13, "concentration_score": 0.3514, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 13, "Q1": 3, "Q2": 12, "Q3": 9}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000307_0514", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000307", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 114, "classified_buildings": 114, "counts": {"no-damage": 37, "minor-damage": 40, "major-damage": 30, "destroyed": 7, "unclassified": 0}, "severe_count": 37, "severe_pct": 32.46, "damaged_pct": 67.54}, "area_summary": {"total_footprint_m2": 353354.4885, "severe_footprint_m2": 120489.2391, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.3514, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 30204, "polygon_severe_count": 37, "pixel_histogram": {"0": 959849, "1": 27689, "2": 30834, "3": 27168, "4": 3036, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 31.207, "2": 34.7515, "3": 30.6198, "4": 3.4217, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000072_0527", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000072", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 67, "minor-damage": 20, "major-damage": 4, "destroyed": 1, "unclassified": 1}, "total_buildings": 93, "classified_buildings": 92, "damaged_count": 25, "severe_count": 5, "damaged_pct": 27.17, "severe_pct": 5.43, "severe_footprint_m2": 11991.6133, "total_footprint_m2": 327898.072, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000072_0528", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000072", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 1, "severe_footprint_m2": 5813.7307, "severe_pct_in_quadrant": 5.88, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 17, "severe_count": 1, "severe_pct": 5.88, "severe_area_m2": 5813.7307, "total_area_m2": 57554.5208, "severity_score": 0.1171, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 22, "severe_count": 2, "severe_pct": 9.09, "severe_area_m2": 4256.9929, "total_area_m2": 84220.2599, "severity_score": 0.1659, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 39, "severe_count": 2, "severe_pct": 5.13, "severe_area_m2": 1920.8897, "total_area_m2": 106216.3709, "severity_score": 0.0682, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 79906.9204, "severity_score": 0.132, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 22, "severe_count": 2, "severe_pct": 9.09, "severe_area_m2": 4256.9929, "total_area_m2": 84220.2599, "severity_score": 0.1659, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 17, "severe_count": 1, "severe_pct": 5.88, "severe_area_m2": 5813.7307, "total_area_m2": 57554.5208, "severity_score": 0.1171, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 39, "severe_count": 2, "severe_pct": 5.13, "severe_area_m2": 1920.8897, "total_area_m2": 106216.3709, "severity_score": 0.0682, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000072_0529", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000072", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G22", "severity_score": 0.2913, "building_count": 8, "severe_count": 2, "class_histogram": {"no_damage": 4, "minor_damage": 2, "major_damage": 1, "destroyed": 1, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G22", "row": 2, "col": 2, "building_count": 8, "severe_count": 2, "severe_pct": 25.0, "severe_area_m2": 1920.8897, "total_area_m2": 24638.1773, "severity_score": 0.2913, "dominant_damage": "no-damage"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 13903.1778, "severity_score": 0.2475, "dominant_damage": "minor-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 9, "severe_count": 2, "severe_pct": 22.22, "severe_area_m2": 4256.9929, "total_area_m2": 27259.0011, "severity_score": 0.2222, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-michael_00000072_0530", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000072", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "From the post-disaster wind scene (hurricane-michael), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 5, "severe_footprint_m2": 11991.6133, "major_damage_area_m2": 11393.0528, "destroyed_area_m2": 598.5605, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-michael_00000072_0531", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000072", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 5, "max_severe_in_cell": 2, "concentration_score": 0.4, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 0, "Q1": 2, "Q2": 1, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000072_0532", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000072", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 93, "classified_buildings": 92, "counts": {"no-damage": 67, "minor-damage": 20, "major-damage": 4, "destroyed": 1, "unclassified": 1}, "severe_count": 5, "severe_pct": 5.43, "damaged_pct": 27.17}, "area_summary": {"total_footprint_m2": 327898.072, "severe_footprint_m2": 11991.6133, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.4, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 3004, "polygon_severe_count": 5, "pixel_histogram": {"0": 966675, "1": 52750, "2": 26147, "3": 2839, "4": 165, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 64.407, "2": 31.9251, "3": 3.4664, "4": 0.2015, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-harvey_00000489_0539", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000489", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (hurricane-harvey) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 7, "minor-damage": 10, "major-damage": 26, "destroyed": 0, "unclassified": 2}, "total_buildings": 45, "classified_buildings": 43, "damaged_count": 36, "severe_count": 26, "damaged_pct": 83.72, "severe_pct": 60.47, "severe_footprint_m2": 453971.9276, "total_footprint_m2": 863491.3199, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-harvey_00000489_0540", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000489", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (hurricane-harvey), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 13, "severe_footprint_m2": 228437.1916, "severe_pct_in_quadrant": 100.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 13, "severe_count": 13, "severe_pct": 100.0, "severe_area_m2": 228437.1916, "total_area_m2": 228437.1916, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 8, "severe_pct": 88.89, "severe_area_m2": 155856.1497, "total_area_m2": 184868.4994, "severity_score": 0.6322, "dominant_damage": "major-damage"}, {"cell_id": "Q0", "building_count": 6, "severe_count": 4, "severe_pct": 66.67, "severe_area_m2": 58740.1763, "total_area_m2": 95461.7375, "severity_score": 0.4467, "dominant_damage": "major-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 6, "severe_count": 4, "severe_pct": 66.67, "severe_area_m2": 58740.1763, "total_area_m2": 95461.7375, "severity_score": 0.4467, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 13, "severe_count": 13, "severe_pct": 100.0, "severe_area_m2": 228437.1916, "total_area_m2": 228437.1916, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 17, "severe_count": 1, "severe_pct": 5.88, "severe_area_m2": 10938.41, "total_area_m2": 354723.8914, "severity_score": 0.2141, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 8, "severe_pct": 88.89, "severe_area_m2": 155856.1497, "total_area_m2": 184868.4994, "severity_score": 0.6322, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-harvey_00000489_0541", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000489", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.67, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 2, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 27569.3791, "total_area_m2": 27569.3791, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 41891.7134, "total_area_m2": 41891.7134, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 31170.7972, "total_area_m2": 31170.7972, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-harvey_00000489_0542", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000489", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (hurricane-harvey), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 26, "severe_footprint_m2": 453971.9276, "major_damage_area_m2": 453971.9276, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-harvey_00000489_0543", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000489", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 26, "max_severe_in_cell": 13, "concentration_score": 0.5, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 4, "Q1": 13, "Q2": 1, "Q3": 8}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-harvey_00000489_0544", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000489", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-harvey). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-harvey", "disaster_type": "flooding", "capture_date": "2017-08-31T17:38:50.685Z", "gsd": 3.024613, "damage_summary": {"total_buildings": 45, "classified_buildings": 43, "counts": {"no-damage": 7, "minor-damage": 10, "major-damage": 26, "destroyed": 0, "unclassified": 2}, "severe_count": 26, "severe_pct": 60.47, "damaged_pct": 83.72}, "area_summary": {"total_footprint_m2": 863491.3199, "severe_footprint_m2": 453971.9276, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 52047, "polygon_severe_count": 26, "pixel_histogram": {"0": 953803, "1": 14360, "2": 28366, "3": 52047, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 15.152, "2": 29.9305, "3": 54.9175, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000601_0545", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000601", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 164, "minor-damage": 18, "major-damage": 2, "destroyed": 0, "unclassified": 0}, "total_buildings": 184, "classified_buildings": 184, "damaged_count": 20, "severe_count": 2, "damaged_pct": 10.87, "severe_pct": 1.09, "severe_footprint_m2": 6754.9936, "total_footprint_m2": 590188.8281, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000601_0546", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000601", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 2, "severe_footprint_m2": 6754.9936, "severe_pct_in_quadrant": 16.67, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 12, "severe_count": 2, "severe_pct": 16.67, "severe_area_m2": 6754.9936, "total_area_m2": 34638.3766, "severity_score": 0.2217, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 41, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 139733.5684, "severity_score": 0.0241, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 43380.7403, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 12, "severe_count": 2, "severe_pct": 16.67, "severe_area_m2": 6754.9936, "total_area_m2": 34638.3766, "severity_score": 0.2217, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 41, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 139733.5684, "severity_score": 0.0241, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 43380.7403, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 119, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 372436.1428, "severity_score": 0.0305, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000601_0547", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000601", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 0.3883, "building_count": 6, "severe_count": 2, "class_histogram": {"no_damage": 1, "minor_damage": 3, "major_damage": 2, "destroyed": 0, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 6, "severe_count": 2, "severe_pct": 33.33, "severe_area_m2": 6754.9936, "total_area_m2": 16356.9057, "severity_score": 0.3883, "dominant_damage": "minor-damage"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 10386.012, "severity_score": 0.0825, "dominant_damage": "no-damage"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 39, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 110097.5967, "severity_score": 0.0677, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000601_0548", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000601", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 2, "severe_footprint_m2": 6754.9936, "major_damage_area_m2": 6754.9936, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000601_0549", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000601", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 2, "max_severe_in_cell": 2, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 2, "Q1": 0, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000601_0550", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000601", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 184, "classified_buildings": 184, "counts": {"no-damage": 164, "minor-damage": 18, "major-damage": 2, "destroyed": 0, "unclassified": 0}, "severe_count": 2, "severe_pct": 1.09, "damaged_pct": 10.87}, "area_summary": {"total_footprint_m2": 590188.8281, "severe_footprint_m2": 6754.9936, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 2}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000467_0551", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000467", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 115, "minor-damage": 38, "major-damage": 14, "destroyed": 2, "unclassified": 2}, "total_buildings": 171, "classified_buildings": 169, "damaged_count": 54, "severe_count": 16, "damaged_pct": 31.95, "severe_pct": 9.47, "severe_footprint_m2": 55990.5149, "total_footprint_m2": 418407.1051, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000467_0552", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000467", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 8, "severe_footprint_m2": 24038.904, "severe_pct_in_quadrant": 25.81, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 31, "severe_count": 8, "severe_pct": 25.81, "severe_area_m2": 24038.904, "total_area_m2": 68721.3232, "severity_score": 0.2261, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 57, "severe_count": 3, "severe_pct": 5.26, "severe_area_m2": 17156.9849, "total_area_m2": 148143.5183, "severity_score": 0.2263, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 11938.0273, "total_area_m2": 11938.0273, "severity_score": 0.67, "dominant_damage": "major-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 31, "severe_count": 8, "severe_pct": 25.81, "severe_area_m2": 24038.904, "total_area_m2": 68721.3232, "severity_score": 0.2261, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 80, "severe_count": 2, "severe_pct": 2.5, "severe_area_m2": 2856.5987, "total_area_m2": 189604.2363, "severity_score": 0.025, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 11938.0273, "total_area_m2": 11938.0273, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 57, "severe_count": 3, "severe_pct": 5.26, "severe_area_m2": 17156.9849, "total_area_m2": 148143.5183, "severity_score": 0.2263, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000467_0553", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000467", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G30", "severity_score": 0.67, "building_count": 3, "severe_count": 3, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 3, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G30", "row": 3, "col": 0, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 11938.0273, "total_area_m2": 11938.0273, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 11, "severe_count": 7, "severe_pct": 63.64, "severe_area_m2": 23582.186, "total_area_m2": 30866.3904, "severity_score": 0.4564, "dominant_damage": "major-damage"}, {"cell_id": "G32", "row": 3, "col": 2, "building_count": 25, "severe_count": 2, "severe_pct": 8.0, "severe_area_m2": 10062.5328, "total_area_m2": 59426.2535, "severity_score": 0.3572, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000467_0554", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000467", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 16, "severe_footprint_m2": 55990.5149, "major_damage_area_m2": 55173.6552, "destroyed_area_m2": 816.8597, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000467_0555", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000467", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 16, "max_severe_in_cell": 8, "concentration_score": 0.5, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 8, "Q1": 2, "Q2": 3, "Q3": 3}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000467_0556", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000467", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 171, "classified_buildings": 169, "counts": {"no-damage": 115, "minor-damage": 38, "major-damage": 14, "destroyed": 2, "unclassified": 2}, "severe_count": 16, "severe_pct": 9.47, "damaged_pct": 31.95}, "area_summary": {"total_footprint_m2": 418407.1051, "severe_footprint_m2": 55990.5149, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 16}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000102_0557", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000102", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 69, "minor-damage": 36, "major-damage": 15, "destroyed": 2, "unclassified": 1}, "total_buildings": 123, "classified_buildings": 122, "damaged_count": 53, "severe_count": 17, "damaged_pct": 43.44, "severe_pct": 13.93, "severe_footprint_m2": 74126.6341, "total_footprint_m2": 619010.2753, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000102_0558", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000102", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 8, "severe_footprint_m2": 28124.7576, "severe_pct_in_quadrant": 15.38, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 52, "severe_count": 8, "severe_pct": 15.38, "severe_area_m2": 28124.7576, "total_area_m2": 189983.4152, "severity_score": 0.1983, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 38, "severe_count": 4, "severe_pct": 10.53, "severe_area_m2": 21836.7091, "total_area_m2": 148666.8252, "severity_score": 0.1921, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 15, "severe_count": 3, "severe_pct": 20.0, "severe_area_m2": 14290.2794, "total_area_m2": 130019.9797, "severity_score": 0.266, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 38, "severe_count": 4, "severe_pct": 10.53, "severe_area_m2": 21836.7091, "total_area_m2": 148666.8252, "severity_score": 0.1921, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 52, "severe_count": 8, "severe_pct": 15.38, "severe_area_m2": 28124.7576, "total_area_m2": 189983.4152, "severity_score": 0.1983, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 15, "severe_count": 3, "severe_pct": 20.0, "severe_area_m2": 14290.2794, "total_area_m2": 130019.9797, "severity_score": 0.266, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 18, "severe_count": 2, "severe_pct": 11.11, "severe_area_m2": 9874.888, "total_area_m2": 150340.0552, "severity_score": 0.1294, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000102_0559", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000102", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G22", "severity_score": 0.5567, "building_count": 3, "severe_count": 2, "class_histogram": {"no_damage": 1, "minor_damage": 0, "major_damage": 1, "destroyed": 1, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G22", "row": 2, "col": 2, "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 9874.888, "total_area_m2": 10756.3131, "severity_score": 0.5567, "dominant_damage": "no-damage"}, {"cell_id": "G20", "row": 2, "col": 0, "building_count": 5, "severe_count": 2, "severe_pct": 40.0, "severe_area_m2": 9466.4122, "total_area_m2": 23179.2899, "severity_score": 0.4, "dominant_damage": "minor-damage"}, {"cell_id": "G30", "row": 3, "col": 0, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1428.9134, "severity_score": 0.33, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-michael_00000102_0560", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000102", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "From the post-disaster wind scene (hurricane-michael), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 17, "severe_footprint_m2": 74126.6341, "major_damage_area_m2": 62641.8844, "destroyed_area_m2": 11484.7497, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-michael_00000102_0561", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000102", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 17, "max_severe_in_cell": 8, "concentration_score": 0.4706, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 4, "Q1": 8, "Q2": 3, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000102_0562", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000102", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 123, "classified_buildings": 122, "counts": {"no-damage": 69, "minor-damage": 36, "major-damage": 15, "destroyed": 2, "unclassified": 1}, "severe_count": 17, "severe_pct": 13.93, "damaged_pct": 43.44}, "area_summary": {"total_footprint_m2": 619010.2753, "severe_footprint_m2": 74126.6341, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.4706, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 18413, "polygon_severe_count": 17, "pixel_histogram": {"0": 896757, "1": 94242, "2": 39164, "3": 15603, "4": 2810, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 62.0752, "2": 25.7965, "3": 10.2774, "4": 1.8509, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000589_0569", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000589", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 17, "minor-damage": 6, "major-damage": 1, "destroyed": 0, "unclassified": 4}, "total_buildings": 28, "classified_buildings": 24, "damaged_count": 7, "severe_count": 1, "damaged_pct": 29.17, "severe_pct": 4.17, "severe_footprint_m2": 3355.6954, "total_footprint_m2": 72870.7446, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000589_0570", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000589", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 1, "severe_footprint_m2": 3355.6954, "severe_pct_in_quadrant": 3.57, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 28, "severe_count": 1, "severe_pct": 3.57, "severe_area_m2": 3355.6954, "total_area_m2": 72870.7446, "severity_score": 0.0946, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 28, "severe_count": 1, "severe_pct": 3.57, "severe_area_m2": 3355.6954, "total_area_m2": 72870.7446, "severity_score": 0.0946, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000589_0571", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000589", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G03", "severity_score": 0.0946, "building_count": 28, "severe_count": 1, "class_histogram": {"no_damage": 17, "minor_damage": 6, "major_damage": 1, "destroyed": 0, "unclassified": 4}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G03", "row": 0, "col": 3, "building_count": 28, "severe_count": 1, "severe_pct": 3.57, "severe_area_m2": 3355.6954, "total_area_m2": 72870.7446, "severity_score": 0.0946, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000589_0572", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000589", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 1, "severe_footprint_m2": 3355.6954, "major_damage_area_m2": 3355.6954, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000589_0573", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000589", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 1, "max_severe_in_cell": 1, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 1, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000589_0574", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000589", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 28, "classified_buildings": 24, "counts": {"no-damage": 17, "minor-damage": 6, "major-damage": 1, "destroyed": 0, "unclassified": 4}, "severe_count": 1, "severe_pct": 4.17, "damaged_pct": 29.17}, "area_summary": {"total_footprint_m2": 72870.7446, "severe_footprint_m2": 3355.6954, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 1}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_midwest-flooding_00000239_0575", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000239", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (midwest-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 12, "minor-damage": 1, "major-damage": 2, "destroyed": 0, "unclassified": 0}, "total_buildings": 15, "classified_buildings": 15, "damaged_count": 3, "severe_count": 2, "damaged_pct": 20.0, "severe_pct": 13.33, "severe_footprint_m2": 3725.6746, "total_footprint_m2": 22945.2298, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_midwest-flooding_00000239_0576", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000239", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (midwest-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 2, "severe_footprint_m2": 3725.6746, "severe_pct_in_quadrant": 50.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 2, "severe_pct": 50.0, "severe_area_m2": 3725.6746, "total_area_m2": 7540.919, "severity_score": 0.4175, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 15404.3108, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 2, "severe_pct": 50.0, "severe_area_m2": 3725.6746, "total_area_m2": 7540.919, "severity_score": 0.4175, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 15404.3108, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_midwest-flooding_00000239_0577", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000239", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 0.4467, "building_count": 3, "severe_count": 2, "class_histogram": {"no_damage": 1, "minor_damage": 0, "major_damage": 2, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 3725.6746, "total_area_m2": 6817.1388, "severity_score": 0.4467, "dominant_damage": "major-damage"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 723.7802, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_midwest-flooding_00000239_0578", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000239", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (midwest-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 2, "severe_footprint_m2": 3725.6746, "major_damage_area_m2": 3725.6746, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_midwest-flooding_00000239_0579", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000239", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 2, "max_severe_in_cell": 2, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 2, "Q1": 0, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_midwest-flooding_00000239_0580", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000239", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (midwest-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "midwest-flooding", "disaster_type": "flooding", "capture_date": "2019-05-30T17:35:22.000Z", "gsd": 1.3587263, "damage_summary": {"total_buildings": 15, "classified_buildings": 15, "counts": {"no-damage": 12, "minor-damage": 1, "major-damage": 2, "destroyed": 0, "unclassified": 0}, "severe_count": 2, "severe_pct": 13.33, "damaged_pct": 20.0}, "area_summary": {"total_footprint_m2": 22945.2298, "severe_footprint_m2": 3725.6746, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 2109, "polygon_severe_count": 2, "pixel_histogram": {"0": 1035433, "1": 10599, "2": 435, "3": 2109, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 80.6437, "2": 3.3097, "3": 16.0466, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00000698_0581", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000698", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 11, "minor-damage": 0, "major-damage": 1, "destroyed": 0, "unclassified": 0}, "total_buildings": 12, "classified_buildings": 12, "damaged_count": 1, "severe_count": 1, "damaged_pct": 8.33, "severe_pct": 8.33, "severe_footprint_m2": 2127.463, "total_footprint_m2": 33089.584, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00000698_0582", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000698", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 1, "severe_footprint_m2": 2127.463, "severe_pct_in_quadrant": 20.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 5, "severe_count": 1, "severe_pct": 20.0, "severe_area_m2": 2127.463, "total_area_m2": 9016.9276, "severity_score": 0.134, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 24072.6564, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 24072.6564, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 1, "severe_pct": 20.0, "severe_area_m2": 2127.463, "total_area_m2": 9016.9276, "severity_score": 0.134, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00000698_0583", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000698", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G12", "severity_score": 0.2233, "building_count": 3, "severe_count": 1, "class_histogram": {"no_damage": 2, "minor_damage": 0, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G12", "row": 1, "col": 2, "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 2127.463, "total_area_m2": 6646.6228, "severity_score": 0.2233, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_portugal-wildfire_00000698_0584", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000698", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "From the post-disaster fire scene (portugal-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 1, "severe_footprint_m2": 2127.463, "major_damage_area_m2": 2127.463, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_portugal-wildfire_00000698_0585", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000698", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 1, "max_severe_in_cell": 1, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 1, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00000698_0586", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000698", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 12, "classified_buildings": 12, "counts": {"no-damage": 11, "minor-damage": 0, "major-damage": 1, "destroyed": 0, "unclassified": 0}, "severe_count": 1, "severe_pct": 8.33, "damaged_pct": 8.33}, "area_summary": {"total_footprint_m2": 33089.584, "severe_footprint_m2": 2127.463, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 1}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_joplin-tornado_00000044_0587", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000044", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a wind event (joplin-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 76, "minor-damage": 21, "major-damage": 2, "destroyed": 0, "unclassified": 18}, "total_buildings": 117, "classified_buildings": 99, "damaged_count": 23, "severe_count": 2, "damaged_pct": 23.23, "severe_pct": 2.02, "severe_footprint_m2": 23618.2394, "total_footprint_m2": 636400.9368, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_joplin-tornado_00000044_0588", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000044", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "For a post-disaster wind scene (joplin-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 2, "severe_footprint_m2": 23618.2394, "severe_pct_in_quadrant": 16.67, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 12, "severe_count": 2, "severe_pct": 16.67, "severe_area_m2": 23618.2394, "total_area_m2": 101289.7843, "severity_score": 0.2217, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 47, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 182372.8393, "severity_score": 0.0562, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 35, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 114355.158, "severity_score": 0.0377, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 47, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 182372.8393, "severity_score": 0.0562, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 35, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 114355.158, "severity_score": 0.0377, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 23, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 238383.1552, "severity_score": 0.0717, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 12, "severe_count": 2, "severe_pct": 16.67, "severe_area_m2": 23618.2394, "total_area_m2": 101289.7843, "severity_score": 0.2217, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_joplin-tornado_00000044_0589", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000044", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G32", "severity_score": 0.415, "building_count": 4, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 3, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G32", "row": 3, "col": 2, "building_count": 4, "severe_count": 1, "severe_pct": 25.0, "severe_area_m2": 7245.4924, "total_area_m2": 64028.6833, "severity_score": 0.415, "dominant_damage": "minor-damage"}, {"cell_id": "G33", "row": 3, "col": 3, "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 16372.747, "total_area_m2": 25962.0419, "severity_score": 0.3333, "dominant_damage": "minor-damage"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 29440.6724, "severity_score": 0.099, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_joplin-tornado_00000044_0590", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000044", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "From the post-disaster wind scene (joplin-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 2, "severe_footprint_m2": 23618.2394, "major_damage_area_m2": 23618.2394, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_joplin-tornado_00000044_0591", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000044", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 2, "max_severe_in_cell": 2, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 0, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_joplin-tornado_00000044_0592", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000044", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (joplin-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "joplin-tornado", "disaster_type": "wind", "capture_date": "2011-05-29T17:29:25.433Z", "gsd": 2.35255861282349, "damage_summary": {"total_buildings": 117, "classified_buildings": 99, "counts": {"no-damage": 76, "minor-damage": 21, "major-damage": 2, "destroyed": 0, "unclassified": 18}, "severe_count": 2, "severe_pct": 2.02, "damaged_pct": 23.23}, "area_summary": {"total_footprint_m2": 636400.9368, "severe_footprint_m2": 23618.2394, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 2}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_woolsey-fire_00000115_0593", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000115", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a fire event (woolsey-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 25, "minor-damage": 0, "major-damage": 2, "destroyed": 0, "unclassified": 0}, "total_buildings": 27, "classified_buildings": 27, "damaged_count": 2, "severe_count": 2, "damaged_pct": 7.41, "severe_pct": 7.41, "severe_footprint_m2": 1740.987, "total_footprint_m2": 72751.1399, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_woolsey-fire_00000115_0594", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000115", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "For a post-disaster fire scene (woolsey-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 2, "severe_footprint_m2": 1740.987, "severe_pct_in_quadrant": 16.67, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 12, "severe_count": 2, "severe_pct": 16.67, "severe_area_m2": 1740.987, "total_area_m2": 23135.5272, "severity_score": 0.1117, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 30548.1277, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 15656.7722, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 30548.1277, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 15656.7722, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 12, "severe_count": 2, "severe_pct": 16.67, "severe_area_m2": 1740.987, "total_area_m2": 23135.5272, "severity_score": 0.1117, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 3410.7128, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_woolsey-fire_00000115_0595", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000115", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G31", "severity_score": 0.2233, "building_count": 6, "severe_count": 2, "class_histogram": {"no_damage": 4, "minor_damage": 0, "major_damage": 2, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G31", "row": 3, "col": 1, "building_count": 6, "severe_count": 2, "severe_pct": 33.33, "severe_area_m2": 1740.987, "total_area_m2": 9680.294, "severity_score": 0.2233, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 605.9286, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 5843.2223, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_woolsey-fire_00000115_0596", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000115", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "From the post-disaster fire scene (woolsey-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 2, "severe_footprint_m2": 1740.987, "major_damage_area_m2": 1740.987, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_woolsey-fire_00000115_0597", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000115", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 2, "max_severe_in_cell": 2, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 2, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_woolsey-fire_00000115_0598", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000115", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (woolsey-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "woolsey-fire", "disaster_type": "fire", "capture_date": "2018-11-18T18:44:07.000Z", "gsd": 1.7726423000000002, "damage_summary": {"total_buildings": 27, "classified_buildings": 27, "counts": {"no-damage": 25, "minor-damage": 0, "major-damage": 2, "destroyed": 0, "unclassified": 0}, "severe_count": 2, "severe_pct": 7.41, "damaged_pct": 7.41}, "area_summary": {"total_footprint_m2": 72751.1399, "severe_footprint_m2": 1740.987, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 2}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-harvey_00000410_0599", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000410", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (hurricane-harvey) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 19, "minor-damage": 24, "major-damage": 30, "destroyed": 0, "unclassified": 0}, "total_buildings": 73, "classified_buildings": 73, "damaged_count": 54, "severe_count": 30, "damaged_pct": 73.97, "severe_pct": 41.1, "severe_footprint_m2": 658095.529, "total_footprint_m2": 1905106.2072, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-harvey_00000410_0600", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000410", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (hurricane-harvey), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 26, "severe_footprint_m2": 614172.9397, "severe_pct_in_quadrant": 72.22, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 36, "severe_count": 26, "severe_pct": 72.22, "severe_area_m2": 614172.9397, "total_area_m2": 909878.2568, "severity_score": 0.5664, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 13, "severe_count": 3, "severe_pct": 23.08, "severe_area_m2": 41727.8621, "total_area_m2": 484443.6662, "severity_score": 0.2562, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 12, "severe_count": 1, "severe_pct": 8.33, "severe_area_m2": 2194.7272, "total_area_m2": 308304.3856, "severity_score": 0.3033, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 36, "severe_count": 26, "severe_pct": 72.22, "severe_area_m2": 614172.9397, "total_area_m2": 909878.2568, "severity_score": 0.5664, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 13, "severe_count": 3, "severe_pct": 23.08, "severe_area_m2": 41727.8621, "total_area_m2": 484443.6662, "severity_score": 0.2562, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 12, "severe_count": 1, "severe_pct": 8.33, "severe_area_m2": 2194.7272, "total_area_m2": 308304.3856, "severity_score": 0.3033, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 202479.8986, "severity_score": 0.055, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-harvey_00000410_0601", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000410", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.67, "building_count": 11, "severe_count": 11, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 11, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 11, "severe_count": 11, "severe_pct": 100.0, "severe_area_m2": 250539.0475, "total_area_m2": 250539.0475, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 178109.8786, "total_area_m2": 178109.8786, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 8, "severe_count": 7, "severe_pct": 87.5, "severe_area_m2": 158345.2494, "total_area_m2": 235977.5621, "severity_score": 0.6275, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-harvey_00000410_0602", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000410", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (hurricane-harvey), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 30, "severe_footprint_m2": 658095.529, "major_damage_area_m2": 658095.529, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-harvey_00000410_0603", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000410", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 30, "max_severe_in_cell": 26, "concentration_score": 0.8667, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 26, "Q1": 3, "Q2": 1, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-harvey_00000410_0604", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000410", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-harvey). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-harvey", "disaster_type": "flooding", "capture_date": "2017-08-31T17:38:50.685Z", "gsd": 3.024613, "damage_summary": {"total_buildings": 73, "classified_buildings": 73, "counts": {"no-damage": 19, "minor-damage": 24, "major-damage": 30, "destroyed": 0, "unclassified": 0}, "severe_count": 30, "severe_pct": 41.1, "damaged_pct": 73.97}, "area_summary": {"total_footprint_m2": 1905106.2072, "severe_footprint_m2": 658095.529, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.8667, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 75089, "polygon_severe_count": 30, "pixel_histogram": {"0": 831082, "1": 64355, "2": 78050, "3": 75089, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 29.5893, "2": 35.886, "3": 34.5246, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000536_0605", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000536", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 9, "minor-damage": 2, "major-damage": 9, "destroyed": 2, "unclassified": 0}, "total_buildings": 22, "classified_buildings": 22, "damaged_count": 13, "severe_count": 11, "damaged_pct": 59.09, "severe_pct": 50.0, "severe_footprint_m2": 46357.425, "total_footprint_m2": 69452.3386, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000536_0606", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000536", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 9, "severe_footprint_m2": 43077.8072, "severe_pct_in_quadrant": 45.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 20, "severe_count": 9, "severe_pct": 45.0, "severe_area_m2": 43077.8072, "total_area_m2": 66172.7208, "severity_score": 0.3675, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 3279.6178, "total_area_m2": 3279.6178, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 20, "severe_count": 9, "severe_pct": 45.0, "severe_area_m2": 43077.8072, "total_area_m2": 66172.7208, "severity_score": 0.3675, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 3279.6178, "total_area_m2": 3279.6178, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000536_0607", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000536", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G10", "severity_score": 0.835, "building_count": 4, "severe_count": 4, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 2, "destroyed": 2, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G10", "row": 1, "col": 0, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 6498.0579, "total_area_m2": 6498.0579, "severity_score": 0.835, "dominant_damage": "major-damage"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 16541.8931, "total_area_m2": 16541.8931, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G20", "row": 2, "col": 0, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 479.0823, "total_area_m2": 479.0823, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000536_0608", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000536", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 11, "severe_footprint_m2": 46357.425, "major_damage_area_m2": 45713.4688, "destroyed_area_m2": 643.9562, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000536_0609", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000536", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 11, "max_severe_in_cell": 9, "concentration_score": 0.8182, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 9, "Q1": 0, "Q2": 2, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000536_0610", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000536", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 22, "classified_buildings": 22, "counts": {"no-damage": 9, "minor-damage": 2, "major-damage": 9, "destroyed": 2, "unclassified": 0}, "severe_count": 11, "severe_pct": 50.0, "damaged_pct": 59.09}, "area_summary": {"total_footprint_m2": 69452.3386, "severe_footprint_m2": 46357.425, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.8182, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 11}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-harvey_00000269_0623", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000269", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (hurricane-harvey) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 0, "major-damage": 15, "destroyed": 0, "unclassified": 0}, "total_buildings": 15, "classified_buildings": 15, "damaged_count": 15, "severe_count": 15, "damaged_pct": 100.0, "severe_pct": 100.0, "severe_footprint_m2": 59808.9531, "total_footprint_m2": 59808.9531, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-harvey_00000269_0624", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000269", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (hurricane-harvey), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 8, "severe_footprint_m2": 31303.2334, "severe_pct_in_quadrant": 100.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 31303.2334, "total_area_m2": 31303.2334, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 7, "severe_count": 7, "severe_pct": 100.0, "severe_area_m2": 28505.7197, "total_area_m2": 28505.7197, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 31303.2334, "total_area_m2": 31303.2334, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 7, "severe_count": 7, "severe_pct": 100.0, "severe_area_m2": 28505.7197, "total_area_m2": 28505.7197, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-harvey_00000269_0625", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000269", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G21", "severity_score": 0.67, "building_count": 8, "severe_count": 8, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 8, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G21", "row": 2, "col": 1, "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 31303.2334, "total_area_m2": 31303.2334, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G22", "row": 2, "col": 2, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 8619.3754, "total_area_m2": 8619.3754, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G32", "row": 3, "col": 2, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 19886.3443, "total_area_m2": 19886.3443, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-harvey_00000269_0626", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000269", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (hurricane-harvey), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 15, "severe_footprint_m2": 59808.9531, "major_damage_area_m2": 59808.9531, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-harvey_00000269_0627", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000269", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 15, "max_severe_in_cell": 8, "concentration_score": 0.5333, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 8, "Q3": 7}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-harvey_00000269_0628", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000269", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-harvey). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-harvey", "disaster_type": "flooding", "capture_date": "2017-08-31T17:38:50.685Z", "gsd": 3.024613, "damage_summary": {"total_buildings": 15, "classified_buildings": 15, "counts": {"no-damage": 0, "minor-damage": 0, "major-damage": 15, "destroyed": 0, "unclassified": 0}, "severe_count": 15, "severe_pct": 100.0, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 59808.9531, "severe_footprint_m2": 59808.9531, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5333, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 7061, "polygon_severe_count": 15, "pixel_histogram": {"0": 1041515, "1": 0, "2": 0, "3": 7061, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 100.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000240_0629", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000240", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 0, "major-damage": 30, "destroyed": 3, "unclassified": 3}, "total_buildings": 36, "classified_buildings": 33, "damaged_count": 33, "severe_count": 33, "damaged_pct": 100.0, "severe_pct": 100.0, "severe_footprint_m2": 90168.3023, "total_footprint_m2": 92455.9742, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000240_0630", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000240", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 22, "severe_footprint_m2": 51353.0231, "severe_pct_in_quadrant": 88.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 25, "severe_count": 22, "severe_pct": 88.0, "severe_area_m2": 51353.0231, "total_area_m2": 53640.695, "severity_score": 0.616, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 7, "severe_pct": 100.0, "severe_area_m2": 22915.5047, "total_area_m2": 22915.5047, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 8299.7057, "total_area_m2": 8299.7057, "severity_score": 0.67, "dominant_damage": "major-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 7600.0688, "total_area_m2": 7600.0688, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 8299.7057, "total_area_m2": 8299.7057, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 7, "severe_pct": 100.0, "severe_area_m2": 22915.5047, "total_area_m2": 22915.5047, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 25, "severe_count": 22, "severe_pct": 88.0, "severe_area_m2": 51353.0231, "total_area_m2": 53640.695, "severity_score": 0.616, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000240_0631", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000240", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G11", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G11", "row": 1, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 7600.0688, "total_area_m2": 7600.0688, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G32", "row": 3, "col": 2, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 5697.4169, "total_area_m2": 5697.4169, "severity_score": 0.7525, "dominant_damage": "major-damage"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 8299.7057, "total_area_m2": 8299.7057, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000240_0632", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000240", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 33, "severe_footprint_m2": 90168.3023, "major_damage_area_m2": 79771.766, "destroyed_area_m2": 10396.5363, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000240_0633", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000240", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 33, "max_severe_in_cell": 22, "concentration_score": 0.6667, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 1, "Q1": 3, "Q2": 7, "Q3": 22}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000240_0634", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000240", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:17:02.085Z", "gsd": 2.9729018, "damage_summary": {"total_buildings": 36, "classified_buildings": 33, "counts": {"no-damage": 0, "minor-damage": 0, "major-damage": 30, "destroyed": 3, "unclassified": 3}, "severe_count": 33, "severe_pct": 100.0, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 92455.9742, "severe_footprint_m2": 90168.3023, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6667, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 33}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-florence_00000389_0635", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000389", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (hurricane-florence) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 0, "major-damage": 27, "destroyed": 0, "unclassified": 17}, "total_buildings": 44, "classified_buildings": 27, "damaged_count": 27, "severe_count": 27, "damaged_pct": 100.0, "severe_pct": 100.0, "severe_footprint_m2": 69751.2643, "total_footprint_m2": 113824.9448, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-florence_00000389_0636", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000389", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (hurricane-florence), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 12, "severe_footprint_m2": 28862.9341, "severe_pct_in_quadrant": 100.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 12, "severe_count": 12, "severe_pct": 100.0, "severe_area_m2": 28862.9341, "total_area_m2": 28862.9341, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q0", "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 24142.3484, "total_area_m2": 24142.3484, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 18, "severe_count": 7, "severe_pct": 38.89, "severe_area_m2": 16745.9818, "total_area_m2": 36050.5907, "severity_score": 0.2606, "dominant_damage": "major-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 24142.3484, "total_area_m2": 24142.3484, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 12, "severe_count": 12, "severe_pct": 100.0, "severe_area_m2": 28862.9341, "total_area_m2": 28862.9341, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 18, "severe_count": 7, "severe_pct": 38.89, "severe_area_m2": 16745.9818, "total_area_m2": 36050.5907, "severity_score": 0.2606, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 24769.0716, "severity_score": 0.0, "dominant_damage": "unclassified"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-florence_00000389_0637", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000389", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 0.67, "building_count": 5, "severe_count": 5, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 5, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 5, "severe_count": 5, "severe_pct": 100.0, "severe_area_m2": 14504.08, "total_area_m2": 14504.08, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G03", "row": 0, "col": 3, "building_count": 7, "severe_count": 7, "severe_pct": 100.0, "severe_area_m2": 14358.8541, "total_area_m2": 14358.8541, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 14772.5395, "total_area_m2": 14772.5395, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-florence_00000389_0638", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000389", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (hurricane-florence), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 27, "severe_footprint_m2": 69751.2643, "major_damage_area_m2": 69751.2643, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-florence_00000389_0639", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000389", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 27, "max_severe_in_cell": 12, "concentration_score": 0.4444, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 8, "Q1": 12, "Q2": 7, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-florence_00000389_0640", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000389", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-florence). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-florence", "disaster_type": "flooding", "capture_date": "2018-09-20T16:04:41.000Z", "gsd": 2.0916247, "damage_summary": {"total_buildings": 44, "classified_buildings": 27, "counts": {"no-damage": 0, "minor-damage": 0, "major-damage": 27, "destroyed": 0, "unclassified": 17}, "severe_count": 27, "severe_pct": 100.0, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 113824.9448, "severe_footprint_m2": 69751.2643, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.4444, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 17210, "polygon_severe_count": 27, "pixel_histogram": {"0": 1031366, "1": 0, "2": 0, "3": 17210, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 100.0, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000157_0647", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000157", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 11, "minor-damage": 18, "major-damage": 4, "destroyed": 1, "unclassified": 0}, "total_buildings": 34, "classified_buildings": 34, "damaged_count": 23, "severe_count": 5, "damaged_pct": 67.65, "severe_pct": 14.71, "severe_footprint_m2": 178982.3323, "total_footprint_m2": 597157.9879, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000157_0648", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000157", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 5, "severe_footprint_m2": 178982.3323, "severe_pct_in_quadrant": 22.73, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 22, "severe_count": 5, "severe_pct": 22.73, "severe_area_m2": 178982.3323, "total_area_m2": 304153.6833, "severity_score": 0.2873, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 229981.5111, "severity_score": 0.297, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 229981.5111, "severity_score": 0.297, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 22, "severe_count": 5, "severe_pct": 22.73, "severe_area_m2": 178982.3323, "total_area_m2": 304153.6833, "severity_score": 0.2873, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 63022.7935, "severity_score": 0.165, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000157_0649", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000157", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G12", "severity_score": 0.67, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G12", "row": 1, "col": 2, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 171227.4536, "total_area_m2": 171227.4536, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 7, "severe_count": 2, "severe_pct": 28.57, "severe_area_m2": 3015.2456, "total_area_m2": 46504.5654, "severity_score": 0.3329, "dominant_damage": "minor-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 53406.4392, "severity_score": 0.33, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-michael_00000157_0650", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000157", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "From the post-disaster wind scene (hurricane-michael), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 5, "severe_footprint_m2": 178982.3323, "major_damage_area_m2": 175572.9854, "destroyed_area_m2": 3409.3469, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-michael_00000157_0651", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000157", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 5, "max_severe_in_cell": 5, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 5, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000157_0652", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000157", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 34, "classified_buildings": 34, "counts": {"no-damage": 11, "minor-damage": 18, "major-damage": 4, "destroyed": 1, "unclassified": 0}, "severe_count": 5, "severe_pct": 14.71, "damaged_pct": 67.65}, "area_summary": {"total_footprint_m2": 597157.9879, "severe_footprint_m2": 178982.3323, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 42339, "polygon_severe_count": 5, "pixel_histogram": {"0": 905546, "1": 10927, "2": 89764, "3": 41500, "4": 839, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 7.6397, "2": 62.7589, "3": 29.0149, "4": 0.5866, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-harvey_00000071_0653", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000071", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (hurricane-harvey) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 0, "major-damage": 14, "destroyed": 2, "unclassified": 0}, "total_buildings": 16, "classified_buildings": 16, "damaged_count": 16, "severe_count": 16, "damaged_pct": 100.0, "severe_pct": 100.0, "severe_footprint_m2": 100548.8365, "total_footprint_m2": 100548.8365, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-harvey_00000071_0654", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000071", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (hurricane-harvey), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 2, "severe_footprint_m2": 42359.3969, "severe_pct_in_quadrant": 100.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 42359.3969, "total_area_m2": 42359.3969, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q0", "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 29205.3987, "total_area_m2": 29205.3987, "severity_score": 0.7525, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 16390.1223, "total_area_m2": 16390.1223, "severity_score": 0.67, "dominant_damage": "major-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 29205.3987, "total_area_m2": 29205.3987, "severity_score": 0.7525, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 42359.3969, "total_area_m2": 42359.3969, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 12593.9186, "total_area_m2": 12593.9186, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 16390.1223, "total_area_m2": 16390.1223, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-harvey_00000071_0655", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000071", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G10", "severity_score": 0.7643, "building_count": 7, "severe_count": 7, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 5, "destroyed": 2, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G10", "row": 1, "col": 0, "building_count": 7, "severe_count": 7, "severe_pct": 100.0, "severe_area_m2": 22888.4032, "total_area_m2": 22888.4032, "severity_score": 0.7643, "dominant_damage": "major-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 6316.9955, "total_area_m2": 6316.9955, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 18858.6531, "total_area_m2": 18858.6531, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-harvey_00000071_0656", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000071", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (hurricane-harvey), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 16, "severe_footprint_m2": 100548.8365, "major_damage_area_m2": 98011.0614, "destroyed_area_m2": 2537.7751, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-harvey_00000071_0657", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000071", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 16, "max_severe_in_cell": 8, "concentration_score": 0.5, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 8, "Q1": 2, "Q2": 4, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-harvey_00000071_0658", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000071", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-harvey). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-harvey", "disaster_type": "flooding", "capture_date": "2017-08-31T17:38:50.685Z", "gsd": 3.024613, "damage_summary": {"total_buildings": 16, "classified_buildings": 16, "counts": {"no-damage": 0, "minor-damage": 0, "major-damage": 14, "destroyed": 2, "unclassified": 0}, "severe_count": 16, "severe_pct": 100.0, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 100548.8365, "severe_footprint_m2": 100548.8365, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 11672, "polygon_severe_count": 16, "pixel_histogram": {"0": 1036904, "1": 0, "2": 0, "3": 11354, "4": 318, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 97.2755, "4": 2.7245, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000597_0659", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000597", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 96, "minor-damage": 6, "major-damage": 2, "destroyed": 0, "unclassified": 3}, "total_buildings": 107, "classified_buildings": 104, "damaged_count": 8, "severe_count": 2, "damaged_pct": 7.69, "severe_pct": 1.92, "severe_footprint_m2": 2744.5802, "total_footprint_m2": 362678.3491, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000597_0660", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000597", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 1, "severe_footprint_m2": 2172.5272, "severe_pct_in_quadrant": 3.85, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 26, "severe_count": 1, "severe_pct": 3.85, "severe_area_m2": 2172.5272, "total_area_m2": 82433.8937, "severity_score": 0.1019, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 6, "severe_count": 1, "severe_pct": 16.67, "severe_area_m2": 572.053, "total_area_m2": 21282.6674, "severity_score": 0.1117, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 6, "severe_count": 1, "severe_pct": 16.67, "severe_area_m2": 572.053, "total_area_m2": 21282.6674, "severity_score": 0.1117, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 26, "severe_count": 1, "severe_pct": 3.85, "severe_area_m2": 2172.5272, "total_area_m2": 82433.8937, "severity_score": 0.1019, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 75, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 258961.788, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000597_0661", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000597", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G03", "severity_score": 0.67, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G03", "row": 0, "col": 3, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 572.053, "total_area_m2": 572.053, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G21", "row": 2, "col": 1, "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 41396.6397, "severity_score": 0.2829, "dominant_damage": "minor-damage"}, {"cell_id": "G31", "row": 3, "col": 1, "building_count": 19, "severe_count": 1, "severe_pct": 5.26, "severe_area_m2": 2172.5272, "total_area_m2": 41037.254, "severity_score": 0.0353, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000597_0662", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000597", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 2, "severe_footprint_m2": 2744.5802, "major_damage_area_m2": 2744.5802, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000597_0663", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000597", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 2, "max_severe_in_cell": 1, "concentration_score": 0.5, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 1, "Q2": 1, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000597_0664", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000597", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 107, "classified_buildings": 104, "counts": {"no-damage": 96, "minor-damage": 6, "major-damage": 2, "destroyed": 0, "unclassified": 3}, "severe_count": 2, "severe_pct": 1.92, "damaged_pct": 7.69}, "area_summary": {"total_footprint_m2": 362678.3491, "severe_footprint_m2": 2744.5802, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 2}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-florence_00000153_0665", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000153", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (hurricane-florence) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 3, "major-damage": 13, "destroyed": 0, "unclassified": 0}, "total_buildings": 16, "classified_buildings": 16, "damaged_count": 16, "severe_count": 13, "damaged_pct": 100.0, "severe_pct": 81.25, "severe_footprint_m2": 41181.3601, "total_footprint_m2": 56622.6332, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-florence_00000153_0666", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000153", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (hurricane-florence), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 9, "severe_footprint_m2": 20563.3536, "severe_pct_in_quadrant": 90.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 10, "severe_count": 9, "severe_pct": 90.0, "severe_area_m2": 20563.3536, "total_area_m2": 26856.2208, "severity_score": 0.636, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 12205.577, "total_area_m2": 17786.8082, "severity_score": 0.5567, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 8412.4295, "total_area_m2": 11979.6042, "severity_score": 0.5567, "dominant_damage": "major-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 10, "severe_count": 9, "severe_pct": 90.0, "severe_area_m2": 20563.3536, "total_area_m2": 26856.2208, "severity_score": 0.636, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 8412.4295, "total_area_m2": 11979.6042, "severity_score": 0.5567, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 12205.577, "total_area_m2": 17786.8082, "severity_score": 0.5567, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-florence_00000153_0667", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000153", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.67, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 2, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 3906.8262, "total_area_m2": 3906.8262, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 8233.4761, "total_area_m2": 8233.4761, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G33", "row": 3, "col": 3, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 9618.4694, "total_area_m2": 9618.4694, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-florence_00000153_0668", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000153", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (hurricane-florence), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 13, "severe_footprint_m2": 41181.3601, "major_damage_area_m2": 41181.3601, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-florence_00000153_0669", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000153", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 13, "max_severe_in_cell": 9, "concentration_score": 0.6923, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 9, "Q1": 2, "Q2": 0, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-florence_00000153_0670", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000153", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-florence). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-florence", "disaster_type": "flooding", "capture_date": "2018-09-20T16:04:41.000Z", "gsd": 2.0916247, "damage_summary": {"total_buildings": 16, "classified_buildings": 16, "counts": {"no-damage": 0, "minor-damage": 3, "major-damage": 13, "destroyed": 0, "unclassified": 0}, "severe_count": 13, "severe_pct": 81.25, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 56622.6332, "severe_footprint_m2": 41181.3601, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6923, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 9934, "polygon_severe_count": 13, "pixel_histogram": {"0": 1034905, "1": 0, "2": 3737, "3": 9934, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 27.3352, "3": 72.6648, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000548_0671", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000548", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 11, "minor-damage": 6, "major-damage": 1, "destroyed": 0, "unclassified": 0}, "total_buildings": 18, "classified_buildings": 18, "damaged_count": 7, "severe_count": 1, "damaged_pct": 38.89, "severe_pct": 5.56, "severe_footprint_m2": 421116.2196, "total_footprint_m2": 767586.5247, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000548_0672", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000548", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 1, "severe_footprint_m2": 421116.2196, "severe_pct_in_quadrant": 50.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 421116.2196, "total_area_m2": 422027.6199, "severity_score": 0.335, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 24395.999, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 173533.959, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 421116.2196, "total_area_m2": 422027.6199, "severity_score": 0.335, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 24395.999, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 173533.959, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 147628.9468, "severity_score": 0.198, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000548_0673", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000548", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G10", "severity_score": 0.67, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G10", "row": 1, "col": 0, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 421116.2196, "total_area_m2": 421116.2196, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G32", "row": 3, "col": 2, "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 44118.3114, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "G33", "row": 3, "col": 3, "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 15291.5092, "severity_score": 0.33, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-michael_00000548_0674", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000548", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "From the post-disaster wind scene (hurricane-michael), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 1, "severe_footprint_m2": 421116.2196, "major_damage_area_m2": 421116.2196, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-michael_00000548_0675", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000548", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 1, "max_severe_in_cell": 1, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 1, "Q1": 0, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000548_0676", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000548", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 18, "classified_buildings": 18, "counts": {"no-damage": 11, "minor-damage": 6, "major-damage": 1, "destroyed": 0, "unclassified": 0}, "severe_count": 1, "severe_pct": 5.56, "damaged_pct": 38.89}, "area_summary": {"total_footprint_m2": 767586.5247, "severe_footprint_m2": 421116.2196, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 98743, "polygon_severe_count": 1, "pixel_histogram": {"0": 866708, "1": 68745, "2": 14380, "3": 98743, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 37.7994, "2": 7.9068, "3": 54.2938, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000049_0677", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000049", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 0, "major-damage": 10, "destroyed": 1, "unclassified": 0}, "total_buildings": 11, "classified_buildings": 11, "damaged_count": 11, "severe_count": 11, "damaged_pct": 100.0, "severe_pct": 100.0, "severe_footprint_m2": 36061.1812, "total_footprint_m2": 36061.1812, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000049_0678", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000049", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 9, "severe_footprint_m2": 32153.4004, "severe_pct_in_quadrant": 100.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 9, "severe_count": 9, "severe_pct": 100.0, "severe_area_m2": 32153.4004, "total_area_m2": 32153.4004, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 3907.7808, "total_area_m2": 3907.7808, "severity_score": 0.835, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 9, "severe_count": 9, "severe_pct": 100.0, "severe_area_m2": 32153.4004, "total_area_m2": 32153.4004, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 3907.7808, "total_area_m2": 3907.7808, "severity_score": 0.835, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000049_0679", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000049", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G21", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G21", "row": 2, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 812.2436, "total_area_m2": 812.2436, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 14192.3557, "total_area_m2": 14192.3557, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 5, "severe_count": 5, "severe_pct": 100.0, "severe_area_m2": 17961.0447, "total_area_m2": 17961.0447, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000049_0680", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000049", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 11, "severe_footprint_m2": 36061.1812, "major_damage_area_m2": 35248.9376, "destroyed_area_m2": 812.2436, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000049_0681", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000049", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 11, "max_severe_in_cell": 9, "concentration_score": 0.8182, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 9, "Q1": 0, "Q2": 2, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000049_0682", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000049", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 11, "classified_buildings": 11, "counts": {"no-damage": 0, "minor-damage": 0, "major-damage": 10, "destroyed": 1, "unclassified": 0}, "severe_count": 11, "severe_pct": 100.0, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 36061.1812, "severe_footprint_m2": 36061.1812, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.8182, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 11}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000398_0683", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000398", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 89, "minor-damage": 20, "major-damage": 6, "destroyed": 1, "unclassified": 2}, "total_buildings": 118, "classified_buildings": 116, "damaged_count": 27, "severe_count": 7, "damaged_pct": 23.28, "severe_pct": 6.03, "severe_footprint_m2": 16811.8934, "total_footprint_m2": 234499.5564, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000398_0684", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000398", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 6, "severe_footprint_m2": 13455.8296, "severe_pct_in_quadrant": 24.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 25, "severe_count": 6, "severe_pct": 24.0, "severe_area_m2": 13455.8296, "total_area_m2": 50066.1992, "severity_score": 0.174, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 18, "severe_count": 1, "severe_pct": 5.56, "severe_area_m2": 3356.0638, "total_area_m2": 22433.8749, "severity_score": 0.1106, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 35, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 70675.8524, "severity_score": 0.1509, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 18, "severe_count": 1, "severe_pct": 5.56, "severe_area_m2": 3356.0638, "total_area_m2": 22433.8749, "severity_score": 0.1106, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 35, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 70675.8524, "severity_score": 0.1509, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 40, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 91323.6299, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 25, "severe_count": 6, "severe_pct": 24.0, "severe_area_m2": 13455.8296, "total_area_m2": 50066.1992, "severity_score": 0.174, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000398_0685", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000398", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G33", "severity_score": 0.67, "building_count": 4, "severe_count": 4, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 4, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G33", "row": 3, "col": 3, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 9186.6783, "total_area_m2": 9186.6783, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 3356.0638, "total_area_m2": 5330.9218, "severity_score": 0.5, "dominant_damage": "minor-damage"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 4, "severe_count": 2, "severe_pct": 50.0, "severe_area_m2": 4269.1513, "total_area_m2": 7802.449, "severity_score": 0.4175, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000398_0686", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000398", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 7, "severe_footprint_m2": 16811.8934, "major_damage_area_m2": 15591.8522, "destroyed_area_m2": 1220.0412, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000398_0687", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000398", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 7, "max_severe_in_cell": 6, "concentration_score": 0.8571, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 1, "Q1": 0, "Q2": 0, "Q3": 6}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000398_0688", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000398", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 118, "classified_buildings": 116, "counts": {"no-damage": 89, "minor-damage": 20, "major-damage": 6, "destroyed": 1, "unclassified": 2}, "severe_count": 7, "severe_pct": 6.03, "damaged_pct": 23.28}, "area_summary": {"total_footprint_m2": 234499.5564, "severe_footprint_m2": 16811.8934, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.8571, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 7}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_palu-tsunami_00000090_0689", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000090", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a tsunami event (palu-tsunami) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 210, "minor-damage": 0, "major-damage": 2, "destroyed": 0, "unclassified": 2}, "total_buildings": 214, "classified_buildings": 212, "damaged_count": 2, "severe_count": 2, "damaged_pct": 0.94, "severe_pct": 0.94, "severe_footprint_m2": 2700.0226, "total_footprint_m2": 743959.5432, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_palu-tsunami_00000090_0690", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000090", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "major-heavy", "question": "For a post-disaster tsunami scene (palu-tsunami), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 2, "severe_footprint_m2": 2700.0226, "severe_pct_in_quadrant": 100.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 2700.0226, "total_area_m2": 2700.0226, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 98, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 365002.7804, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 37, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 102525.2824, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 2700.0226, "total_area_m2": 2700.0226, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 98, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 365002.7804, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 37, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 102525.2824, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 77, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 273731.4578, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_palu-tsunami_00000090_0691", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000090", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster tsunami satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G10", "severity_score": 0.67, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 2, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G10", "row": 1, "col": 0, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 2700.0226, "total_area_m2": 2700.0226, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_palu-tsunami_00000090_0692", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000090", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "major-heavy", "question": "From the post-disaster tsunami scene (palu-tsunami), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 2, "severe_footprint_m2": 2700.0226, "major_damage_area_m2": 2700.0226, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_palu-tsunami_00000090_0693", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000090", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster tsunami scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 2, "max_severe_in_cell": 2, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 2, "Q1": 0, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_palu-tsunami_00000090_0694", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000090", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster tsunami scene (palu-tsunami). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "palu-tsunami", "disaster_type": "tsunami", "capture_date": "2018-10-01T02:26:02.000Z", "gsd": 2.8247435, "damage_summary": {"total_buildings": 214, "classified_buildings": 212, "counts": {"no-damage": 210, "minor-damage": 0, "major-damage": 2, "destroyed": 0, "unclassified": 2}, "severe_count": 2, "severe_pct": 0.94, "damaged_pct": 0.94}, "area_summary": {"total_footprint_m2": 743959.5432, "severe_footprint_m2": 2700.0226, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 394, "polygon_severe_count": 2, "pixel_histogram": {"0": 948123, "1": 100059, "2": 0, "3": 394, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 99.6078, "2": 0.0, "3": 0.3922, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_tuscaloosa-tornado_00000278_0695", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000278", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a wind event (tuscaloosa-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 4, "minor-damage": 0, "major-damage": 1, "destroyed": 0, "unclassified": 5}, "total_buildings": 10, "classified_buildings": 5, "damaged_count": 1, "severe_count": 1, "damaged_pct": 20.0, "severe_pct": 20.0, "severe_footprint_m2": 3145.848, "total_footprint_m2": 12977.1241, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_tuscaloosa-tornado_00000278_0696", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000278", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "For a post-disaster wind scene (tuscaloosa-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 1, "severe_footprint_m2": 3145.848, "severe_pct_in_quadrant": 20.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 5, "severe_count": 1, "severe_pct": 20.0, "severe_area_m2": 3145.848, "total_area_m2": 7046.392, "severity_score": 0.134, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 5930.7321, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 5930.7321, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 5, "severe_count": 1, "severe_pct": 20.0, "severe_area_m2": 3145.848, "total_area_m2": 7046.392, "severity_score": 0.134, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_tuscaloosa-tornado_00000278_0697", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000278", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G21", "severity_score": 0.67, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G21", "row": 2, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 3145.848, "total_area_m2": 3145.848, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_tuscaloosa-tornado_00000278_0698", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000278", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "From the post-disaster wind scene (tuscaloosa-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 1, "severe_footprint_m2": 3145.848, "major_damage_area_m2": 3145.848, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_tuscaloosa-tornado_00000278_0699", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000278", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 1, "max_severe_in_cell": 1, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 1, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_tuscaloosa-tornado_00000278_0700", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000278", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (tuscaloosa-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "tuscaloosa-tornado", "disaster_type": "wind", "capture_date": "2011-05-19T16:48:01.085Z", "gsd": 1.85690224170685, "damage_summary": {"total_buildings": 10, "classified_buildings": 5, "counts": {"no-damage": 4, "minor-damage": 0, "major-damage": 1, "destroyed": 0, "unclassified": 5}, "severe_count": 1, "severe_pct": 20.0, "damaged_pct": 20.0}, "area_summary": {"total_footprint_m2": 12977.1241, "severe_footprint_m2": 3145.848, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 1}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-harvey_00000491_0701", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000491", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (hurricane-harvey) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 60, "major-damage": 114, "destroyed": 1, "unclassified": 2}, "total_buildings": 177, "classified_buildings": 175, "damaged_count": 175, "severe_count": 115, "damaged_pct": 100.0, "severe_pct": 65.71, "severe_footprint_m2": 1622768.118, "total_footprint_m2": 2428745.3844, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-harvey_00000491_0702", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000491", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (hurricane-harvey), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 25, "severe_footprint_m2": 589553.2842, "severe_pct_in_quadrant": 100.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 25, "severe_count": 25, "severe_pct": 100.0, "severe_area_m2": 589553.2842, "total_area_m2": 589553.2842, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 72, "severe_count": 42, "severe_pct": 58.33, "severe_area_m2": 465260.8068, "total_area_m2": 906642.517, "severity_score": 0.5238, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 72, "severe_count": 40, "severe_pct": 55.56, "severe_area_m2": 445945.7547, "total_area_m2": 810541.3109, "severity_score": 0.5189, "dominant_damage": "major-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 122008.2723, "total_area_m2": 122008.2723, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 72, "severe_count": 42, "severe_pct": 58.33, "severe_area_m2": 465260.8068, "total_area_m2": 906642.517, "severity_score": 0.5238, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 25, "severe_count": 25, "severe_pct": 100.0, "severe_area_m2": 589553.2842, "total_area_m2": 589553.2842, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 72, "severe_count": 40, "severe_pct": 55.56, "severe_area_m2": 445945.7547, "total_area_m2": 810541.3109, "severity_score": 0.5189, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-harvey_00000491_0703", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000491", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.67, "building_count": 3, "severe_count": 3, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 3, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 37110.3788, "total_area_m2": 37110.3788, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 5, "severe_count": 5, "severe_pct": 100.0, "severe_area_m2": 84897.8935, "total_area_m2": 84897.8935, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G20", "row": 2, "col": 0, "building_count": 11, "severe_count": 11, "severe_pct": 100.0, "severe_area_m2": 304789.1406, "total_area_m2": 304789.1406, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-harvey_00000491_0704", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000491", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (hurricane-harvey), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 115, "severe_footprint_m2": 1622768.118, "major_damage_area_m2": 1621078.8601, "destroyed_area_m2": 1689.2579, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-harvey_00000491_0705", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000491", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 115, "max_severe_in_cell": 42, "concentration_score": 0.3652, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 8, "Q1": 42, "Q2": 25, "Q3": 40}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-harvey_00000491_0706", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000491", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-harvey). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-harvey", "disaster_type": "flooding", "capture_date": "2017-08-31T17:38:50.685Z", "gsd": 3.024613, "damage_summary": {"total_buildings": 177, "classified_buildings": 175, "counts": {"no-damage": 0, "minor-damage": 60, "major-damage": 114, "destroyed": 1, "unclassified": 2}, "severe_count": 115, "severe_pct": 65.71, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 2428745.3844, "severe_footprint_m2": 1622768.118, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.3652, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 185774, "polygon_severe_count": 115, "pixel_histogram": {"0": 776838, "1": 0, "2": 85964, "3": 185560, "4": 214, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 31.6349, "3": 68.2864, "4": 0.0788, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000287_0713", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000287", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 50, "minor-damage": 31, "major-damage": 26, "destroyed": 0, "unclassified": 0}, "total_buildings": 107, "classified_buildings": 107, "damaged_count": 57, "severe_count": 26, "damaged_pct": 53.27, "severe_pct": 24.3, "severe_footprint_m2": 178917.4639, "total_footprint_m2": 484549.9102, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000287_0714", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000287", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 9, "severe_footprint_m2": 102243.6372, "severe_pct_in_quadrant": 40.91, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 22, "severe_count": 9, "severe_pct": 40.91, "severe_area_m2": 102243.6372, "total_area_m2": 149235.0811, "severity_score": 0.3191, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 35, "severe_count": 6, "severe_pct": 17.14, "severe_area_m2": 39297.7761, "total_area_m2": 168369.6447, "severity_score": 0.2751, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 38, "severe_count": 8, "severe_pct": 21.05, "severe_area_m2": 24529.7465, "total_area_m2": 121547.579, "severity_score": 0.2018, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 22, "severe_count": 9, "severe_pct": 40.91, "severe_area_m2": 102243.6372, "total_area_m2": 149235.0811, "severity_score": 0.3191, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 35, "severe_count": 6, "severe_pct": 17.14, "severe_area_m2": 39297.7761, "total_area_m2": 168369.6447, "severity_score": 0.2751, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 38, "severe_count": 8, "severe_pct": 21.05, "severe_area_m2": 24529.7465, "total_area_m2": 121547.579, "severity_score": 0.2018, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 12, "severe_count": 3, "severe_pct": 25.0, "severe_area_m2": 12846.3041, "total_area_m2": 45397.6054, "severity_score": 0.2775, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000287_0715", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000287", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 0.67, "building_count": 3, "severe_count": 3, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 3, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 34394.8595, "total_area_m2": 34394.8595, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 10, "severe_count": 5, "severe_pct": 50.0, "severe_area_m2": 37749.2889, "total_area_m2": 97419.8046, "severity_score": 0.434, "dominant_damage": "major-damage"}, {"cell_id": "G32", "row": 3, "col": 2, "building_count": 5, "severe_count": 3, "severe_pct": 60.0, "severe_area_m2": 12846.3041, "total_area_m2": 16623.5674, "severity_score": 0.402, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-michael_00000287_0716", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000287", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "From the post-disaster wind scene (hurricane-michael), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 26, "severe_footprint_m2": 178917.4639, "major_damage_area_m2": 178917.4639, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-michael_00000287_0717", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000287", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 26, "max_severe_in_cell": 9, "concentration_score": 0.3462, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 9, "Q1": 6, "Q2": 8, "Q3": 3}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000287_0718", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000287", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 107, "classified_buildings": 107, "counts": {"no-damage": 50, "minor-damage": 31, "major-damage": 26, "destroyed": 0, "unclassified": 0}, "severe_count": 26, "severe_pct": 24.3, "damaged_pct": 53.27}, "area_summary": {"total_footprint_m2": 484549.9102, "severe_footprint_m2": 178917.4639, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.3462, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 43936, "polygon_severe_count": 26, "pixel_histogram": {"0": 928477, "1": 40770, "2": 35393, "3": 43936, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 33.947, "2": 29.4699, "3": 36.5832, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-matthew_00000034_0719", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000034", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a wind event (hurricane-matthew) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 2, "minor-damage": 10, "major-damage": 4, "destroyed": 0, "unclassified": 3}, "total_buildings": 19, "classified_buildings": 16, "damaged_count": 14, "severe_count": 4, "damaged_pct": 87.5, "severe_pct": 25.0, "severe_footprint_m2": 13668.6899, "total_footprint_m2": 53764.107, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-matthew_00000034_0720", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000034", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "For a post-disaster wind scene (hurricane-matthew), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 3, "severe_footprint_m2": 9925.8803, "severe_pct_in_quadrant": 21.43, "dominant_damage": "minor-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 14, "severe_count": 3, "severe_pct": 21.43, "severe_area_m2": 9925.8803, "total_area_m2": 32942.2898, "severity_score": 0.285, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 3742.8096, "total_area_m2": 3742.8096, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q0", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1924.8497, "severity_score": 0.33, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1924.8497, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 15154.1579, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 3742.8096, "total_area_m2": 3742.8096, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 14, "severe_count": 3, "severe_pct": 21.43, "severe_area_m2": 9925.8803, "total_area_m2": 32942.2898, "severity_score": 0.285, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-matthew_00000034_0721", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000034", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G21", "severity_score": 0.67, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G21", "row": 2, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 3742.8096, "total_area_m2": 3742.8096, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G33", "row": 3, "col": 3, "building_count": 5, "severe_count": 2, "severe_pct": 40.0, "severe_area_m2": 5590.005, "total_area_m2": 11508.192, "severity_score": 0.4, "dominant_damage": "minor-damage"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1924.8497, "severity_score": 0.33, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-matthew_00000034_0722", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000034", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "From the post-disaster wind scene (hurricane-matthew), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 4, "severe_footprint_m2": 13668.6899, "major_damage_area_m2": 13668.6899, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-matthew_00000034_0723", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000034", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 4, "max_severe_in_cell": 3, "concentration_score": 0.75, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 1, "Q3": 3}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-matthew_00000034_0724", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000034", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-matthew). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-matthew", "disaster_type": "wind", "capture_date": "2016-10-01T15:29:45.000Z", "gsd": 2.37887692, "damage_summary": {"total_buildings": 19, "classified_buildings": 16, "counts": {"no-damage": 2, "minor-damage": 10, "major-damage": 4, "destroyed": 0, "unclassified": 3}, "severe_count": 4, "severe_pct": 25.0, "damaged_pct": 87.5}, "area_summary": {"total_footprint_m2": 53764.107, "severe_footprint_m2": 13668.6899, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.75, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 2552, "polygon_severe_count": 4, "pixel_histogram": {"0": 1039148, "1": 731, "2": 6145, "3": 2552, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 7.7535, "2": 65.1782, "3": 27.0683, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_socal-fire_00001105_0725", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "socal-fire_00001105", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a fire event (socal-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 13, "minor-damage": 1, "major-damage": 1, "destroyed": 0, "unclassified": 0}, "total_buildings": 15, "classified_buildings": 15, "damaged_count": 2, "severe_count": 1, "damaged_pct": 13.33, "severe_pct": 6.67, "severe_footprint_m2": 18787.3755, "total_footprint_m2": 180455.2601, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_socal-fire_00001105_0726", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "socal-fire_00001105", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "For a post-disaster fire scene (socal-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 1, "severe_footprint_m2": 18787.3755, "severe_pct_in_quadrant": 25.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 4, "severe_count": 1, "severe_pct": 25.0, "severe_area_m2": 18787.3755, "total_area_m2": 48551.105, "severity_score": 0.1675, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 62102.2246, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 62102.2246, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 4, "severe_count": 1, "severe_pct": 25.0, "severe_area_m2": 18787.3755, "total_area_m2": 48551.105, "severity_score": 0.1675, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 69801.9305, "severity_score": 0.055, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_socal-fire_00001105_0727", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "socal-fire_00001105", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G23", "severity_score": 0.33, "building_count": 1, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G23", "row": 2, "col": 3, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 19569.3041, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "G30", "row": 3, "col": 0, "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 18787.3755, "total_area_m2": 21889.3916, "severity_score": 0.2233, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_socal-fire_00001105_0728", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "socal-fire_00001105", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "From the post-disaster fire scene (socal-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 1, "severe_footprint_m2": 18787.3755, "major_damage_area_m2": 18787.3755, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_socal-fire_00001105_0729", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "socal-fire_00001105", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 1, "max_severe_in_cell": 1, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 1, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_socal-fire_00001105_0730", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "socal-fire_00001105", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (socal-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "socal-fire", "disaster_type": "fire", "capture_date": "2018-11-14T18:42:58.000Z", "gsd": 2.5700748, "damage_summary": {"total_buildings": 15, "classified_buildings": 15, "counts": {"no-damage": 13, "minor-damage": 1, "major-damage": 1, "destroyed": 0, "unclassified": 0}, "severe_count": 1, "severe_pct": 6.67, "damaged_pct": 13.33}, "area_summary": {"total_footprint_m2": 180455.2601, "severe_footprint_m2": 18787.3755, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 2983, "polygon_severe_count": 1, "pixel_histogram": {"0": 1019867, "1": 22619, "2": 3107, "3": 2983, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 78.7871, "2": 10.8224, "3": 10.3905, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-matthew_00000002_0731", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000002", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a wind event (hurricane-matthew) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 4, "minor-damage": 27, "major-damage": 6, "destroyed": 1, "unclassified": 0}, "total_buildings": 38, "classified_buildings": 38, "damaged_count": 34, "severe_count": 7, "damaged_pct": 89.47, "severe_pct": 18.42, "severe_footprint_m2": 4250.4423, "total_footprint_m2": 51880.5327, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-matthew_00000002_0732", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000002", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "For a post-disaster wind scene (hurricane-matthew), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 7, "severe_footprint_m2": 4250.4423, "severe_pct_in_quadrant": 24.14, "dominant_damage": "minor-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 29, "severe_count": 7, "severe_pct": 24.14, "severe_area_m2": 4250.4423, "total_area_m2": 37401.5749, "severity_score": 0.4007, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 14478.9578, "severity_score": 0.2567, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 29, "severe_count": 7, "severe_pct": 24.14, "severe_area_m2": 4250.4423, "total_area_m2": 37401.5749, "severity_score": 0.4007, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 14478.9578, "severity_score": 0.2567, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-matthew_00000002_0733", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000002", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 0.4806, "building_count": 18, "severe_count": 7, "class_histogram": {"no_damage": 0, "minor_damage": 11, "major_damage": 6, "destroyed": 1, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 18, "severe_count": 7, "severe_pct": 38.89, "severe_area_m2": 4250.4423, "total_area_m2": 20703.5517, "severity_score": 0.4806, "dominant_damage": "minor-damage"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 10824.4113, "severity_score": 0.2888, "dominant_damage": "minor-damage"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 14478.9578, "severity_score": 0.2567, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-matthew_00000002_0734", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000002", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "From the post-disaster wind scene (hurricane-matthew), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 7, "severe_footprint_m2": 4250.4423, "major_damage_area_m2": 3298.9519, "destroyed_area_m2": 951.4904, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-matthew_00000002_0735", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000002", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 7, "max_severe_in_cell": 7, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 7, "Q1": 0, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-matthew_00000002_0736", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000002", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-matthew). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-matthew", "disaster_type": "wind", "capture_date": "2016-10-09T15:32:03.000Z", "gsd": 2.77336645, "damage_summary": {"total_buildings": 38, "classified_buildings": 38, "counts": {"no-damage": 4, "minor-damage": 27, "major-damage": 6, "destroyed": 1, "unclassified": 0}, "severe_count": 7, "severe_pct": 18.42, "damaged_pct": 89.47}, "area_summary": {"total_footprint_m2": 51880.5327, "severe_footprint_m2": 4250.4423, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 658, "polygon_severe_count": 7, "pixel_histogram": {"0": 1040911, "1": 1061, "2": 5946, "3": 515, "4": 143, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 13.8421, "2": 77.5734, "3": 6.7189, "4": 1.8656, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_mexico-earthquake_00000049_0737", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000049", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a earthquake event (mexico-earthquake) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 513, "minor-damage": 1, "major-damage": 3, "destroyed": 0, "unclassified": 0}, "total_buildings": 517, "classified_buildings": 517, "damaged_count": 4, "severe_count": 3, "damaged_pct": 0.77, "severe_pct": 0.58, "severe_footprint_m2": 9020.1076, "total_footprint_m2": 1597269.3217, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_mexico-earthquake_00000049_0738", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000049", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "major-heavy", "question": "For a post-disaster earthquake scene (mexico-earthquake), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 3, "severe_footprint_m2": 9020.1076, "severe_pct_in_quadrant": 2.17, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 138, "severe_count": 3, "severe_pct": 2.17, "severe_area_m2": 9020.1076, "total_area_m2": 466541.645, "severity_score": 0.017, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 310, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 698645.3667, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 24, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 180467.2933, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 310, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 698645.3667, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 24, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 180467.2933, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 138, "severe_count": 3, "severe_pct": 2.17, "severe_area_m2": 9020.1076, "total_area_m2": 466541.645, "severity_score": 0.017, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 45, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 251615.0167, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_mexico-earthquake_00000049_0739", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000049", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster earthquake satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G30", "severity_score": 0.0359, "building_count": 56, "severe_count": 3, "class_histogram": {"no_damage": 53, "minor_damage": 0, "major_damage": 3, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G30", "row": 3, "col": 0, "building_count": 56, "severe_count": 3, "severe_pct": 5.36, "severe_area_m2": 9020.1076, "total_area_m2": 208850.8959, "severity_score": 0.0359, "dominant_damage": "no-damage"}, {"cell_id": "G31", "row": 3, "col": 1, "building_count": 21, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 79454.6855, "severity_score": 0.0157, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 160, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 262127.9399, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_mexico-earthquake_00000049_0740", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000049", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "major-heavy", "question": "From the post-disaster earthquake scene (mexico-earthquake), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 3, "severe_footprint_m2": 9020.1076, "major_damage_area_m2": 9020.1076, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_mexico-earthquake_00000049_0741", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000049", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster earthquake scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 3, "max_severe_in_cell": 3, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 3, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_mexico-earthquake_00000049_0742", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000049", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster earthquake scene (mexico-earthquake). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "mexico-earthquake", "disaster_type": "earthquake", "capture_date": "2017-09-20T17:46:11.000Z", "gsd": 2.6503215, "damage_summary": {"total_buildings": 517, "classified_buildings": 517, "counts": {"no-damage": 513, "minor-damage": 1, "major-damage": 3, "destroyed": 0, "unclassified": 0}, "severe_count": 3, "severe_pct": 0.58, "damaged_pct": 0.77}, "area_summary": {"total_footprint_m2": 1597269.3217, "severe_footprint_m2": 9020.1076, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 1400, "polygon_severe_count": 3, "pixel_histogram": {"0": 804622, "1": 241400, "2": 1154, "3": 1400, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 98.9531, "2": 0.473, "3": 0.5739, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000552_0743", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000552", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 6, "minor-damage": 11, "major-damage": 9, "destroyed": 0, "unclassified": 0}, "total_buildings": 26, "classified_buildings": 26, "damaged_count": 20, "severe_count": 9, "damaged_pct": 76.92, "severe_pct": 34.62, "severe_footprint_m2": 31881.6092, "total_footprint_m2": 60856.3446, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000552_0744", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000552", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 8, "severe_footprint_m2": 18402.4191, "severe_pct_in_quadrant": 32.0, "dominant_damage": "minor-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 25, "severe_count": 8, "severe_pct": 32.0, "severe_area_m2": 18402.4191, "total_area_m2": 47377.1545, "severity_score": 0.3596, "dominant_damage": "minor-damage"}, {"cell_id": "Q0", "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 13479.1901, "total_area_m2": 13479.1901, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 13479.1901, "total_area_m2": 13479.1901, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 25, "severe_count": 8, "severe_pct": 32.0, "severe_area_m2": 18402.4191, "total_area_m2": 47377.1545, "severity_score": 0.3596, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000552_0745", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000552", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 0.67, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 13479.1901, "total_area_m2": 13479.1901, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 8322.99, "total_area_m2": 8322.99, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G03", "row": 0, "col": 3, "building_count": 7, "severe_count": 5, "severe_pct": 71.43, "severe_area_m2": 10079.4291, "total_area_m2": 13487.6112, "severity_score": 0.5729, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000552_0746", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000552", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 9, "severe_footprint_m2": 31881.6092, "major_damage_area_m2": 31881.6092, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000552_0747", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000552", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 9, "max_severe_in_cell": 8, "concentration_score": 0.8889, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 1, "Q1": 8, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000552_0748", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000552", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 26, "classified_buildings": 26, "counts": {"no-damage": 6, "minor-damage": 11, "major-damage": 9, "destroyed": 0, "unclassified": 0}, "severe_count": 9, "severe_pct": 34.62, "damaged_pct": 76.92}, "area_summary": {"total_footprint_m2": 60856.3446, "severe_footprint_m2": 31881.6092, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.8889, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 9}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000456_0749", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000456", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 31, "minor-damage": 11, "major-damage": 8, "destroyed": 2, "unclassified": 1}, "total_buildings": 53, "classified_buildings": 52, "damaged_count": 21, "severe_count": 10, "damaged_pct": 40.38, "severe_pct": 19.23, "severe_footprint_m2": 16597.0464, "total_footprint_m2": 120930.2473, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000456_0750", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000456", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 9, "severe_footprint_m2": 15546.8151, "severe_pct_in_quadrant": 23.08, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 39, "severe_count": 9, "severe_pct": 23.08, "severe_area_m2": 15546.8151, "total_area_m2": 82274.1554, "severity_score": 0.2054, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 14, "severe_count": 1, "severe_pct": 7.14, "severe_area_m2": 1050.2313, "total_area_m2": 38656.0919, "severity_score": 0.2129, "dominant_damage": "minor-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 39, "severe_count": 9, "severe_pct": 23.08, "severe_area_m2": 15546.8151, "total_area_m2": 82274.1554, "severity_score": 0.2054, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 14, "severe_count": 1, "severe_pct": 7.14, "severe_area_m2": 1050.2313, "total_area_m2": 38656.0919, "severity_score": 0.2129, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000456_0751", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000456", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G13", "severity_score": 0.39, "building_count": 6, "severe_count": 3, "class_histogram": {"no_damage": 3, "minor_damage": 0, "major_damage": 2, "destroyed": 1, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G13", "row": 1, "col": 3, "building_count": 6, "severe_count": 3, "severe_pct": 50.0, "severe_area_m2": 4966.884, "total_area_m2": 11998.4804, "severity_score": 0.39, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 9, "severe_count": 4, "severe_pct": 44.44, "severe_area_m2": 4250.0078, "total_area_m2": 13711.5066, "severity_score": 0.3344, "dominant_damage": "no-damage"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 20317.9668, "severity_score": 0.2357, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000456_0752", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000456", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 10, "severe_footprint_m2": 16597.0464, "major_damage_area_m2": 15099.3373, "destroyed_area_m2": 1497.7091, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000456_0753", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000456", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 10, "max_severe_in_cell": 9, "concentration_score": 0.9, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 9, "Q2": 0, "Q3": 1}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000456_0754", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000456", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 53, "classified_buildings": 52, "counts": {"no-damage": 31, "minor-damage": 11, "major-damage": 8, "destroyed": 2, "unclassified": 1}, "severe_count": 10, "severe_pct": 19.23, "damaged_pct": 40.38}, "area_summary": {"total_footprint_m2": 120930.2473, "severe_footprint_m2": 16597.0464, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.9, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 10}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-harvey_00000351_0755", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000351", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (hurricane-harvey) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 47, "major-damage": 61, "destroyed": 0, "unclassified": 3}, "total_buildings": 111, "classified_buildings": 108, "damaged_count": 108, "severe_count": 61, "damaged_pct": 100.0, "severe_pct": 56.48, "severe_footprint_m2": 755352.0537, "total_footprint_m2": 1361531.3412, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-harvey_00000351_0756", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000351", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (hurricane-harvey), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 34, "severe_footprint_m2": 385174.3117, "severe_pct_in_quadrant": 73.91, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 46, "severe_count": 34, "severe_pct": 73.91, "severe_area_m2": 385174.3117, "total_area_m2": 546280.8765, "severity_score": 0.5813, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 57, "severe_count": 22, "severe_pct": 38.6, "severe_area_m2": 281799.7619, "total_area_m2": 663174.2291, "severity_score": 0.4439, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 5, "severe_pct": 62.5, "severe_area_m2": 88377.9801, "total_area_m2": 152076.2356, "severity_score": 0.5425, "dominant_damage": "major-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 46, "severe_count": 34, "severe_pct": 73.91, "severe_area_m2": 385174.3117, "total_area_m2": 546280.8765, "severity_score": 0.5813, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 57, "severe_count": 22, "severe_pct": 38.6, "severe_area_m2": 281799.7619, "total_area_m2": 663174.2291, "severity_score": 0.4439, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 5, "severe_pct": 62.5, "severe_area_m2": 88377.9801, "total_area_m2": 152076.2356, "severity_score": 0.5425, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-harvey_00000351_0757", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000351", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.67, "building_count": 11, "severe_count": 11, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 11, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 11, "severe_count": 11, "severe_pct": 100.0, "severe_area_m2": 138543.3539, "total_area_m2": 138543.3539, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 11, "severe_count": 11, "severe_pct": 100.0, "severe_area_m2": 132007.2131, "total_area_m2": 132007.2131, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G30", "row": 3, "col": 0, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 68891.0746, "total_area_m2": 68891.0746, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-harvey_00000351_0758", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000351", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (hurricane-harvey), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 61, "severe_footprint_m2": 755352.0537, "major_damage_area_m2": 755352.0537, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-harvey_00000351_0759", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000351", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 61, "max_severe_in_cell": 34, "concentration_score": 0.5574, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 34, "Q1": 22, "Q2": 5, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-harvey_00000351_0760", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000351", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-harvey). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-harvey", "disaster_type": "flooding", "capture_date": "2017-08-31T17:38:50.685Z", "gsd": 3.024613, "damage_summary": {"total_buildings": 111, "classified_buildings": 108, "counts": {"no-damage": 0, "minor-damage": 47, "major-damage": 61, "destroyed": 0, "unclassified": 3}, "severe_count": 61, "severe_pct": 56.48, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 1361531.3412, "severe_footprint_m2": 755352.0537, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5574, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 87450, "polygon_severe_count": 61, "pixel_histogram": {"0": 892395, "1": 0, "2": 68731, "3": 87450, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 44.0073, "3": 55.9927, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000461_0761", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000461", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 55, "minor-damage": 14, "major-damage": 9, "destroyed": 0, "unclassified": 0}, "total_buildings": 78, "classified_buildings": 78, "damaged_count": 23, "severe_count": 9, "damaged_pct": 29.49, "severe_pct": 11.54, "severe_footprint_m2": 68177.7086, "total_footprint_m2": 365187.1145, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000461_0762", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000461", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 5, "severe_footprint_m2": 43354.1883, "severe_pct_in_quadrant": 21.74, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 23, "severe_count": 5, "severe_pct": 21.74, "severe_area_m2": 43354.1883, "total_area_m2": 130434.6184, "severity_score": 0.203, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 10, "severe_count": 3, "severe_pct": 30.0, "severe_area_m2": 22229.855, "total_area_m2": 88711.9961, "severity_score": 0.3, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 20, "severe_count": 1, "severe_pct": 5.0, "severe_area_m2": 2593.6653, "total_area_m2": 61250.828, "severity_score": 0.0665, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 20, "severe_count": 1, "severe_pct": 5.0, "severe_area_m2": 2593.6653, "total_area_m2": 61250.828, "severity_score": 0.0665, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 10, "severe_count": 3, "severe_pct": 30.0, "severe_area_m2": 22229.855, "total_area_m2": 88711.9961, "severity_score": 0.3, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 25, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 84789.672, "severity_score": 0.066, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 23, "severe_count": 5, "severe_pct": 21.74, "severe_area_m2": 43354.1883, "total_area_m2": 130434.6184, "severity_score": 0.203, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000461_0763", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000461", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G03", "severity_score": 0.335, "building_count": 2, "severe_count": 1, "class_histogram": {"no_damage": 1, "minor_damage": 0, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G03", "row": 0, "col": 3, "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 10862.7238, "total_area_m2": 26978.9032, "severity_score": 0.335, "dominant_damage": "no-damage"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 8020.4088, "total_area_m2": 40152.2072, "severity_score": 0.335, "dominant_damage": "no-damage"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 17066.4889, "total_area_m2": 38409.7317, "severity_score": 0.3333, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-michael_00000461_0764", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000461", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "From the post-disaster wind scene (hurricane-michael), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 9, "severe_footprint_m2": 68177.7086, "major_damage_area_m2": 68177.7086, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-michael_00000461_0765", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000461", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 9, "max_severe_in_cell": 5, "concentration_score": 0.5556, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 1, "Q1": 3, "Q2": 0, "Q3": 5}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000461_0766", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000461", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 78, "classified_buildings": 78, "counts": {"no-damage": 55, "minor-damage": 14, "major-damage": 9, "destroyed": 0, "unclassified": 0}, "severe_count": 9, "severe_pct": 11.54, "damaged_pct": 29.49}, "area_summary": {"total_footprint_m2": 365187.1145, "severe_footprint_m2": 68177.7086, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5556, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 16769, "polygon_severe_count": 9, "pixel_histogram": {"0": 958025, "1": 58209, "2": 15573, "3": 16769, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 64.2831, "2": 17.198, "3": 18.5188, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_sunda-tsunami_00000023_0767", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000023", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a tsunami event (sunda-tsunami) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 1192, "minor-damage": 0, "major-damage": 3, "destroyed": 0, "unclassified": 0}, "total_buildings": 1195, "classified_buildings": 1195, "damaged_count": 3, "severe_count": 3, "damaged_pct": 0.25, "severe_pct": 0.25, "severe_footprint_m2": 6788.7146, "total_footprint_m2": 1173908.7056, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_sunda-tsunami_00000023_0768", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000023", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "major-heavy", "question": "For a post-disaster tsunami scene (sunda-tsunami), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 1, "severe_footprint_m2": 3253.9771, "severe_pct_in_quadrant": 0.31, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 327, "severe_count": 1, "severe_pct": 0.31, "severe_area_m2": 3253.9771, "total_area_m2": 320017.7222, "severity_score": 0.002, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 388, "severe_count": 1, "severe_pct": 0.26, "severe_area_m2": 2546.7644, "total_area_m2": 368291.3913, "severity_score": 0.0017, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 300, "severe_count": 1, "severe_pct": 0.33, "severe_area_m2": 987.9731, "total_area_m2": 264770.0651, "severity_score": 0.0022, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 388, "severe_count": 1, "severe_pct": 0.26, "severe_area_m2": 2546.7644, "total_area_m2": 368291.3913, "severity_score": 0.0017, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 300, "severe_count": 1, "severe_pct": 0.33, "severe_area_m2": 987.9731, "total_area_m2": 264770.0651, "severity_score": 0.0022, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 327, "severe_count": 1, "severe_pct": 0.31, "severe_area_m2": 3253.9771, "total_area_m2": 320017.7222, "severity_score": 0.002, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 180, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 220829.527, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_sunda-tsunami_00000023_0769", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000023", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster tsunami satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G13", "severity_score": 0.0108, "building_count": 62, "severe_count": 1, "class_histogram": {"no_damage": 61, "minor_damage": 0, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G13", "row": 1, "col": 3, "building_count": 62, "severe_count": 1, "severe_pct": 1.61, "severe_area_m2": 987.9731, "total_area_m2": 38263.8732, "severity_score": 0.0108, "dominant_damage": "no-damage"}, {"cell_id": "G30", "row": 3, "col": 0, "building_count": 69, "severe_count": 1, "severe_pct": 1.45, "severe_area_m2": 3253.9771, "total_area_m2": 90061.3618, "severity_score": 0.0097, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 109, "severe_count": 1, "severe_pct": 0.92, "severe_area_m2": 2546.7644, "total_area_m2": 97951.6769, "severity_score": 0.0061, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_sunda-tsunami_00000023_0770", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000023", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "major-heavy", "question": "From the post-disaster tsunami scene (sunda-tsunami), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 3, "severe_footprint_m2": 6788.7146, "major_damage_area_m2": 6788.7146, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_sunda-tsunami_00000023_0771", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000023", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster tsunami scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 3, "max_severe_in_cell": 1, "concentration_score": 0.3333, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 1, "Q1": 1, "Q2": 1, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_sunda-tsunami_00000023_0772", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000023", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster tsunami scene (sunda-tsunami). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "sunda-tsunami", "disaster_type": "tsunami", "capture_date": "2019-01-02T03:17:07.000Z", "gsd": 1.7746705, "damage_summary": {"total_buildings": 1195, "classified_buildings": 1195, "counts": {"no-damage": 1192, "minor-damage": 0, "major-damage": 3, "destroyed": 0, "unclassified": 0}, "severe_count": 3, "severe_pct": 0.25, "damaged_pct": 0.25}, "area_summary": {"total_footprint_m2": 1173908.7056, "severe_footprint_m2": 6788.7146, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.3333, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 3}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000307_0773", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000307", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 25, "minor-damage": 57, "major-damage": 23, "destroyed": 0, "unclassified": 0}, "total_buildings": 105, "classified_buildings": 105, "damaged_count": 80, "severe_count": 23, "damaged_pct": 76.19, "severe_pct": 21.9, "severe_footprint_m2": 84605.9189, "total_footprint_m2": 412317.7034, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000307_0774", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000307", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 12, "severe_footprint_m2": 49550.7195, "severe_pct_in_quadrant": 80.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 15, "severe_count": 12, "severe_pct": 80.0, "severe_area_m2": 49550.7195, "total_area_m2": 57238.6214, "severity_score": 0.602, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 78, "severe_count": 10, "severe_pct": 12.82, "severe_area_m2": 27423.3724, "total_area_m2": 299029.3198, "severity_score": 0.289, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 12, "severe_count": 1, "severe_pct": 8.33, "severe_area_m2": 7631.827, "total_area_m2": 56049.7622, "severity_score": 0.2208, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 15, "severe_count": 12, "severe_pct": 80.0, "severe_area_m2": 49550.7195, "total_area_m2": 57238.6214, "severity_score": 0.602, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 78, "severe_count": 10, "severe_pct": 12.82, "severe_area_m2": 27423.3724, "total_area_m2": 299029.3198, "severity_score": 0.289, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 12, "severe_count": 1, "severe_pct": 8.33, "severe_area_m2": 7631.827, "total_area_m2": 56049.7622, "severity_score": 0.2208, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000307_0775", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000307", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.67, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 6719.6041, "total_area_m2": 6719.6041, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 17681.7819, "total_area_m2": 17681.7819, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1008.3485, "total_area_m2": 1008.3485, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000307_0776", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000307", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 23, "severe_footprint_m2": 84605.9189, "major_damage_area_m2": 84605.9189, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000307_0777", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000307", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 23, "max_severe_in_cell": 12, "concentration_score": 0.5217, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 12, "Q1": 10, "Q2": 0, "Q3": 1}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000307_0778", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000307", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:17:02.085Z", "gsd": 2.9729018, "damage_summary": {"total_buildings": 105, "classified_buildings": 105, "counts": {"no-damage": 25, "minor-damage": 57, "major-damage": 23, "destroyed": 0, "unclassified": 0}, "severe_count": 23, "severe_pct": 21.9, "damaged_pct": 76.19}, "area_summary": {"total_footprint_m2": 412317.7034, "severe_footprint_m2": 84605.9189, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5217, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 23}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000201_0779", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000201", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 201, "minor-damage": 11, "major-damage": 1, "destroyed": 0, "unclassified": 1}, "total_buildings": 214, "classified_buildings": 213, "damaged_count": 12, "severe_count": 1, "damaged_pct": 5.63, "severe_pct": 0.47, "severe_footprint_m2": 2001.9066, "total_footprint_m2": 981416.965, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000201_0780", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000201", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 1, "severe_footprint_m2": 2001.9066, "severe_pct_in_quadrant": 1.35, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 74, "severe_count": 1, "severe_pct": 1.35, "severe_area_m2": 2001.9066, "total_area_m2": 275025.6186, "severity_score": 0.0447, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 83, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 324253.1166, "severity_score": 0.0119, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 14, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 38584.7092, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 83, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 324253.1166, "severity_score": 0.0119, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 74, "severe_count": 1, "severe_pct": 1.35, "severe_area_m2": 2001.9066, "total_area_m2": 275025.6186, "severity_score": 0.0447, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 14, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 38584.7092, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 43, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 343553.5206, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000201_0781", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000201", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G13", "severity_score": 0.0711, "building_count": 28, "severe_count": 1, "class_histogram": {"no_damage": 23, "minor_damage": 4, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G13", "row": 1, "col": 3, "building_count": 28, "severe_count": 1, "severe_pct": 3.57, "severe_area_m2": 2001.9066, "total_area_m2": 123847.3182, "severity_score": 0.0711, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 22142.8819, "severity_score": 0.03, "dominant_damage": "no-damage"}, {"cell_id": "G03", "row": 0, "col": 3, "building_count": 34, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 127759.0335, "severity_score": 0.0291, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000201_0782", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000201", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 1, "severe_footprint_m2": 2001.9066, "major_damage_area_m2": 2001.9066, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000201_0783", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000201", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 1, "max_severe_in_cell": 1, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 1, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000201_0784", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000201", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:17:02.085Z", "gsd": 2.9729018, "damage_summary": {"total_buildings": 214, "classified_buildings": 213, "counts": {"no-damage": 201, "minor-damage": 11, "major-damage": 1, "destroyed": 0, "unclassified": 1}, "severe_count": 1, "severe_pct": 0.47, "damaged_pct": 5.63}, "area_summary": {"total_footprint_m2": 981416.965, "severe_footprint_m2": 2001.9066, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 1}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000547_0791", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000547", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 14, "minor-damage": 5, "major-damage": 1, "destroyed": 0, "unclassified": 0}, "total_buildings": 20, "classified_buildings": 20, "damaged_count": 6, "severe_count": 1, "damaged_pct": 30.0, "severe_pct": 5.0, "severe_footprint_m2": 473.0473, "total_footprint_m2": 73226.434, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000547_0792", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000547", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 1, "severe_footprint_m2": 473.0473, "severe_pct_in_quadrant": 6.67, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 15, "severe_count": 1, "severe_pct": 6.67, "severe_area_m2": 473.0473, "total_area_m2": 57633.7723, "severity_score": 0.1107, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 15, "severe_count": 1, "severe_pct": 6.67, "severe_area_m2": 473.0473, "total_area_m2": 57633.7723, "severity_score": 0.1107, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 15592.6617, "severity_score": 0.132, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000547_0793", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000547", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G33", "severity_score": 0.33, "building_count": 1, "severe_count": 0, "class_histogram": {"no_damage": 0, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G33", "row": 3, "col": 3, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1278.3695, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "G20", "row": 2, "col": 0, "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 473.0473, "total_area_m2": 2833.1894, "severity_score": 0.2233, "dominant_damage": "no-damage"}, {"cell_id": "G31", "row": 3, "col": 1, "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 22584.1089, "severity_score": 0.11, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-michael_00000547_0794", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000547", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "From the post-disaster wind scene (hurricane-michael), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 1, "severe_footprint_m2": 473.0473, "major_damage_area_m2": 473.0473, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-michael_00000547_0795", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000547", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 1, "max_severe_in_cell": 1, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 1, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000547_0796", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000547", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 20, "classified_buildings": 20, "counts": {"no-damage": 14, "minor-damage": 5, "major-damage": 1, "destroyed": 0, "unclassified": 0}, "severe_count": 1, "severe_pct": 5.0, "damaged_pct": 30.0}, "area_summary": {"total_footprint_m2": 73226.434, "severe_footprint_m2": 473.0473, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 132, "polygon_severe_count": 1, "pixel_histogram": {"0": 1030367, "1": 12696, "2": 5381, "3": 132, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 69.7238, "2": 29.5513, "3": 0.7249, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-florence_00000018_0797", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000018", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (hurricane-florence) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 0, "major-damage": 42, "destroyed": 5, "unclassified": 0}, "total_buildings": 47, "classified_buildings": 47, "damaged_count": 47, "severe_count": 47, "damaged_pct": 100.0, "severe_pct": 100.0, "severe_footprint_m2": 109144.6691, "total_footprint_m2": 109144.6691, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-florence_00000018_0798", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000018", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (hurricane-florence), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 16, "severe_footprint_m2": 51129.5529, "severe_pct_in_quadrant": 100.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 16, "severe_count": 16, "severe_pct": 100.0, "severe_area_m2": 51129.5529, "total_area_m2": 51129.5529, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 18, "severe_count": 18, "severe_pct": 100.0, "severe_area_m2": 35854.2434, "total_area_m2": 35854.2434, "severity_score": 0.725, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 9, "severe_pct": 100.0, "severe_area_m2": 14260.6696, "total_area_m2": 14260.6696, "severity_score": 0.7433, "dominant_damage": "major-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 16, "severe_count": 16, "severe_pct": 100.0, "severe_area_m2": 51129.5529, "total_area_m2": 51129.5529, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 7900.2032, "total_area_m2": 7900.2032, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 18, "severe_count": 18, "severe_pct": 100.0, "severe_area_m2": 35854.2434, "total_area_m2": 35854.2434, "severity_score": 0.725, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 9, "severe_pct": 100.0, "severe_area_m2": 14260.6696, "total_area_m2": 14260.6696, "severity_score": 0.7433, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-florence_00000018_0799", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000018", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G21", "severity_score": 0.835, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 1, "destroyed": 1, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G21", "row": 2, "col": 1, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 5049.0112, "total_area_m2": 5049.0112, "severity_score": 0.835, "dominant_damage": "major-damage"}, {"cell_id": "G31", "row": 3, "col": 1, "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 8630.4408, "total_area_m2": 8630.4408, "severity_score": 0.7525, "dominant_damage": "major-damage"}, {"cell_id": "G33", "row": 3, "col": 3, "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 9710.5571, "total_area_m2": 9710.5571, "severity_score": 0.7525, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-florence_00000018_0800", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000018", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (hurricane-florence), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 47, "severe_footprint_m2": 109144.6691, "major_damage_area_m2": 108184.3016, "destroyed_area_m2": 960.3675, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-florence_00000018_0801", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000018", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 47, "max_severe_in_cell": 18, "concentration_score": 0.383, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 16, "Q1": 4, "Q2": 18, "Q3": 9}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-florence_00000018_0802", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000018", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-florence). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-florence", "disaster_type": "flooding", "capture_date": "2018-09-20T16:04:41.000Z", "gsd": 2.0916247, "damage_summary": {"total_buildings": 47, "classified_buildings": 47, "counts": {"no-damage": 0, "minor-damage": 0, "major-damage": 42, "destroyed": 5, "unclassified": 0}, "severe_count": 47, "severe_pct": 100.0, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 109144.6691, "severe_footprint_m2": 109144.6691, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.383, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 26766, "polygon_severe_count": 47, "pixel_histogram": {"0": 1021810, "1": 0, "2": 0, "3": 26492, "4": 274, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 98.9763, "4": 1.0237, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000105_0803", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000105", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 45, "minor-damage": 14, "major-damage": 17, "destroyed": 0, "unclassified": 0}, "total_buildings": 76, "classified_buildings": 76, "damaged_count": 31, "severe_count": 17, "damaged_pct": 40.79, "severe_pct": 22.37, "severe_footprint_m2": 24354.9309, "total_footprint_m2": 138880.6687, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000105_0804", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000105", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 11, "severe_footprint_m2": 14577.7392, "severe_pct_in_quadrant": 15.71, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 70, "severe_count": 11, "severe_pct": 15.71, "severe_area_m2": 14577.7392, "total_area_m2": 129103.477, "severity_score": 0.1713, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 9777.1917, "total_area_m2": 9777.1917, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 9777.1917, "total_area_m2": 9777.1917, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 70, "severe_count": 11, "severe_pct": 15.71, "severe_area_m2": 14577.7392, "total_area_m2": 129103.477, "severity_score": 0.1713, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000105_0805", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000105", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G10", "severity_score": 0.67, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 2, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G10", "row": 1, "col": 0, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 1884.6842, "total_area_m2": 1884.6842, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 7892.5075, "total_area_m2": 7892.5075, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 15, "severe_count": 4, "severe_pct": 26.67, "severe_area_m2": 2781.9525, "total_area_m2": 16182.8764, "severity_score": 0.3327, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000105_0806", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000105", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 17, "severe_footprint_m2": 24354.9309, "major_damage_area_m2": 24354.9309, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000105_0807", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000105", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 17, "max_severe_in_cell": 11, "concentration_score": 0.6471, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 6, "Q1": 11, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000105_0808", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000105", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 76, "classified_buildings": 76, "counts": {"no-damage": 45, "minor-damage": 14, "major-damage": 17, "destroyed": 0, "unclassified": 0}, "severe_count": 17, "severe_pct": 22.37, "damaged_pct": 40.79}, "area_summary": {"total_footprint_m2": 138880.6687, "severe_footprint_m2": 24354.9309, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6471, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 17}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-florence_00000409_0815", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000409", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A post-disaster satellite scene from a flooding event (hurricane-florence) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 8, "minor-damage": 3, "major-damage": 20, "destroyed": 0, "unclassified": 0}, "total_buildings": 31, "classified_buildings": 31, "damaged_count": 23, "severe_count": 20, "damaged_pct": 74.19, "severe_pct": 64.52, "severe_footprint_m2": 163973.0118, "total_footprint_m2": 240787.8065, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-florence_00000409_0816", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000409", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "For a post-disaster flooding scene (hurricane-florence), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 8, "severe_footprint_m2": 65870.6905, "severe_pct_in_quadrant": 100.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 65870.6905, "total_area_m2": 65870.6905, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 48854.5982, "total_area_m2": 48854.5982, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 4, "severe_pct": 57.14, "severe_area_m2": 32230.5807, "total_area_m2": 54198.3397, "severity_score": 0.43, "dominant_damage": "major-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 10, "severe_count": 2, "severe_pct": 20.0, "severe_area_m2": 17017.1424, "total_area_m2": 71864.1781, "severity_score": 0.2, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 48854.5982, "total_area_m2": 48854.5982, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 4, "severe_pct": 57.14, "severe_area_m2": 32230.5807, "total_area_m2": 54198.3397, "severity_score": 0.43, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 65870.6905, "total_area_m2": 65870.6905, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-florence_00000409_0817", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000409", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 0.67, "building_count": 3, "severe_count": 3, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 3, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 25352.2713, "total_area_m2": 25352.2713, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G03", "row": 0, "col": 3, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 5325.986, "total_area_m2": 5325.986, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 8565.5786, "total_area_m2": 8565.5786, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-florence_00000409_0818", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000409", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "From the post-disaster flooding scene (hurricane-florence), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 20, "severe_footprint_m2": 163973.0118, "major_damage_area_m2": 163973.0118, "destroyed_area_m2": 0.0, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-florence_00000409_0819", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000409", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 20, "max_severe_in_cell": 8, "concentration_score": 0.4, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 2, "Q1": 6, "Q2": 4, "Q3": 8}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-florence_00000409_0820", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000409", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-florence). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-florence", "disaster_type": "flooding", "capture_date": "2018-09-20T16:04:41.000Z", "gsd": 2.0916247, "damage_summary": {"total_buildings": 31, "classified_buildings": 31, "counts": {"no-damage": 8, "minor-damage": 3, "major-damage": 20, "destroyed": 0, "unclassified": 0}, "severe_count": 20, "severe_pct": 64.52, "damaged_pct": 74.19}, "area_summary": {"total_footprint_m2": 240787.8065, "severe_footprint_m2": 163973.0118, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.4, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 39184, "polygon_severe_count": 20, "pixel_histogram": {"0": 990956, "1": 11465, "2": 6971, "3": 39184, "4": 0, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 19.8976, "2": 12.0982, "3": 68.0042, "4": 0.0, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_woolsey-fire_00000532_0827", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000532", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (woolsey-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 2, "minor-damage": 3, "major-damage": 0, "destroyed": 5, "unclassified": 1}, "total_buildings": 11, "classified_buildings": 10, "damaged_count": 8, "severe_count": 5, "damaged_pct": 80.0, "severe_pct": 50.0, "severe_footprint_m2": 2491.7538, "total_footprint_m2": 6599.7861, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_woolsey-fire_00000532_0828", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000532", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (woolsey-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 5, "severe_footprint_m2": 2491.7538, "severe_pct_in_quadrant": 50.0, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 10, "severe_count": 5, "severe_pct": 50.0, "severe_area_m2": 2491.7538, "total_area_m2": 6421.147, "severity_score": 0.599, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 178.6391, "severity_score": 0.0, "dominant_damage": "unclassified"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 178.6391, "severity_score": 0.0, "dominant_damage": "unclassified"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 10, "severe_count": 5, "severe_pct": 50.0, "severe_area_m2": 2491.7538, "total_area_m2": 6421.147, "severity_score": 0.599, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_woolsey-fire_00000532_0829", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000532", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G33", "severity_score": 0.599, "building_count": 10, "severe_count": 5, "class_histogram": {"no_damage": 2, "minor_damage": 3, "major_damage": 0, "destroyed": 5, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G33", "row": 3, "col": 3, "building_count": 10, "severe_count": 5, "severe_pct": 50.0, "severe_area_m2": 2491.7538, "total_area_m2": 6421.147, "severity_score": 0.599, "dominant_damage": "destroyed"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 178.6391, "severity_score": 0.0, "dominant_damage": "unclassified"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_woolsey-fire_00000532_0830", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000532", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (woolsey-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 5, "severe_footprint_m2": 2491.7538, "major_damage_area_m2": 0.0, "destroyed_area_m2": 2491.7538, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_woolsey-fire_00000532_0831", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000532", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 5, "max_severe_in_cell": 5, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 0, "Q3": 5}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_woolsey-fire_00000532_0832", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000532", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (woolsey-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "woolsey-fire", "disaster_type": "fire", "capture_date": "2018-11-18T18:44:07.000Z", "gsd": 1.7726423000000002, "damage_summary": {"total_buildings": 11, "classified_buildings": 10, "counts": {"no-damage": 2, "minor-damage": 3, "major-damage": 0, "destroyed": 5, "unclassified": 1}, "severe_count": 5, "severe_pct": 50.0, "damaged_pct": 80.0}, "area_summary": {"total_footprint_m2": 6599.7861, "severe_footprint_m2": 2491.7538, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 5}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_santa-rosa-wildfire_00000253_0833", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000253", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (santa-rosa-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 10, "minor-damage": 0, "major-damage": 0, "destroyed": 5, "unclassified": 0}, "total_buildings": 15, "classified_buildings": 15, "damaged_count": 5, "severe_count": 5, "damaged_pct": 33.33, "severe_pct": 33.33, "severe_footprint_m2": 13382.2172, "total_footprint_m2": 35334.0188, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_santa-rosa-wildfire_00000253_0834", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000253", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (santa-rosa-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 5, "severe_footprint_m2": 13382.2172, "severe_pct_in_quadrant": 45.45, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 11, "severe_count": 5, "severe_pct": 45.45, "severe_area_m2": 13382.2172, "total_area_m2": 33387.9172, "severity_score": 0.4545, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1946.1016, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 11, "severe_count": 5, "severe_pct": 45.45, "severe_area_m2": 13382.2172, "total_area_m2": 33387.9172, "severity_score": 0.4545, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1946.1016, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_santa-rosa-wildfire_00000253_0835", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000253", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.7143, "building_count": 7, "severe_count": 5, "class_histogram": {"no_damage": 2, "minor_damage": 0, "major_damage": 0, "destroyed": 5, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 7, "severe_count": 5, "severe_pct": 71.43, "severe_area_m2": 13382.2172, "total_area_m2": 20698.9673, "severity_score": 0.7143, "dominant_damage": "destroyed"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 8182.5883, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_santa-rosa-wildfire_00000253_0836", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000253", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (santa-rosa-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 5, "severe_footprint_m2": 13382.2172, "major_damage_area_m2": 0.0, "destroyed_area_m2": 13382.2172, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_santa-rosa-wildfire_00000253_0837", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000253", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 5, "max_severe_in_cell": 5, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 5, "Q1": 0, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_santa-rosa-wildfire_00000253_0838", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000253", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (santa-rosa-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "santa-rosa-wildfire", "disaster_type": "fire", "capture_date": "2017-10-11T19:19:41.000Z", "gsd": 1.8769937, "damage_summary": {"total_buildings": 15, "classified_buildings": 15, "counts": {"no-damage": 10, "minor-damage": 0, "major-damage": 0, "destroyed": 5, "unclassified": 0}, "severe_count": 5, "severe_pct": 33.33, "damaged_pct": 33.33}, "area_summary": {"total_footprint_m2": 35334.0188, "severe_footprint_m2": 13382.2172, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 4083, "polygon_severe_count": 5, "pixel_histogram": {"0": 1037863, "1": 6630, "2": 0, "3": 0, "4": 4083, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 61.8874, "2": 0.0, "3": 0.0, "4": 38.1126, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_guatemala-volcano_00000025_0839", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "guatemala-volcano_00000025", "split": "train", "qa_split": "train", "disaster": "guatemala-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a volcano event (guatemala-volcano) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 0, "major-damage": 6, "destroyed": 21, "unclassified": 174}, "total_buildings": 201, "classified_buildings": 27, "damaged_count": 27, "severe_count": 27, "damaged_pct": 100.0, "severe_pct": 100.0, "severe_footprint_m2": 15332.6583, "total_footprint_m2": 135770.1672, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_guatemala-volcano_00000025_0840", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "guatemala-volcano_00000025", "split": "train", "qa_split": "train", "disaster": "guatemala-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster volcano scene (guatemala-volcano), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 13, "severe_footprint_m2": 8794.2757, "severe_pct_in_quadrant": 41.94, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 31, "severe_count": 13, "severe_pct": 41.94, "severe_area_m2": 8794.2757, "total_area_m2": 22447.3527, "severity_score": 0.3555, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 170, "severe_count": 14, "severe_pct": 8.24, "severe_area_m2": 6538.3826, "total_area_m2": 113322.8145, "severity_score": 0.0824, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 170, "severe_count": 14, "severe_pct": 8.24, "severe_area_m2": 6538.3826, "total_area_m2": 113322.8145, "severity_score": 0.0824, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 31, "severe_count": 13, "severe_pct": 41.94, "severe_area_m2": 8794.2757, "total_area_m2": 22447.3527, "severity_score": 0.3555, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_guatemala-volcano_00000025_0841", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "guatemala-volcano_00000025", "split": "train", "qa_split": "train", "disaster": "guatemala-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster volcano satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G12", "severity_score": 1.0, "building_count": 6, "severe_count": 6, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 6, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G12", "row": 1, "col": 2, "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 1955.6778, "total_area_m2": 1955.6778, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G33", "row": 3, "col": 3, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 2041.7122, "total_area_m2": 2041.7122, "severity_score": 0.89, "dominant_damage": "destroyed"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 28, "severe_count": 10, "severe_pct": 35.71, "severe_area_m2": 6752.5635, "total_area_m2": 20405.6405, "severity_score": 0.2982, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_guatemala-volcano_00000025_0842", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "guatemala-volcano_00000025", "split": "train", "qa_split": "train", "disaster": "guatemala-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster volcano scene (guatemala-volcano), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 27, "severe_footprint_m2": 15332.6583, "major_damage_area_m2": 3602.7144, "destroyed_area_m2": 11729.9439, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_guatemala-volcano_00000025_0843", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "guatemala-volcano_00000025", "split": "train", "qa_split": "train", "disaster": "guatemala-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster volcano scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 27, "max_severe_in_cell": 14, "concentration_score": 0.5185, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 14, "Q2": 0, "Q3": 13}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_guatemala-volcano_00000025_0844", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "guatemala-volcano_00000025", "split": "train", "qa_split": "train", "disaster": "guatemala-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster volcano scene (guatemala-volcano). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "guatemala-volcano", "disaster_type": "volcano", "capture_date": "2018-06-22T16:55:40.000Z", "gsd": 1.4085245, "damage_summary": {"total_buildings": 201, "classified_buildings": 27, "counts": {"no-damage": 0, "minor-damage": 0, "major-damage": 6, "destroyed": 21, "unclassified": 174}, "severe_count": 27, "severe_pct": 100.0, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 135770.1672, "severe_footprint_m2": 15332.6583, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5185, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 8535, "polygon_severe_count": 27, "pixel_histogram": {"0": 1040041, "1": 0, "2": 0, "3": 1984, "4": 6551, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 23.2455, "4": 76.7545, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_woolsey-fire_00000088_0851", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000088", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (woolsey-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 11, "minor-damage": 0, "major-damage": 0, "destroyed": 5, "unclassified": 0}, "total_buildings": 16, "classified_buildings": 16, "damaged_count": 5, "severe_count": 5, "damaged_pct": 31.25, "severe_pct": 31.25, "severe_footprint_m2": 11109.9175, "total_footprint_m2": 68122.02, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_woolsey-fire_00000088_0852", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000088", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (woolsey-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 5, "severe_footprint_m2": 11109.9175, "severe_pct_in_quadrant": 71.43, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 7, "severe_count": 5, "severe_pct": 71.43, "severe_area_m2": 11109.9175, "total_area_m2": 16229.1563, "severity_score": 0.7143, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 8997.9041, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 8997.9041, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 42894.9596, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 7, "severe_count": 5, "severe_pct": 71.43, "severe_area_m2": 11109.9175, "total_area_m2": 16229.1563, "severity_score": 0.7143, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_woolsey-fire_00000088_0853", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000088", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G33", "severity_score": 1.0, "building_count": 5, "severe_count": 5, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 5, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G33", "row": 3, "col": 3, "building_count": 5, "severe_count": 5, "severe_pct": 100.0, "severe_area_m2": 11109.9175, "total_area_m2": 11109.9175, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_woolsey-fire_00000088_0854", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000088", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (woolsey-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 5, "severe_footprint_m2": 11109.9175, "major_damage_area_m2": 0.0, "destroyed_area_m2": 11109.9175, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_woolsey-fire_00000088_0855", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000088", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 5, "max_severe_in_cell": 5, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 0, "Q3": 5}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_woolsey-fire_00000088_0856", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000088", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (woolsey-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "woolsey-fire", "disaster_type": "fire", "capture_date": "2018-11-18T18:44:40.000Z", "gsd": 2.1112316, "damage_summary": {"total_buildings": 16, "classified_buildings": 16, "counts": {"no-damage": 11, "minor-damage": 0, "major-damage": 0, "destroyed": 5, "unclassified": 0}, "severe_count": 5, "severe_pct": 31.25, "damaged_pct": 31.25}, "area_summary": {"total_footprint_m2": 68122.02, "severe_footprint_m2": 11109.9175, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 5}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000211_0857", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000211", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 13, "minor-damage": 52, "major-damage": 5, "destroyed": 18, "unclassified": 0}, "total_buildings": 88, "classified_buildings": 88, "damaged_count": 75, "severe_count": 23, "damaged_pct": 85.23, "severe_pct": 26.14, "severe_footprint_m2": 111843.5923, "total_footprint_m2": 552180.2045, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000211_0858", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000211", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 7, "severe_footprint_m2": 56845.4527, "severe_pct_in_quadrant": 29.17, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 24, "severe_count": 7, "severe_pct": 29.17, "severe_area_m2": 56845.4527, "total_area_m2": 134065.2097, "severity_score": 0.3604, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 20, "severe_count": 4, "severe_pct": 20.0, "severe_area_m2": 28793.2861, "total_area_m2": 87573.3674, "severity_score": 0.431, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 30, "severe_count": 8, "severe_pct": 26.67, "severe_area_m2": 13236.5515, "total_area_m2": 202695.2197, "severity_score": 0.5087, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 20, "severe_count": 4, "severe_pct": 20.0, "severe_area_m2": 28793.2861, "total_area_m2": 87573.3674, "severity_score": 0.431, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 30, "severe_count": 8, "severe_pct": 26.67, "severe_area_m2": 13236.5515, "total_area_m2": 202695.2197, "severity_score": 0.5087, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 24, "severe_count": 7, "severe_pct": 29.17, "severe_area_m2": 56845.4527, "total_area_m2": 134065.2097, "severity_score": 0.3604, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 14, "severe_count": 4, "severe_pct": 28.57, "severe_area_m2": 12968.302, "total_area_m2": 127846.4077, "severity_score": 0.4271, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000211_0859", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000211", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G03", "severity_score": 0.6908, "building_count": 13, "severe_count": 7, "class_histogram": {"no_damage": 0, "minor_damage": 6, "major_damage": 0, "destroyed": 7, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G03", "row": 0, "col": 3, "building_count": 13, "severe_count": 7, "severe_pct": 53.85, "severe_area_m2": 12804.1685, "total_area_m2": 28893.5323, "severity_score": 0.6908, "dominant_damage": "destroyed"}, {"cell_id": "G20", "row": 2, "col": 0, "building_count": 5, "severe_count": 3, "severe_pct": 60.0, "severe_area_m2": 26624.2458, "total_area_m2": 38386.5249, "severity_score": 0.666, "dominant_damage": "minor-damage"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 4, "severe_count": 2, "severe_pct": 50.0, "severe_area_m2": 6664.3171, "total_area_m2": 19009.704, "severity_score": 0.5825, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-michael_00000211_0860", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000211", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster wind scene (hurricane-michael), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 23, "severe_footprint_m2": 111843.5923, "major_damage_area_m2": 65343.0925, "destroyed_area_m2": 46500.4998, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-michael_00000211_0861", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000211", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 23, "max_severe_in_cell": 8, "concentration_score": 0.3478, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 4, "Q1": 8, "Q2": 7, "Q3": 4}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000211_0862", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000211", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 88, "classified_buildings": 88, "counts": {"no-damage": 13, "minor-damage": 52, "major-damage": 5, "destroyed": 18, "unclassified": 0}, "severe_count": 23, "severe_pct": 26.14, "damaged_pct": 85.23}, "area_summary": {"total_footprint_m2": 552180.2045, "severe_footprint_m2": 111843.5923, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.3478, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 27513, "polygon_severe_count": 23, "pixel_histogram": {"0": 913574, "1": 26790, "2": 80699, "3": 15783, "4": 11730, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 19.8442, "2": 59.7762, "3": 11.6909, "4": 8.6888, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_socal-fire_00000358_0863", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "socal-fire_00000358", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (socal-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 1, "minor-damage": 0, "major-damage": 0, "destroyed": 16, "unclassified": 0}, "total_buildings": 17, "classified_buildings": 17, "damaged_count": 16, "severe_count": 16, "damaged_pct": 94.12, "severe_pct": 94.12, "severe_footprint_m2": 28226.1103, "total_footprint_m2": 28859.3781, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_socal-fire_00000358_0864", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "socal-fire_00000358", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (socal-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 5, "severe_footprint_m2": 16909.5094, "severe_pct_in_quadrant": 100.0, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 5, "severe_count": 5, "severe_pct": 100.0, "severe_area_m2": 16909.5094, "total_area_m2": 16909.5094, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 9, "severe_count": 8, "severe_pct": 88.89, "severe_area_m2": 9534.852, "total_area_m2": 10168.1198, "severity_score": 0.8889, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 1390.0945, "total_area_m2": 1390.0945, "severity_score": 1.0, "dominant_damage": "destroyed"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 9, "severe_count": 8, "severe_pct": 88.89, "severe_area_m2": 9534.852, "total_area_m2": 10168.1198, "severity_score": 0.8889, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 391.6544, "total_area_m2": 391.6544, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 5, "severe_count": 5, "severe_pct": 100.0, "severe_area_m2": 16909.5094, "total_area_m2": 16909.5094, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 1390.0945, "total_area_m2": 1390.0945, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_socal-fire_00000358_0865", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "socal-fire_00000358", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 1.0, "building_count": 4, "severe_count": 4, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 4, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 6737.4776, "total_area_m2": 6737.4776, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 295.0905, "total_area_m2": 295.0905, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 391.6544, "total_area_m2": 391.6544, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_socal-fire_00000358_0866", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "socal-fire_00000358", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (socal-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 16, "severe_footprint_m2": 28226.1103, "major_damage_area_m2": 0.0, "destroyed_area_m2": 28226.1103, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_socal-fire_00000358_0867", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "socal-fire_00000358", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 16, "max_severe_in_cell": 8, "concentration_score": 0.5, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 8, "Q1": 1, "Q2": 5, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_socal-fire_00000358_0868", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "socal-fire_00000358", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (socal-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "socal-fire", "disaster_type": "fire", "capture_date": "2018-11-14T18:42:58.000Z", "gsd": 2.5700748, "damage_summary": {"total_buildings": 17, "classified_buildings": 17, "counts": {"no-damage": 1, "minor-damage": 0, "major-damage": 0, "destroyed": 16, "unclassified": 0}, "severe_count": 16, "severe_pct": 94.12, "damaged_pct": 94.12}, "area_summary": {"total_footprint_m2": 28859.3781, "severe_footprint_m2": 28226.1103, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 4733, "polygon_severe_count": 16, "pixel_histogram": {"0": 1043723, "1": 120, "2": 0, "3": 0, "4": 4733, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 2.4727, "2": 0.0, "3": 0.0, "4": 97.5273, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_midwest-flooding_00000002_0869", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000002", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a flooding event (midwest-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 6, "minor-damage": 0, "major-damage": 1, "destroyed": 4, "unclassified": 0}, "total_buildings": 11, "classified_buildings": 11, "damaged_count": 5, "severe_count": 5, "damaged_pct": 45.45, "severe_pct": 45.45, "severe_footprint_m2": 6390.7094, "total_footprint_m2": 47451.6685, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_midwest-flooding_00000002_0870", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000002", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster flooding scene (midwest-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 4, "severe_footprint_m2": 4976.776, "severe_pct_in_quadrant": 100.0, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 4976.776, "total_area_m2": 4976.776, "severity_score": 0.9175, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 1413.9334, "total_area_m2": 7579.5843, "severity_score": 0.5, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 34895.3082, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 1413.9334, "total_area_m2": 7579.5843, "severity_score": 0.5, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 4976.776, "total_area_m2": 4976.776, "severity_score": 0.9175, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 34895.3082, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_midwest-flooding_00000002_0871", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000002", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G11", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G11", "row": 1, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1413.9334, "total_area_m2": 1413.9334, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1521.4233, "total_area_m2": 1521.4233, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 3455.3527, "total_area_m2": 3455.3527, "severity_score": 0.89, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_midwest-flooding_00000002_0872", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000002", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster flooding scene (midwest-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 5, "severe_footprint_m2": 6390.7094, "major_damage_area_m2": 633.8931, "destroyed_area_m2": 5756.8163, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_midwest-flooding_00000002_0873", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000002", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 5, "max_severe_in_cell": 4, "concentration_score": 0.8, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 1, "Q1": 4, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_midwest-flooding_00000002_0874", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000002", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (midwest-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "midwest-flooding", "disaster_type": "flooding", "capture_date": "2019-05-31T16:54:31.000Z", "gsd": 1.7392077, "damage_summary": {"total_buildings": 11, "classified_buildings": 11, "counts": {"no-damage": 6, "minor-damage": 0, "major-damage": 1, "destroyed": 4, "unclassified": 0}, "severe_count": 5, "severe_pct": 45.45, "damaged_pct": 45.45}, "area_summary": {"total_footprint_m2": 47451.6685, "severe_footprint_m2": 6390.7094, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.8, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 2329, "polygon_severe_count": 5, "pixel_histogram": {"0": 1032035, "1": 14212, "2": 0, "3": 248, "4": 2081, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 85.9198, "2": 0.0, "3": 1.4993, "4": 12.5809, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00000360_0875", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000360", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 41, "minor-damage": 0, "major-damage": 0, "destroyed": 3, "unclassified": 6}, "total_buildings": 50, "classified_buildings": 44, "damaged_count": 3, "severe_count": 3, "damaged_pct": 6.82, "severe_pct": 6.82, "severe_footprint_m2": 18649.2943, "total_footprint_m2": 168355.5106, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00000360_0876", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000360", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 2, "severe_footprint_m2": 15716.2082, "severe_pct_in_quadrant": 6.06, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 33, "severe_count": 2, "severe_pct": 6.06, "severe_area_m2": 15716.2082, "total_area_m2": 108528.5896, "severity_score": 0.0606, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 12, "severe_count": 1, "severe_pct": 8.33, "severe_area_m2": 2933.0861, "total_area_m2": 39182.0074, "severity_score": 0.0833, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 12, "severe_count": 1, "severe_pct": 8.33, "severe_area_m2": 2933.0861, "total_area_m2": 39182.0074, "severity_score": 0.0833, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 33, "severe_count": 2, "severe_pct": 6.06, "severe_area_m2": 15716.2082, "total_area_m2": 108528.5896, "severity_score": 0.0606, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 20644.9136, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00000360_0877", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000360", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G10", "severity_score": 0.5, "building_count": 2, "severe_count": 1, "class_histogram": {"no_damage": 1, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G10", "row": 1, "col": 0, "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 2933.0861, "total_area_m2": 4686.1074, "severity_score": 0.5, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 22, "severe_count": 2, "severe_pct": 9.09, "severe_area_m2": 15716.2082, "total_area_m2": 80159.0637, "severity_score": 0.0909, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_portugal-wildfire_00000360_0878", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000360", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (portugal-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 3, "severe_footprint_m2": 18649.2943, "major_damage_area_m2": 0.0, "destroyed_area_m2": 18649.2943, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_portugal-wildfire_00000360_0879", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000360", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 3, "max_severe_in_cell": 2, "concentration_score": 0.6667, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 1, "Q1": 2, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00000360_0880", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000360", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 50, "classified_buildings": 44, "counts": {"no-damage": 41, "minor-damage": 0, "major-damage": 0, "destroyed": 3, "unclassified": 6}, "severe_count": 3, "severe_pct": 6.82, "damaged_pct": 6.82}, "area_summary": {"total_footprint_m2": 168355.5106, "severe_footprint_m2": 18649.2943, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6667, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 3}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_woolsey-fire_00000419_0881", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000419", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (woolsey-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 9, "minor-damage": 0, "major-damage": 0, "destroyed": 1, "unclassified": 1}, "total_buildings": 11, "classified_buildings": 10, "damaged_count": 1, "severe_count": 1, "damaged_pct": 10.0, "severe_pct": 10.0, "severe_footprint_m2": 768.5366, "total_footprint_m2": 12297.7376, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_woolsey-fire_00000419_0882", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000419", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (woolsey-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 1, "severe_footprint_m2": 768.5366, "severe_pct_in_quadrant": 9.09, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 11, "severe_count": 1, "severe_pct": 9.09, "severe_area_m2": 768.5366, "total_area_m2": 12297.7376, "severity_score": 0.0909, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 11, "severe_count": 1, "severe_pct": 9.09, "severe_area_m2": 768.5366, "total_area_m2": 12297.7376, "severity_score": 0.0909, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_woolsey-fire_00000419_0883", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000419", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 0.1667, "building_count": 6, "severe_count": 1, "class_histogram": {"no_damage": 5, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 6, "severe_count": 1, "severe_pct": 16.67, "severe_area_m2": 768.5366, "total_area_m2": 10378.2393, "severity_score": 0.1667, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_woolsey-fire_00000419_0884", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000419", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (woolsey-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 1, "severe_footprint_m2": 768.5366, "major_damage_area_m2": 0.0, "destroyed_area_m2": 768.5366, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_woolsey-fire_00000419_0885", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000419", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 1, "max_severe_in_cell": 1, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 1, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_woolsey-fire_00000419_0886", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000419", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (woolsey-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "woolsey-fire", "disaster_type": "fire", "capture_date": "2018-11-18T18:44:40.000Z", "gsd": 2.1112316, "damage_summary": {"total_buildings": 11, "classified_buildings": 10, "counts": {"no-damage": 9, "minor-damage": 0, "major-damage": 0, "destroyed": 1, "unclassified": 1}, "severe_count": 1, "severe_pct": 10.0, "damaged_pct": 10.0}, "area_summary": {"total_footprint_m2": 12297.7376, "severe_footprint_m2": 768.5366, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 1}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_moore-tornado_00000120_0899", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "moore-tornado_00000120", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a wind event (moore-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 2, "major-damage": 2, "destroyed": 16, "unclassified": 0}, "total_buildings": 20, "classified_buildings": 20, "damaged_count": 20, "severe_count": 18, "damaged_pct": 100.0, "severe_pct": 90.0, "severe_footprint_m2": 85329.6156, "total_footprint_m2": 95608.1605, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_moore-tornado_00000120_0900", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "moore-tornado_00000120", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster wind scene (moore-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 7, "severe_footprint_m2": 33079.8153, "severe_pct_in_quadrant": 77.78, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 9, "severe_count": 7, "severe_pct": 77.78, "severe_area_m2": 33079.8153, "total_area_m2": 43358.3602, "severity_score": 0.8144, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 32281.2338, "total_area_m2": 32281.2338, "severity_score": 0.945, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 5, "severe_count": 5, "severe_pct": 100.0, "severe_area_m2": 19968.5665, "total_area_m2": 19968.5665, "severity_score": 1.0, "dominant_damage": "destroyed"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 32281.2338, "total_area_m2": 32281.2338, "severity_score": 0.945, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 9, "severe_count": 7, "severe_pct": 77.78, "severe_area_m2": 33079.8153, "total_area_m2": 43358.3602, "severity_score": 0.8144, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 5, "severe_count": 5, "severe_pct": 100.0, "severe_area_m2": 19968.5665, "total_area_m2": 19968.5665, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_moore-tornado_00000120_0901", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "moore-tornado_00000120", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 1.0, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 2, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 11124.6805, "total_area_m2": 11124.6805, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 8564.3028, "total_area_m2": 8564.3028, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 15455.4615, "total_area_m2": 15455.4615, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_moore-tornado_00000120_0902", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "moore-tornado_00000120", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster wind scene (moore-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 18, "severe_footprint_m2": 85329.6156, "major_damage_area_m2": 13368.6634, "destroyed_area_m2": 71960.9522, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_moore-tornado_00000120_0903", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "moore-tornado_00000120", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 18, "max_severe_in_cell": 7, "concentration_score": 0.3889, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 6, "Q1": 7, "Q2": 5, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_moore-tornado_00000120_0904", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "moore-tornado_00000120", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (moore-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "moore-tornado", "disaster_type": "wind", "capture_date": "2013-05-22T17:26:30.085Z", "gsd": 1.72984850406647, "damage_summary": {"total_buildings": 20, "classified_buildings": 20, "counts": {"no-damage": 0, "minor-damage": 2, "major-damage": 2, "destroyed": 16, "unclassified": 0}, "severe_count": 18, "severe_pct": 90.0, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 95608.1605, "severe_footprint_m2": 85329.6156, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.3889, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 18}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00001804_0917", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001804", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 9, "minor-damage": 0, "major-damage": 0, "destroyed": 5, "unclassified": 0}, "total_buildings": 14, "classified_buildings": 14, "damaged_count": 5, "severe_count": 5, "damaged_pct": 35.71, "severe_pct": 35.71, "severe_footprint_m2": 11885.0597, "total_footprint_m2": 31103.8256, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00001804_0918", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001804", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 3, "severe_footprint_m2": 9801.7269, "severe_pct_in_quadrant": 60.0, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 5, "severe_count": 3, "severe_pct": 60.0, "severe_area_m2": 9801.7269, "total_area_m2": 15360.8794, "severity_score": 0.6, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 9, "severe_count": 2, "severe_pct": 22.22, "severe_area_m2": 2083.3328, "total_area_m2": 15742.9462, "severity_score": 0.2222, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 9, "severe_count": 2, "severe_pct": 22.22, "severe_area_m2": 2083.3328, "total_area_m2": 15742.9462, "severity_score": 0.2222, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 5, "severe_count": 3, "severe_pct": 60.0, "severe_area_m2": 9801.7269, "total_area_m2": 15360.8794, "severity_score": 0.6, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00001804_0919", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001804", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G20", "severity_score": 0.6, "building_count": 5, "severe_count": 3, "class_histogram": {"no_damage": 2, "minor_damage": 0, "major_damage": 0, "destroyed": 3, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G20", "row": 2, "col": 0, "building_count": 5, "severe_count": 3, "severe_pct": 60.0, "severe_area_m2": 9801.7269, "total_area_m2": 15360.8794, "severity_score": 0.6, "dominant_damage": "destroyed"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 9, "severe_count": 2, "severe_pct": 22.22, "severe_area_m2": 2083.3328, "total_area_m2": 15742.9462, "severity_score": 0.2222, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_portugal-wildfire_00001804_0920", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001804", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (portugal-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 5, "severe_footprint_m2": 11885.0597, "major_damage_area_m2": 0.0, "destroyed_area_m2": 11885.0597, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_portugal-wildfire_00001804_0921", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001804", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 5, "max_severe_in_cell": 3, "concentration_score": 0.6, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 2, "Q1": 0, "Q2": 3, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00001804_0922", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001804", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 14, "classified_buildings": 14, "counts": {"no-damage": 9, "minor-damage": 0, "major-damage": 0, "destroyed": 5, "unclassified": 0}, "severe_count": 5, "severe_pct": 35.71, "damaged_pct": 35.71}, "area_summary": {"total_footprint_m2": 31103.8256, "severe_footprint_m2": 11885.0597, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 5}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00000923_0923", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000923", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 20, "minor-damage": 0, "major-damage": 0, "destroyed": 1, "unclassified": 1}, "total_buildings": 22, "classified_buildings": 21, "damaged_count": 1, "severe_count": 1, "damaged_pct": 4.76, "severe_pct": 4.76, "severe_footprint_m2": 820.4356, "total_footprint_m2": 56766.2528, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00000923_0924", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000923", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 1, "severe_footprint_m2": 820.4356, "severe_pct_in_quadrant": 100.0, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 820.4356, "total_area_m2": 820.4356, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 21, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 55945.8172, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 21, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 55945.8172, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 820.4356, "total_area_m2": 820.4356, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00000923_0925", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000923", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G20", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G20", "row": 2, "col": 0, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 820.4356, "total_area_m2": 820.4356, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 13, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 34497.9729, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_portugal-wildfire_00000923_0926", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000923", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (portugal-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 1, "severe_footprint_m2": 820.4356, "major_damage_area_m2": 0.0, "destroyed_area_m2": 820.4356, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_portugal-wildfire_00000923_0927", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000923", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 1, "max_severe_in_cell": 1, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 1, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00000923_0928", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000923", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 22, "classified_buildings": 21, "counts": {"no-damage": 20, "minor-damage": 0, "major-damage": 0, "destroyed": 1, "unclassified": 1}, "severe_count": 1, "severe_pct": 4.76, "damaged_pct": 4.76}, "area_summary": {"total_footprint_m2": 56766.2528, "severe_footprint_m2": 820.4356, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 1}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_socal-fire_00000966_0929", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "socal-fire_00000966", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (socal-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 16, "minor-damage": 0, "major-damage": 0, "destroyed": 4, "unclassified": 0}, "total_buildings": 20, "classified_buildings": 20, "damaged_count": 4, "severe_count": 4, "damaged_pct": 20.0, "severe_pct": 20.0, "severe_footprint_m2": 29935.6188, "total_footprint_m2": 259061.6036, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_socal-fire_00000966_0930", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "socal-fire_00000966", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (socal-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 4, "severe_footprint_m2": 29935.6188, "severe_pct_in_quadrant": 26.67, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 15, "severe_count": 4, "severe_pct": 26.67, "severe_area_m2": 29935.6188, "total_area_m2": 181626.6297, "severity_score": 0.2667, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 29529.0003, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 15, "severe_count": 4, "severe_pct": 26.67, "severe_area_m2": 29935.6188, "total_area_m2": 181626.6297, "severity_score": 0.2667, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 29529.0003, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 47905.9736, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_socal-fire_00000966_0931", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "socal-fire_00000966", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 0.3636, "building_count": 11, "severe_count": 4, "class_histogram": {"no_damage": 7, "minor_damage": 0, "major_damage": 0, "destroyed": 4, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 11, "severe_count": 4, "severe_pct": 36.36, "severe_area_m2": 29935.6188, "total_area_m2": 56019.5418, "severity_score": 0.3636, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_socal-fire_00000966_0932", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "socal-fire_00000966", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (socal-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 4, "severe_footprint_m2": 29935.6188, "major_damage_area_m2": 0.0, "destroyed_area_m2": 29935.6188, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_socal-fire_00000966_0933", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "socal-fire_00000966", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 4, "max_severe_in_cell": 4, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 4, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_socal-fire_00000966_0934", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "socal-fire_00000966", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (socal-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "socal-fire", "disaster_type": "fire", "capture_date": "2018-11-14T18:42:58.000Z", "gsd": 2.5700748, "damage_summary": {"total_buildings": 20, "classified_buildings": 20, "counts": {"no-damage": 16, "minor-damage": 0, "major-damage": 0, "destroyed": 4, "unclassified": 0}, "severe_count": 4, "severe_pct": 20.0, "damaged_pct": 20.0}, "area_summary": {"total_footprint_m2": 259061.6036, "severe_footprint_m2": 29935.6188, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 4820, "polygon_severe_count": 4, "pixel_histogram": {"0": 1007966, "1": 35790, "2": 0, "3": 0, "4": 4820, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 88.131, "2": 0.0, "3": 0.0, "4": 11.869, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_socal-fire_00000436_0935", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "socal-fire_00000436", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (socal-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 5, "minor-damage": 0, "major-damage": 0, "destroyed": 13, "unclassified": 0}, "total_buildings": 18, "classified_buildings": 18, "damaged_count": 13, "severe_count": 13, "damaged_pct": 72.22, "severe_pct": 72.22, "severe_footprint_m2": 37870.0961, "total_footprint_m2": 62558.6483, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_socal-fire_00000436_0936", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "socal-fire_00000436", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (socal-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 13, "severe_footprint_m2": 37870.0961, "severe_pct_in_quadrant": 100.0, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 13, "severe_count": 13, "severe_pct": 100.0, "severe_area_m2": 37870.0961, "total_area_m2": 37870.0961, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 24688.5522, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 13, "severe_count": 13, "severe_pct": 100.0, "severe_area_m2": 37870.0961, "total_area_m2": 37870.0961, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 24688.5522, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_socal-fire_00000436_0937", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "socal-fire_00000436", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 1.0, "building_count": 4, "severe_count": 4, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 4, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 15112.6595, "total_area_m2": 15112.6595, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 17493.6395, "total_area_m2": 17493.6395, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 5263.7971, "total_area_m2": 5263.7971, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_socal-fire_00000436_0938", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "socal-fire_00000436", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (socal-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 13, "severe_footprint_m2": 37870.0961, "major_damage_area_m2": 0.0, "destroyed_area_m2": 37870.0961, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_socal-fire_00000436_0939", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "socal-fire_00000436", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 13, "max_severe_in_cell": 13, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 13, "Q1": 0, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_socal-fire_00000436_0940", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "socal-fire_00000436", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (socal-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "socal-fire", "disaster_type": "fire", "capture_date": "2018-11-14T18:42:58.000Z", "gsd": 2.5700748, "damage_summary": {"total_buildings": 18, "classified_buildings": 18, "counts": {"no-damage": 5, "minor-damage": 0, "major-damage": 0, "destroyed": 13, "unclassified": 0}, "severe_count": 13, "severe_pct": 72.22, "damaged_pct": 72.22}, "area_summary": {"total_footprint_m2": 62558.6483, "severe_footprint_m2": 37870.0961, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 6295, "polygon_severe_count": 13, "pixel_histogram": {"0": 1038335, "1": 3946, "2": 0, "3": 0, "4": 6295, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 38.5314, "2": 0.0, "3": 0.0, "4": 61.4686, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_woolsey-fire_00000441_0941", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000441", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (woolsey-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 101, "minor-damage": 0, "major-damage": 0, "destroyed": 1, "unclassified": 0}, "total_buildings": 102, "classified_buildings": 102, "damaged_count": 1, "severe_count": 1, "damaged_pct": 0.98, "severe_pct": 0.98, "severe_footprint_m2": 742.974, "total_footprint_m2": 245648.9665, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_woolsey-fire_00000441_0942", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000441", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (woolsey-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 1, "severe_footprint_m2": 742.974, "severe_pct_in_quadrant": 25.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 1, "severe_pct": 25.0, "severe_area_m2": 742.974, "total_area_m2": 8154.2523, "severity_score": 0.25, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 33, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 57212.7563, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 26, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 82342.4167, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 1, "severe_pct": 25.0, "severe_area_m2": 742.974, "total_area_m2": 8154.2523, "severity_score": 0.25, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 33, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 57212.7563, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 26, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 82342.4167, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 39, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 97939.5412, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_woolsey-fire_00000441_0943", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000441", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 742.974, "total_area_m2": 742.974, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 7023.1015, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 16149.5953, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_woolsey-fire_00000441_0944", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000441", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (woolsey-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 1, "severe_footprint_m2": 742.974, "major_damage_area_m2": 0.0, "destroyed_area_m2": 742.974, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_woolsey-fire_00000441_0945", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000441", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 1, "max_severe_in_cell": 1, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 1, "Q1": 0, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_woolsey-fire_00000441_0946", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000441", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (woolsey-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "woolsey-fire", "disaster_type": "fire", "capture_date": "2018-11-18T18:44:07.000Z", "gsd": 1.7726423000000002, "damage_summary": {"total_buildings": 102, "classified_buildings": 102, "counts": {"no-damage": 101, "minor-damage": 0, "major-damage": 0, "destroyed": 1, "unclassified": 0}, "severe_count": 1, "severe_pct": 0.98, "damaged_pct": 0.98}, "area_summary": {"total_footprint_m2": 245648.9665, "severe_footprint_m2": 742.974, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 1}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_palu-tsunami_00000000_0947", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000000", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a tsunami event (palu-tsunami) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 81, "minor-damage": 0, "major-damage": 1, "destroyed": 9, "unclassified": 0}, "total_buildings": 91, "classified_buildings": 91, "damaged_count": 10, "severe_count": 10, "damaged_pct": 10.99, "severe_pct": 10.99, "severe_footprint_m2": 24289.6055, "total_footprint_m2": 515864.1082, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_palu-tsunami_00000000_0948", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000000", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster tsunami scene (palu-tsunami), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 6, "severe_footprint_m2": 12876.2887, "severe_pct_in_quadrant": 85.71, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 7, "severe_count": 6, "severe_pct": 85.71, "severe_area_m2": 12876.2887, "total_area_m2": 14161.8662, "severity_score": 0.81, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 5, "severe_count": 1, "severe_pct": 20.0, "severe_area_m2": 9669.3285, "total_area_m2": 57600.4747, "severity_score": 0.2, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 43, "severe_count": 3, "severe_pct": 6.98, "severe_area_m2": 1743.9883, "total_area_m2": 248982.5583, "severity_score": 0.0698, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 36, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 195119.209, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 7, "severe_count": 6, "severe_pct": 85.71, "severe_area_m2": 12876.2887, "total_area_m2": 14161.8662, "severity_score": 0.81, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 5, "severe_count": 1, "severe_pct": 20.0, "severe_area_m2": 9669.3285, "total_area_m2": 57600.4747, "severity_score": 0.2, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 43, "severe_count": 3, "severe_pct": 6.98, "severe_area_m2": 1743.9883, "total_area_m2": 248982.5583, "severity_score": 0.0698, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_palu-tsunami_00000000_0949", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000000", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster tsunami satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 1.0, "building_count": 3, "severe_count": 3, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 3, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 6200.2134, "total_area_m2": 6200.2134, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 4, "severe_count": 3, "severe_pct": 75.0, "severe_area_m2": 6676.0753, "total_area_m2": 7961.6528, "severity_score": 0.6675, "dominant_damage": "destroyed"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 1156.8607, "total_area_m2": 4090.0317, "severity_score": 0.6667, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_palu-tsunami_00000000_0950", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000000", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster tsunami scene (palu-tsunami), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 10, "severe_footprint_m2": 24289.6055, "major_damage_area_m2": 3466.1856, "destroyed_area_m2": 20823.4199, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_palu-tsunami_00000000_0951", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000000", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster tsunami scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 10, "max_severe_in_cell": 6, "concentration_score": 0.6, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 6, "Q2": 1, "Q3": 3}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_palu-tsunami_00000000_0952", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000000", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster tsunami scene (palu-tsunami). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "palu-tsunami", "disaster_type": "tsunami", "capture_date": "2018-10-01T02:26:02.000Z", "gsd": 2.8247435, "damage_summary": {"total_buildings": 91, "classified_buildings": 91, "counts": {"no-damage": 81, "minor-damage": 0, "major-damage": 1, "destroyed": 9, "unclassified": 0}, "severe_count": 10, "severe_pct": 10.99, "damaged_pct": 10.99}, "area_summary": {"total_footprint_m2": 515864.1082, "severe_footprint_m2": 24289.6055, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 3284, "polygon_severe_count": 10, "pixel_histogram": {"0": 979974, "1": 65318, "2": 0, "3": 480, "4": 2804, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 95.213, "2": 0.0, "3": 0.6997, "4": 4.0873, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_moore-tornado_00000056_0959", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "moore-tornado_00000056", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a wind event (moore-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 175, "minor-damage": 37, "major-damage": 12, "destroyed": 47, "unclassified": 0}, "total_buildings": 271, "classified_buildings": 271, "damaged_count": 96, "severe_count": 59, "damaged_pct": 35.42, "severe_pct": 21.77, "severe_footprint_m2": 167613.8527, "total_footprint_m2": 693308.4913, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_moore-tornado_00000056_0960", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "moore-tornado_00000056", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster wind scene (moore-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 31, "severe_footprint_m2": 97275.4761, "severe_pct_in_quadrant": 49.21, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 63, "severe_count": 31, "severe_pct": 49.21, "severe_area_m2": 97275.4761, "total_area_m2": 190708.7004, "severity_score": 0.513, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 61, "severe_count": 28, "severe_pct": 45.9, "severe_area_m2": 70338.3766, "total_area_m2": 166512.9881, "severity_score": 0.5077, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 77, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 180673.4807, "severity_score": 0.0343, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 63, "severe_count": 31, "severe_pct": 49.21, "severe_area_m2": 97275.4761, "total_area_m2": 190708.7004, "severity_score": 0.513, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 61, "severe_count": 28, "severe_pct": 45.9, "severe_area_m2": 70338.3766, "total_area_m2": 166512.9881, "severity_score": 0.5077, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 77, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 180673.4807, "severity_score": 0.0343, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 70, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 155413.3221, "severity_score": 0.0189, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_moore-tornado_00000056_0961", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "moore-tornado_00000056", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 0.9333, "building_count": 15, "severe_count": 14, "class_histogram": {"no_damage": 0, "minor_damage": 1, "major_damage": 1, "destroyed": 13, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 15, "severe_count": 14, "severe_pct": 93.33, "severe_area_m2": 51567.9908, "total_area_m2": 54969.501, "severity_score": 0.9333, "dominant_damage": "destroyed"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 10, "severe_count": 10, "severe_pct": 100.0, "severe_area_m2": 26577.6532, "total_area_m2": 26577.6532, "severity_score": 0.868, "dominant_damage": "destroyed"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 13, "severe_count": 11, "severe_pct": 84.62, "severe_area_m2": 35122.9241, "total_area_m2": 40779.6159, "severity_score": 0.8462, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_moore-tornado_00000056_0962", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "moore-tornado_00000056", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster wind scene (moore-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 59, "severe_footprint_m2": 167613.8527, "major_damage_area_m2": 37425.3159, "destroyed_area_m2": 130188.5368, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_moore-tornado_00000056_0963", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "moore-tornado_00000056", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 59, "max_severe_in_cell": 31, "concentration_score": 0.5254, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 31, "Q1": 28, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_moore-tornado_00000056_0964", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "moore-tornado_00000056", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (moore-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "moore-tornado", "disaster_type": "wind", "capture_date": "2013-05-22T17:26:30.085Z", "gsd": 1.72984850406647, "damage_summary": {"total_buildings": 271, "classified_buildings": 271, "counts": {"no-damage": 175, "minor-damage": 37, "major-damage": 12, "destroyed": 47, "unclassified": 0}, "severe_count": 59, "severe_pct": 21.77, "damaged_pct": 35.42}, "area_summary": {"total_footprint_m2": 693308.4913, "severe_footprint_m2": 167613.8527, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5254, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 59}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_tuscaloosa-tornado_00000324_0965", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000324", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a wind event (tuscaloosa-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 62, "minor-damage": 1, "major-damage": 0, "destroyed": 4, "unclassified": 6}, "total_buildings": 73, "classified_buildings": 67, "damaged_count": 5, "severe_count": 4, "damaged_pct": 7.46, "severe_pct": 5.97, "severe_footprint_m2": 3005.1628, "total_footprint_m2": 238463.7435, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_tuscaloosa-tornado_00000324_0966", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000324", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster wind scene (tuscaloosa-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 2, "severe_footprint_m2": 1969.8396, "severe_pct_in_quadrant": 6.45, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 31, "severe_count": 2, "severe_pct": 6.45, "severe_area_m2": 1969.8396, "total_area_m2": 114065.8749, "severity_score": 0.0645, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 13, "severe_count": 2, "severe_pct": 15.38, "severe_area_m2": 1035.3232, "total_area_m2": 29055.0568, "severity_score": 0.1538, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 39942.5445, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 13, "severe_count": 2, "severe_pct": 15.38, "severe_area_m2": 1035.3232, "total_area_m2": 29055.0568, "severity_score": 0.1538, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 39942.5445, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 18, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 55400.2673, "severity_score": 0.0183, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 31, "severe_count": 2, "severe_pct": 6.45, "severe_area_m2": 1969.8396, "total_area_m2": 114065.8749, "severity_score": 0.0645, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_tuscaloosa-tornado_00000324_0967", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000324", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.4, "building_count": 5, "severe_count": 2, "class_histogram": {"no_damage": 3, "minor_damage": 0, "major_damage": 0, "destroyed": 2, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 5, "severe_count": 2, "severe_pct": 40.0, "severe_area_m2": 1035.3232, "total_area_m2": 11550.7934, "severity_score": 0.4, "dominant_damage": "no-damage"}, {"cell_id": "G33", "row": 3, "col": 3, "building_count": 10, "severe_count": 2, "severe_pct": 20.0, "severe_area_m2": 1969.8396, "total_area_m2": 60061.7355, "severity_score": 0.2, "dominant_damage": "no-damage"}, {"cell_id": "G21", "row": 2, "col": 1, "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 15376.2276, "severity_score": 0.0471, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_tuscaloosa-tornado_00000324_0968", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000324", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster wind scene (tuscaloosa-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 4, "severe_footprint_m2": 3005.1628, "major_damage_area_m2": 0.0, "destroyed_area_m2": 3005.1628, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_tuscaloosa-tornado_00000324_0969", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000324", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 4, "max_severe_in_cell": 2, "concentration_score": 0.5, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 2, "Q1": 0, "Q2": 0, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_tuscaloosa-tornado_00000324_0970", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000324", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (tuscaloosa-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "tuscaloosa-tornado", "disaster_type": "wind", "capture_date": "2011-05-19T16:48:01.085Z", "gsd": 1.85690224170685, "damage_summary": {"total_buildings": 73, "classified_buildings": 67, "counts": {"no-damage": 62, "minor-damage": 1, "major-damage": 0, "destroyed": 4, "unclassified": 6}, "severe_count": 4, "severe_pct": 5.97, "damaged_pct": 7.46}, "area_summary": {"total_footprint_m2": 238463.7435, "severe_footprint_m2": 3005.1628, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 4}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-matthew_00000322_0971", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000322", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a wind event (hurricane-matthew) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 11, "major-damage": 3, "destroyed": 13, "unclassified": 0}, "total_buildings": 27, "classified_buildings": 27, "damaged_count": 27, "severe_count": 16, "damaged_pct": 100.0, "severe_pct": 59.26, "severe_footprint_m2": 7700.969, "total_footprint_m2": 13601.7877, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-matthew_00000322_0972", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000322", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster wind scene (hurricane-matthew), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 5, "severe_footprint_m2": 2579.7259, "severe_pct_in_quadrant": 71.43, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 7, "severe_count": 5, "severe_pct": 71.43, "severe_area_m2": 2579.7259, "total_area_m2": 3667.8201, "severity_score": 0.8086, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 10, "severe_count": 4, "severe_pct": 40.0, "severe_area_m2": 2043.2786, "total_area_m2": 5279.9604, "severity_score": 0.565, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 1852.0604, "total_area_m2": 1852.0604, "severity_score": 0.9175, "dominant_damage": "destroyed"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 6, "severe_count": 3, "severe_pct": 50.0, "severe_area_m2": 1225.9041, "total_area_m2": 2801.9468, "severity_score": 0.61, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 1852.0604, "total_area_m2": 1852.0604, "severity_score": 0.9175, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 5, "severe_pct": 71.43, "severe_area_m2": 2579.7259, "total_area_m2": 3667.8201, "severity_score": 0.8086, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 10, "severe_count": 4, "severe_pct": 40.0, "severe_area_m2": 2043.2786, "total_area_m2": 5279.9604, "severity_score": 0.565, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-matthew_00000322_0973", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000322", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 289.2266, "total_area_m2": 289.2266, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G03", "row": 0, "col": 3, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 799.8909, "total_area_m2": 799.8909, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G21", "row": 2, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 162.0301, "total_area_m2": 162.0301, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-matthew_00000322_0974", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000322", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster wind scene (hurricane-matthew), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 16, "severe_footprint_m2": 7700.969, "major_damage_area_m2": 1996.979, "destroyed_area_m2": 5703.99, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-matthew_00000322_0975", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000322", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 16, "max_severe_in_cell": 5, "concentration_score": 0.3125, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 3, "Q1": 4, "Q2": 5, "Q3": 4}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-matthew_00000322_0976", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000322", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-matthew). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-matthew", "disaster_type": "wind", "capture_date": "2016-10-10T16:04:56.646Z", "gsd": 1.59234964847565, "damage_summary": {"total_buildings": 27, "classified_buildings": 27, "counts": {"no-damage": 0, "minor-damage": 11, "major-damage": 3, "destroyed": 13, "unclassified": 0}, "severe_count": 16, "severe_pct": 59.26, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 13601.7877, "severe_footprint_m2": 7700.969, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.3125, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 3444, "polygon_severe_count": 16, "pixel_histogram": {"0": 1042474, "1": 0, "2": 2658, "3": 880, "4": 2564, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 43.5595, "3": 14.4215, "4": 42.019, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00000795_0977", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000795", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 31, "minor-damage": 0, "major-damage": 0, "destroyed": 2, "unclassified": 0}, "total_buildings": 33, "classified_buildings": 33, "damaged_count": 2, "severe_count": 2, "damaged_pct": 6.06, "severe_pct": 6.06, "severe_footprint_m2": 2911.5751, "total_footprint_m2": 231395.5234, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00000795_0978", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000795", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 1, "severe_footprint_m2": 1584.9965, "severe_pct_in_quadrant": 5.26, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 19, "severe_count": 1, "severe_pct": 5.26, "severe_area_m2": 1584.9965, "total_area_m2": 116462.4466, "severity_score": 0.0526, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 1326.5786, "total_area_m2": 8627.7631, "severity_score": 0.3333, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 75003.9843, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 75003.9843, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 19, "severe_count": 1, "severe_pct": 5.26, "severe_area_m2": 1584.9965, "total_area_m2": 116462.4466, "severity_score": 0.0526, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 1326.5786, "total_area_m2": 8627.7631, "severity_score": 0.3333, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 31301.3294, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00000795_0979", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000795", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G31", "severity_score": 0.5, "building_count": 2, "severe_count": 1, "class_histogram": {"no_damage": 1, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G31", "row": 3, "col": 1, "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 1326.5786, "total_area_m2": 3815.9167, "severity_score": 0.5, "dominant_damage": "no-damage"}, {"cell_id": "G03", "row": 0, "col": 3, "building_count": 10, "severe_count": 1, "severe_pct": 10.0, "severe_area_m2": 1584.9965, "total_area_m2": 39361.8912, "severity_score": 0.1, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_portugal-wildfire_00000795_0980", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000795", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (portugal-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 2, "severe_footprint_m2": 2911.5751, "major_damage_area_m2": 0.0, "destroyed_area_m2": 2911.5751, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_portugal-wildfire_00000795_0981", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000795", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 2, "max_severe_in_cell": 1, "concentration_score": 0.5, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 1, "Q2": 1, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00000795_0982", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000795", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 33, "classified_buildings": 33, "counts": {"no-damage": 31, "minor-damage": 0, "major-damage": 0, "destroyed": 2, "unclassified": 0}, "severe_count": 2, "severe_pct": 6.06, "damaged_pct": 6.06}, "area_summary": {"total_footprint_m2": 231395.5234, "severe_footprint_m2": 2911.5751, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 2}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_lower-puna-volcano_00000135_0983", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000135", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a volcano event (lower-puna-volcano) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 1, "minor-damage": 0, "major-damage": 0, "destroyed": 30, "unclassified": 0}, "total_buildings": 31, "classified_buildings": 31, "damaged_count": 30, "severe_count": 30, "damaged_pct": 96.77, "severe_pct": 96.77, "severe_footprint_m2": 64773.7271, "total_footprint_m2": 72774.972, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_lower-puna-volcano_00000135_0984", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000135", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster volcano scene (lower-puna-volcano), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 12, "severe_footprint_m2": 23630.1366, "severe_pct_in_quadrant": 100.0, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 12, "severe_count": 12, "severe_pct": 100.0, "severe_area_m2": 23630.1366, "total_area_m2": 23630.1366, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 14652.1905, "total_area_m2": 14652.1905, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 7, "severe_count": 7, "severe_pct": 100.0, "severe_area_m2": 13720.9988, "total_area_m2": 13720.9988, "severity_score": 1.0, "dominant_damage": "destroyed"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 7, "severe_count": 7, "severe_pct": 100.0, "severe_area_m2": 13720.9988, "total_area_m2": 13720.9988, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 14652.1905, "total_area_m2": 14652.1905, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 12, "severe_count": 12, "severe_pct": 100.0, "severe_area_m2": 23630.1366, "total_area_m2": 23630.1366, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 4, "severe_count": 3, "severe_pct": 75.0, "severe_area_m2": 12770.4012, "total_area_m2": 20771.6461, "severity_score": 0.75, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_lower-puna-volcano_00000135_0985", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000135", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster volcano satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G10", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G10", "row": 1, "col": 0, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 744.7971, "total_area_m2": 744.7971, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 12976.2017, "total_area_m2": 12976.2017, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 13631.8516, "total_area_m2": 13631.8516, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_lower-puna-volcano_00000135_0986", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000135", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster volcano scene (lower-puna-volcano), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 30, "severe_footprint_m2": 64773.7271, "major_damage_area_m2": 0.0, "destroyed_area_m2": 64773.7271, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_lower-puna-volcano_00000135_0987", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000135", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster volcano scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 30, "max_severe_in_cell": 12, "concentration_score": 0.4, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 7, "Q1": 8, "Q2": 12, "Q3": 3}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_lower-puna-volcano_00000135_0988", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000135", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster volcano scene (lower-puna-volcano). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "lower-puna-volcano", "disaster_type": "volcano", "capture_date": "2018-05-23T20:59:21.000Z", "gsd": 2.2423599, "damage_summary": {"total_buildings": 31, "classified_buildings": 31, "counts": {"no-damage": 1, "minor-damage": 0, "major-damage": 0, "destroyed": 30, "unclassified": 0}, "severe_count": 30, "severe_pct": 96.77, "damaged_pct": 96.77}, "area_summary": {"total_footprint_m2": 72774.972, "severe_footprint_m2": 64773.7271, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.4, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 30}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_socal-fire_00000661_0989", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "socal-fire_00000661", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (socal-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 21, "minor-damage": 0, "major-damage": 0, "destroyed": 4, "unclassified": 5}, "total_buildings": 30, "classified_buildings": 25, "damaged_count": 4, "severe_count": 4, "damaged_pct": 16.0, "severe_pct": 16.0, "severe_footprint_m2": 10762.0231, "total_footprint_m2": 168345.2916, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_socal-fire_00000661_0990", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "socal-fire_00000661", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (socal-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 2, "severe_footprint_m2": 8242.0703, "severe_pct_in_quadrant": 11.11, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 18, "severe_count": 2, "severe_pct": 11.11, "severe_area_m2": 8242.0703, "total_area_m2": 57736.8986, "severity_score": 0.1111, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 4, "severe_count": 2, "severe_pct": 50.0, "severe_area_m2": 2519.9528, "total_area_m2": 25993.1957, "severity_score": 0.5, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 2, "severe_pct": 50.0, "severe_area_m2": 2519.9528, "total_area_m2": 25993.1957, "severity_score": 0.5, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 84615.1973, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 18, "severe_count": 2, "severe_pct": 11.11, "severe_area_m2": 8242.0703, "total_area_m2": 57736.8986, "severity_score": 0.1111, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_socal-fire_00000661_0991", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "socal-fire_00000661", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 1.0, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 2, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 2519.9528, "total_area_m2": 2519.9528, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G33", "row": 3, "col": 3, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 8242.0703, "total_area_m2": 8242.0703, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_socal-fire_00000661_0992", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "socal-fire_00000661", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (socal-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 4, "severe_footprint_m2": 10762.0231, "major_damage_area_m2": 0.0, "destroyed_area_m2": 10762.0231, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_socal-fire_00000661_0993", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "socal-fire_00000661", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 4, "max_severe_in_cell": 2, "concentration_score": 0.5, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 2, "Q1": 0, "Q2": 0, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_socal-fire_00000661_0994", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "socal-fire_00000661", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (socal-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "socal-fire", "disaster_type": "fire", "capture_date": "2018-11-14T18:42:58.000Z", "gsd": 2.5700748, "damage_summary": {"total_buildings": 30, "classified_buildings": 25, "counts": {"no-damage": 21, "minor-damage": 0, "major-damage": 0, "destroyed": 4, "unclassified": 5}, "severe_count": 4, "severe_pct": 16.0, "damaged_pct": 16.0}, "area_summary": {"total_footprint_m2": 168345.2916, "severe_footprint_m2": 10762.0231, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 1796, "polygon_severe_count": 4, "pixel_histogram": {"0": 1022557, "1": 24223, "2": 0, "3": 0, "4": 1796, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 93.0974, "2": 0.0, "3": 0.0, "4": 6.9026, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000391_0995", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000391", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 3, "major-damage": 2, "destroyed": 14, "unclassified": 2}, "total_buildings": 21, "classified_buildings": 19, "damaged_count": 19, "severe_count": 16, "damaged_pct": 100.0, "severe_pct": 84.21, "severe_footprint_m2": 19142.4327, "total_footprint_m2": 25145.5658, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000391_0996", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000391", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 14, "severe_footprint_m2": 17876.4207, "severe_pct_in_quadrant": 73.68, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 19, "severe_count": 14, "severe_pct": 73.68, "severe_area_m2": 17876.4207, "total_area_m2": 23879.5538, "severity_score": 0.7889, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 1266.012, "total_area_m2": 1266.012, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 19, "severe_count": 14, "severe_pct": 73.68, "severe_area_m2": 17876.4207, "total_area_m2": 23879.5538, "severity_score": 0.7889, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 1266.012, "total_area_m2": 1266.012, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000391_0997", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000391", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G13", "severity_score": 1.0, "building_count": 9, "severe_count": 9, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 9, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G13", "row": 1, "col": 3, "building_count": 9, "severe_count": 9, "severe_pct": 100.0, "severe_area_m2": 6053.6362, "total_area_m2": 6053.6362, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G22", "row": 2, "col": 2, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 956.0472, "total_area_m2": 956.0472, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 309.9648, "total_area_m2": 309.9648, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000391_0998", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000391", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 16, "severe_footprint_m2": 19142.4327, "major_damage_area_m2": 1266.012, "destroyed_area_m2": 17876.4207, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000391_0999", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000391", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 16, "max_severe_in_cell": 14, "concentration_score": 0.875, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 14, "Q2": 0, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000391_1000", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000391", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 21, "classified_buildings": 19, "counts": {"no-damage": 0, "minor-damage": 3, "major-damage": 2, "destroyed": 14, "unclassified": 2}, "severe_count": 16, "severe_pct": 84.21, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 25145.5658, "severe_footprint_m2": 19142.4327, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.875, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 16}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_pinery-bushfire_00000690_1001", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000690", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (pinery-bushfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 6, "minor-damage": 0, "major-damage": 0, "destroyed": 3, "unclassified": 1}, "total_buildings": 10, "classified_buildings": 9, "damaged_count": 3, "severe_count": 3, "damaged_pct": 33.33, "severe_pct": 33.33, "severe_footprint_m2": 4114.8143, "total_footprint_m2": 11544.9264, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_pinery-bushfire_00000690_1002", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000690", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (pinery-bushfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 3, "severe_footprint_m2": 4114.8143, "severe_pct_in_quadrant": 30.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 10, "severe_count": 3, "severe_pct": 30.0, "severe_area_m2": 4114.8143, "total_area_m2": 11544.9264, "severity_score": 0.3, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 10, "severe_count": 3, "severe_pct": 30.0, "severe_area_m2": 4114.8143, "total_area_m2": 11544.9264, "severity_score": 0.3, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_pinery-bushfire_00000690_1003", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000690", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.3, "building_count": 10, "severe_count": 3, "class_histogram": {"no_damage": 6, "minor_damage": 0, "major_damage": 0, "destroyed": 3, "unclassified": 1}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 10, "severe_count": 3, "severe_pct": 30.0, "severe_area_m2": 4114.8143, "total_area_m2": 11544.9264, "severity_score": 0.3, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_pinery-bushfire_00000690_1004", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000690", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (pinery-bushfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 3, "severe_footprint_m2": 4114.8143, "major_damage_area_m2": 0.0, "destroyed_area_m2": 4114.8143, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_pinery-bushfire_00000690_1005", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000690", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 3, "max_severe_in_cell": 3, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 3, "Q1": 0, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_pinery-bushfire_00000690_1006", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000690", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (pinery-bushfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "pinery-bushfire", "disaster_type": "fire", "capture_date": "2015-11-29T01:09:22.998Z", "gsd": 1.65928483009338, "damage_summary": {"total_buildings": 10, "classified_buildings": 9, "counts": {"no-damage": 6, "minor-damage": 0, "major-damage": 0, "destroyed": 3, "unclassified": 1}, "severe_count": 3, "severe_pct": 33.33, "damaged_pct": 33.33}, "area_summary": {"total_footprint_m2": 11544.9264, "severe_footprint_m2": 4114.8143, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 3}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00001093_1013", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001093", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 143, "minor-damage": 0, "major-damage": 0, "destroyed": 2, "unclassified": 2}, "total_buildings": 147, "classified_buildings": 145, "damaged_count": 2, "severe_count": 2, "damaged_pct": 1.38, "severe_pct": 1.38, "severe_footprint_m2": 5882.978, "total_footprint_m2": 652755.9591, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00001093_1014", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001093", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 2, "severe_footprint_m2": 5882.978, "severe_pct_in_quadrant": 13.33, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 15, "severe_count": 2, "severe_pct": 13.33, "severe_area_m2": 5882.978, "total_area_m2": 57237.9493, "severity_score": 0.1333, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 36, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 97227.9886, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 46, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 330587.0649, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 36, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 97227.9886, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 46, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 330587.0649, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 50, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 167702.9563, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 15, "severe_count": 2, "severe_pct": 13.33, "severe_area_m2": 5882.978, "total_area_m2": 57237.9493, "severity_score": 0.1333, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00001093_1015", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001093", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G22", "severity_score": 0.3333, "building_count": 3, "severe_count": 1, "class_histogram": {"no_damage": 2, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G22", "row": 2, "col": 2, "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 5636.2199, "total_area_m2": 14073.7368, "severity_score": 0.3333, "dominant_damage": "no-damage"}, {"cell_id": "G33", "row": 3, "col": 3, "building_count": 9, "severe_count": 1, "severe_pct": 11.11, "severe_area_m2": 246.7581, "total_area_m2": 39144.1752, "severity_score": 0.1111, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 29741.7159, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_portugal-wildfire_00001093_1016", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001093", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (portugal-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 2, "severe_footprint_m2": 5882.978, "major_damage_area_m2": 0.0, "destroyed_area_m2": 5882.978, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_portugal-wildfire_00001093_1017", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001093", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 2, "max_severe_in_cell": 2, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 0, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00001093_1018", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001093", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 147, "classified_buildings": 145, "counts": {"no-damage": 143, "minor-damage": 0, "major-damage": 0, "destroyed": 2, "unclassified": 2}, "severe_count": 2, "severe_pct": 1.38, "damaged_pct": 1.38}, "area_summary": {"total_footprint_m2": 652755.9591, "severe_footprint_m2": 5882.978, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 2}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_joplin-tornado_00000075_1025", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000075", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a wind event (joplin-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 7, "minor-damage": 15, "major-damage": 7, "destroyed": 45, "unclassified": 12}, "total_buildings": 86, "classified_buildings": 74, "damaged_count": 67, "severe_count": 52, "damaged_pct": 90.54, "severe_pct": 70.27, "severe_footprint_m2": 262303.9588, "total_footprint_m2": 392620.1626, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_joplin-tornado_00000075_1026", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000075", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster wind scene (joplin-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 30, "severe_footprint_m2": 150345.2682, "severe_pct_in_quadrant": 93.75, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 32, "severe_count": 30, "severe_pct": 93.75, "severe_area_m2": 150345.2682, "total_area_m2": 164052.9862, "severity_score": 0.9066, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 26, "severe_count": 16, "severe_pct": 61.54, "severe_area_m2": 91185.3865, "total_area_m2": 124413.0032, "severity_score": 0.6535, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 14, "severe_count": 4, "severe_pct": 28.57, "severe_area_m2": 14224.4425, "total_area_m2": 60582.4356, "severity_score": 0.4271, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 26, "severe_count": 16, "severe_pct": 61.54, "severe_area_m2": 91185.3865, "total_area_m2": 124413.0032, "severity_score": 0.6535, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 14, "severe_count": 4, "severe_pct": 28.57, "severe_area_m2": 14224.4425, "total_area_m2": 60582.4356, "severity_score": 0.4271, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 32, "severe_count": 30, "severe_pct": 93.75, "severe_area_m2": 150345.2682, "total_area_m2": 164052.9862, "severity_score": 0.9066, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 14, "severe_count": 2, "severe_pct": 14.29, "severe_area_m2": 6548.8616, "total_area_m2": 43571.7376, "severity_score": 0.19, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_joplin-tornado_00000075_1027", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000075", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G30", "severity_score": 1.0, "building_count": 9, "severe_count": 9, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 9, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G30", "row": 3, "col": 0, "building_count": 9, "severe_count": 9, "severe_pct": 100.0, "severe_area_m2": 31060.0138, "total_area_m2": 31060.0138, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G20", "row": 2, "col": 0, "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 39971.9123, "total_area_m2": 39971.9123, "severity_score": 0.9587, "dominant_damage": "destroyed"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 7, "severe_count": 7, "severe_pct": 100.0, "severe_area_m2": 36072.6531, "total_area_m2": 36072.6531, "severity_score": 0.9057, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_joplin-tornado_00000075_1028", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000075", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster wind scene (joplin-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 52, "severe_footprint_m2": 262303.9588, "major_damage_area_m2": 59614.7392, "destroyed_area_m2": 202689.2196, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_joplin-tornado_00000075_1029", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000075", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 52, "max_severe_in_cell": 30, "concentration_score": 0.5769, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 16, "Q1": 4, "Q2": 30, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_joplin-tornado_00000075_1030", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000075", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (joplin-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "joplin-tornado", "disaster_type": "wind", "capture_date": "2011-05-29T17:29:25.433Z", "gsd": 2.35255861282349, "damage_summary": {"total_buildings": 86, "classified_buildings": 74, "counts": {"no-damage": 7, "minor-damage": 15, "major-damage": 7, "destroyed": 45, "unclassified": 12}, "severe_count": 52, "severe_pct": 70.27, "damaged_pct": 90.54}, "area_summary": {"total_footprint_m2": 392620.1626, "severe_footprint_m2": 262303.9588, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5769, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 52}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_palu-tsunami_00000019_1031", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000019", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a tsunami event (palu-tsunami) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 130, "minor-damage": 0, "major-damage": 8, "destroyed": 103, "unclassified": 53}, "total_buildings": 294, "classified_buildings": 241, "damaged_count": 111, "severe_count": 111, "damaged_pct": 46.06, "severe_pct": 46.06, "severe_footprint_m2": 264507.2173, "total_footprint_m2": 897151.672, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_palu-tsunami_00000019_1032", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000019", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster tsunami scene (palu-tsunami), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 102, "severe_footprint_m2": 244492.445, "severe_pct_in_quadrant": 58.29, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 175, "severe_count": 102, "severe_pct": 58.29, "severe_area_m2": 244492.445, "total_area_m2": 428814.5419, "severity_score": 0.5791, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 107, "severe_count": 9, "severe_pct": 8.41, "severe_area_m2": 20014.7723, "total_area_m2": 312329.8442, "severity_score": 0.0656, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 5595.3046, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 175, "severe_count": 102, "severe_pct": 58.29, "severe_area_m2": 244492.445, "total_area_m2": 428814.5419, "severity_score": 0.5791, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 107, "severe_count": 9, "severe_pct": 8.41, "severe_area_m2": 20014.7723, "total_area_m2": 312329.8442, "severity_score": 0.0656, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 5595.3046, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 150411.9813, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_palu-tsunami_00000019_1033", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000019", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster tsunami satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G10", "severity_score": 1.0, "building_count": 81, "severe_count": 81, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 81, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G10", "row": 1, "col": 0, "building_count": 81, "severe_count": 81, "severe_pct": 100.0, "severe_area_m2": 149496.2013, "total_area_m2": 149496.2013, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 59099.6277, "total_area_m2": 59099.6277, "severity_score": 0.945, "dominant_damage": "destroyed"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 38, "severe_count": 7, "severe_pct": 18.42, "severe_area_m2": 15008.7725, "total_area_m2": 96968.532, "severity_score": 0.1842, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_palu-tsunami_00000019_1034", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000019", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster tsunami scene (palu-tsunami), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 111, "severe_footprint_m2": 264507.2173, "major_damage_area_m2": 67785.4374, "destroyed_area_m2": 196721.7799, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_palu-tsunami_00000019_1035", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000019", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster tsunami scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 111, "max_severe_in_cell": 102, "concentration_score": 0.9189, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 102, "Q1": 9, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_palu-tsunami_00000019_1036", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000019", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster tsunami scene (palu-tsunami). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "palu-tsunami", "disaster_type": "tsunami", "capture_date": "2018-10-01T02:26:02.000Z", "gsd": 2.8247435, "damage_summary": {"total_buildings": 294, "classified_buildings": 241, "counts": {"no-damage": 130, "minor-damage": 0, "major-damage": 8, "destroyed": 103, "unclassified": 53}, "severe_count": 111, "severe_pct": 46.06, "damaged_pct": 46.06}, "area_summary": {"total_footprint_m2": 897151.672, "severe_footprint_m2": 264507.2173, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.9189, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 35988, "polygon_severe_count": 111, "pixel_histogram": {"0": 955450, "1": 57138, "2": 0, "3": 8841, "4": 27147, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 61.3556, "2": 0.0, "3": 9.4936, "4": 29.1508, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00001016_1037", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001016", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 17, "minor-damage": 2, "major-damage": 0, "destroyed": 1, "unclassified": 3}, "total_buildings": 23, "classified_buildings": 20, "damaged_count": 3, "severe_count": 1, "damaged_pct": 15.0, "severe_pct": 5.0, "severe_footprint_m2": 1657.897, "total_footprint_m2": 53776.242, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00001016_1038", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001016", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 1, "severe_footprint_m2": 1657.897, "severe_pct_in_quadrant": 5.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 20, "severe_count": 1, "severe_pct": 5.0, "severe_area_m2": 1657.897, "total_area_m2": 45428.6086, "severity_score": 0.0665, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 8347.6334, "severity_score": 0.11, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 8347.6334, "severity_score": 0.11, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 20, "severe_count": 1, "severe_pct": 5.0, "severe_area_m2": 1657.897, "total_area_m2": 45428.6086, "severity_score": 0.0665, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00001016_1039", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001016", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G12", "severity_score": 0.11, "building_count": 3, "severe_count": 0, "class_histogram": {"no_damage": 1, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 1}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G12", "row": 1, "col": 2, "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 8347.6334, "severity_score": 0.11, "dominant_damage": "no-damage"}, {"cell_id": "G22", "row": 2, "col": 2, "building_count": 16, "severe_count": 1, "severe_pct": 6.25, "severe_area_m2": 1657.897, "total_area_m2": 34589.5939, "severity_score": 0.0831, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_portugal-wildfire_00001016_1040", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001016", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (portugal-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 1, "severe_footprint_m2": 1657.897, "major_damage_area_m2": 0.0, "destroyed_area_m2": 1657.897, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_portugal-wildfire_00001016_1041", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001016", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 1, "max_severe_in_cell": 1, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 0, "Q3": 1}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00001016_1042", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001016", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 23, "classified_buildings": 20, "counts": {"no-damage": 17, "minor-damage": 2, "major-damage": 0, "destroyed": 1, "unclassified": 3}, "severe_count": 1, "severe_pct": 5.0, "damaged_pct": 15.0}, "area_summary": {"total_footprint_m2": 53776.242, "severe_footprint_m2": 1657.897, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 1}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-matthew_00000233_1043", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000233", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a wind event (hurricane-matthew) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 6, "major-damage": 0, "destroyed": 6, "unclassified": 0}, "total_buildings": 12, "classified_buildings": 12, "damaged_count": 12, "severe_count": 6, "damaged_pct": 100.0, "severe_pct": 50.0, "severe_footprint_m2": 3618.9793, "total_footprint_m2": 7175.7533, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-matthew_00000233_1044", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000233", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster wind scene (hurricane-matthew), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 3, "severe_footprint_m2": 1999.9114, "severe_pct_in_quadrant": 75.0, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 4, "severe_count": 3, "severe_pct": 75.0, "severe_area_m2": 1999.9114, "total_area_m2": 2724.1641, "severity_score": 0.8325, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 2, "severe_pct": 40.0, "severe_area_m2": 1204.26, "total_area_m2": 2925.7487, "severity_score": 0.598, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 414.8079, "total_area_m2": 1525.8405, "severity_score": 0.5533, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 2, "severe_pct": 40.0, "severe_area_m2": 1204.26, "total_area_m2": 2925.7487, "severity_score": 0.598, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 4, "severe_count": 3, "severe_pct": 75.0, "severe_area_m2": 1999.9114, "total_area_m2": 2724.1641, "severity_score": 0.8325, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 414.8079, "total_area_m2": 1525.8405, "severity_score": 0.5533, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-matthew_00000233_1045", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000233", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G20", "severity_score": 1.0, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 2, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G20", "row": 2, "col": 0, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 1421.8622, "total_area_m2": 1421.8622, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G22", "row": 2, "col": 2, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 414.8079, "total_area_m2": 414.8079, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 1204.26, "total_area_m2": 1934.7218, "severity_score": 0.7767, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-matthew_00000233_1046", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000233", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster wind scene (hurricane-matthew), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 6, "severe_footprint_m2": 3618.9793, "major_damage_area_m2": 0.0, "destroyed_area_m2": 3618.9793, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-matthew_00000233_1047", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000233", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 6, "max_severe_in_cell": 3, "concentration_score": 0.5, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 2, "Q2": 3, "Q3": 1}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-matthew_00000233_1048", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000233", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-matthew). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-matthew", "disaster_type": "wind", "capture_date": "2016-10-10T16:04:56.646Z", "gsd": 1.59234964847565, "damage_summary": {"total_buildings": 12, "classified_buildings": 12, "counts": {"no-damage": 0, "minor-damage": 6, "major-damage": 0, "destroyed": 6, "unclassified": 0}, "severe_count": 6, "severe_pct": 50.0, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 7175.7533, "severe_footprint_m2": 3618.9793, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 1604, "polygon_severe_count": 6, "pixel_histogram": {"0": 1045390, "1": 0, "2": 1582, "3": 0, "4": 1604, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 49.6547, "3": 0.0, "4": 50.3453, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_palu-tsunami_00000168_1049", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000168", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a tsunami event (palu-tsunami) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 13, "minor-damage": 0, "major-damage": 0, "destroyed": 175, "unclassified": 1}, "total_buildings": 189, "classified_buildings": 188, "damaged_count": 175, "severe_count": 175, "damaged_pct": 93.09, "severe_pct": 93.09, "severe_footprint_m2": 599615.0504, "total_footprint_m2": 626686.2214, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_palu-tsunami_00000168_1050", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000168", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster tsunami scene (palu-tsunami), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 91, "severe_footprint_m2": 316522.6232, "severe_pct_in_quadrant": 100.0, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 91, "severe_count": 91, "severe_pct": 100.0, "severe_area_m2": 316522.6232, "total_area_m2": 316522.6232, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 62, "severe_count": 62, "severe_pct": 100.0, "severe_area_m2": 234519.5308, "total_area_m2": 234519.5308, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 30, "severe_count": 16, "severe_pct": 53.33, "severe_area_m2": 36369.5683, "total_area_m2": 63440.7393, "severity_score": 0.5333, "dominant_damage": "destroyed"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 30, "severe_count": 16, "severe_pct": 53.33, "severe_area_m2": 36369.5683, "total_area_m2": 63440.7393, "severity_score": 0.5333, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 12203.3281, "total_area_m2": 12203.3281, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 62, "severe_count": 62, "severe_pct": 100.0, "severe_area_m2": 234519.5308, "total_area_m2": 234519.5308, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 91, "severe_count": 91, "severe_pct": 100.0, "severe_area_m2": 316522.6232, "total_area_m2": 316522.6232, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_palu-tsunami_00000168_1051", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000168", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster tsunami satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G11", "severity_score": 1.0, "building_count": 11, "severe_count": 11, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 11, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G11", "row": 1, "col": 1, "building_count": 11, "severe_count": 11, "severe_pct": 100.0, "severe_area_m2": 29210.0489, "total_area_m2": 29210.0489, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 4073.2132, "total_area_m2": 4073.2132, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 8130.1149, "total_area_m2": 8130.1149, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_palu-tsunami_00000168_1052", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000168", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster tsunami scene (palu-tsunami), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 175, "severe_footprint_m2": 599615.0504, "major_damage_area_m2": 0.0, "destroyed_area_m2": 599615.0504, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_palu-tsunami_00000168_1053", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000168", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster tsunami scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 175, "max_severe_in_cell": 91, "concentration_score": 0.52, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 16, "Q1": 6, "Q2": 62, "Q3": 91}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_palu-tsunami_00000168_1054", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000168", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster tsunami scene (palu-tsunami). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "palu-tsunami", "disaster_type": "tsunami", "capture_date": "2018-10-01T02:26:02.000Z", "gsd": 2.8247435, "damage_summary": {"total_buildings": 189, "classified_buildings": 188, "counts": {"no-damage": 13, "minor-damage": 0, "major-damage": 0, "destroyed": 175, "unclassified": 1}, "severe_count": 175, "severe_pct": 93.09, "damaged_pct": 93.09}, "area_summary": {"total_footprint_m2": 626686.2214, "severe_footprint_m2": 599615.0504, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.52, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 81436, "polygon_severe_count": 175, "pixel_histogram": {"0": 963398, "1": 3742, "2": 0, "3": 0, "4": 81436, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 4.3932, "2": 0.0, "3": 0.0, "4": 95.6068, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00000823_1067", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000823", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 78, "minor-damage": 0, "major-damage": 0, "destroyed": 1, "unclassified": 5}, "total_buildings": 84, "classified_buildings": 79, "damaged_count": 1, "severe_count": 1, "damaged_pct": 1.27, "severe_pct": 1.27, "severe_footprint_m2": 260.3752, "total_footprint_m2": 241781.82, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00000823_1068", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000823", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 1, "severe_footprint_m2": 260.3752, "severe_pct_in_quadrant": 7.69, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 13, "severe_count": 1, "severe_pct": 7.69, "severe_area_m2": 260.3752, "total_area_m2": 46822.1645, "severity_score": 0.0769, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 52, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 147959.7925, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 15732.4039, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 52, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 147959.7925, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 13, "severe_count": 1, "severe_pct": 7.69, "severe_area_m2": 260.3752, "total_area_m2": 46822.1645, "severity_score": 0.0769, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 15732.4039, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 31267.4591, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00000823_1069", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000823", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 0.2, "building_count": 5, "severe_count": 1, "class_histogram": {"no_damage": 4, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 5, "severe_count": 1, "severe_pct": 20.0, "severe_area_m2": 260.3752, "total_area_m2": 23658.1248, "severity_score": 0.2, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 42296.5467, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 16, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 56220.263, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_portugal-wildfire_00000823_1070", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000823", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (portugal-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 1, "severe_footprint_m2": 260.3752, "major_damage_area_m2": 0.0, "destroyed_area_m2": 260.3752, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_portugal-wildfire_00000823_1071", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000823", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 1, "max_severe_in_cell": 1, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 1, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00000823_1072", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000823", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 84, "classified_buildings": 79, "counts": {"no-damage": 78, "minor-damage": 0, "major-damage": 0, "destroyed": 1, "unclassified": 5}, "severe_count": 1, "severe_pct": 1.27, "damaged_pct": 1.27}, "area_summary": {"total_footprint_m2": 241781.82, "severe_footprint_m2": 260.3752, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 1}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-matthew_00000384_1073", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000384", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a wind event (hurricane-matthew) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 0, "major-damage": 1, "destroyed": 13, "unclassified": 0}, "total_buildings": 14, "classified_buildings": 14, "damaged_count": 14, "severe_count": 14, "damaged_pct": 100.0, "severe_pct": 100.0, "severe_footprint_m2": 6765.5052, "total_footprint_m2": 6765.5052, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-matthew_00000384_1074", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000384", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster wind scene (hurricane-matthew), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 10, "severe_footprint_m2": 4923.9672, "severe_pct_in_quadrant": 100.0, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 10, "severe_count": 10, "severe_pct": 100.0, "severe_area_m2": 4923.9672, "total_area_m2": 4923.9672, "severity_score": 0.967, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 1841.538, "total_area_m2": 1841.538, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 10, "severe_count": 10, "severe_pct": 100.0, "severe_area_m2": 4923.9672, "total_area_m2": 4923.9672, "severity_score": 0.967, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 1841.538, "total_area_m2": 1841.538, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-matthew_00000384_1075", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000384", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G12", "severity_score": 1.0, "building_count": 4, "severe_count": 4, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 4, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G12", "row": 1, "col": 2, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 1839.3697, "total_area_m2": 1839.3697, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G32", "row": 3, "col": 2, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 1841.538, "total_area_m2": 1841.538, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 3084.5975, "total_area_m2": 3084.5975, "severity_score": 0.945, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-matthew_00000384_1076", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000384", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster wind scene (hurricane-matthew), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 14, "severe_footprint_m2": 6765.5052, "major_damage_area_m2": 762.7466, "destroyed_area_m2": 6002.7586, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-matthew_00000384_1077", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000384", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 14, "max_severe_in_cell": 10, "concentration_score": 0.7143, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 10, "Q2": 0, "Q3": 4}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-matthew_00000384_1078", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000384", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-matthew). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-matthew", "disaster_type": "wind", "capture_date": "2016-10-10T16:04:56.646Z", "gsd": 1.59234964847565, "damage_summary": {"total_buildings": 14, "classified_buildings": 14, "counts": {"no-damage": 0, "minor-damage": 0, "major-damage": 1, "destroyed": 13, "unclassified": 0}, "severe_count": 14, "severe_pct": 100.0, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 6765.5052, "severe_footprint_m2": 6765.5052, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.7143, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 2972, "polygon_severe_count": 14, "pixel_histogram": {"0": 1045604, "1": 0, "2": 0, "3": 320, "4": 2652, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 10.7672, "4": 89.2328, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_woolsey-fire_00000487_1085", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000487", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (woolsey-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 5, "minor-damage": 0, "major-damage": 0, "destroyed": 33, "unclassified": 0}, "total_buildings": 38, "classified_buildings": 38, "damaged_count": 33, "severe_count": 33, "damaged_pct": 86.84, "severe_pct": 86.84, "severe_footprint_m2": 53071.9586, "total_footprint_m2": 63897.5504, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_woolsey-fire_00000487_1086", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000487", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (woolsey-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 26, "severe_footprint_m2": 31812.9516, "severe_pct_in_quadrant": 96.3, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 27, "severe_count": 26, "severe_pct": 96.3, "severe_area_m2": 31812.9516, "total_area_m2": 36076.2945, "severity_score": 0.963, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 9890.7933, "total_area_m2": 9890.7933, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 5, "severe_count": 1, "severe_pct": 20.0, "severe_area_m2": 5758.7656, "total_area_m2": 12321.0145, "severity_score": 0.2, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 9890.7933, "total_area_m2": 9890.7933, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 5609.4481, "total_area_m2": 5609.4481, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 5, "severe_count": 1, "severe_pct": 20.0, "severe_area_m2": 5758.7656, "total_area_m2": 12321.0145, "severity_score": 0.2, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 27, "severe_count": 26, "severe_pct": 96.3, "severe_area_m2": 31812.9516, "total_area_m2": 36076.2945, "severity_score": 0.963, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_woolsey-fire_00000487_1087", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000487", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 9497.9982, "total_area_m2": 9497.9982, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 392.7951, "total_area_m2": 392.7951, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 5609.4481, "total_area_m2": 5609.4481, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_woolsey-fire_00000487_1088", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000487", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (woolsey-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 33, "severe_footprint_m2": 53071.9586, "major_damage_area_m2": 0.0, "destroyed_area_m2": 53071.9586, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_woolsey-fire_00000487_1089", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000487", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 33, "max_severe_in_cell": 26, "concentration_score": 0.7879, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 2, "Q1": 4, "Q2": 1, "Q3": 26}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_woolsey-fire_00000487_1090", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000487", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (woolsey-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "woolsey-fire", "disaster_type": "fire", "capture_date": "2018-11-18T18:44:07.000Z", "gsd": 1.7726423000000002, "damage_summary": {"total_buildings": 38, "classified_buildings": 38, "counts": {"no-damage": 5, "minor-damage": 0, "major-damage": 0, "destroyed": 33, "unclassified": 0}, "severe_count": 33, "severe_pct": 86.84, "damaged_pct": 86.84}, "area_summary": {"total_footprint_m2": 63897.5504, "severe_footprint_m2": 53071.9586, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.7879, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 33}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00001000_1091", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001000", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 6, "minor-damage": 0, "major-damage": 0, "destroyed": 5, "unclassified": 0}, "total_buildings": 11, "classified_buildings": 11, "damaged_count": 5, "severe_count": 5, "damaged_pct": 45.45, "severe_pct": 45.45, "severe_footprint_m2": 13014.9374, "total_footprint_m2": 41718.8114, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00001000_1092", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001000", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 2, "severe_footprint_m2": 7943.9073, "severe_pct_in_quadrant": 66.67, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 7943.9073, "total_area_m2": 15439.6137, "severity_score": 0.6667, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 5, "severe_count": 2, "severe_pct": 40.0, "severe_area_m2": 3920.342, "total_area_m2": 18875.1332, "severity_score": 0.4, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 1150.6881, "total_area_m2": 7404.0645, "severity_score": 0.3333, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 7943.9073, "total_area_m2": 15439.6137, "severity_score": 0.6667, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 1150.6881, "total_area_m2": 7404.0645, "severity_score": 0.3333, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 5, "severe_count": 2, "severe_pct": 40.0, "severe_area_m2": 3920.342, "total_area_m2": 18875.1332, "severity_score": 0.4, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00001000_1093", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001000", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 1.0, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 2, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 7943.9073, "total_area_m2": 7943.9073, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G22", "row": 2, "col": 2, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 3920.342, "total_area_m2": 3920.342, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 1150.6881, "total_area_m2": 5930.1988, "severity_score": 0.5, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_portugal-wildfire_00001000_1094", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001000", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (portugal-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 5, "severe_footprint_m2": 13014.9374, "major_damage_area_m2": 0.0, "destroyed_area_m2": 13014.9374, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_portugal-wildfire_00001000_1095", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001000", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 5, "max_severe_in_cell": 2, "concentration_score": 0.4, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 2, "Q1": 1, "Q2": 0, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00001000_1096", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001000", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 11, "classified_buildings": 11, "counts": {"no-damage": 6, "minor-damage": 0, "major-damage": 0, "destroyed": 5, "unclassified": 0}, "severe_count": 5, "severe_pct": 45.45, "damaged_pct": 45.45}, "area_summary": {"total_footprint_m2": 41718.8114, "severe_footprint_m2": 13014.9374, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.4, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 5}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000110_1097", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000110", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 4, "minor-damage": 4, "major-damage": 0, "destroyed": 1, "unclassified": 5}, "total_buildings": 14, "classified_buildings": 9, "damaged_count": 5, "severe_count": 1, "damaged_pct": 55.56, "severe_pct": 11.11, "severe_footprint_m2": 392.7013, "total_footprint_m2": 11282.4444, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000110_1098", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000110", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 1, "severe_footprint_m2": 392.7013, "severe_pct_in_quadrant": 9.09, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 11, "severe_count": 1, "severe_pct": 9.09, "severe_area_m2": 392.7013, "total_area_m2": 10270.0123, "severity_score": 0.2109, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1012.4321, "severity_score": 0.0, "dominant_damage": "unclassified"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 11, "severe_count": 1, "severe_pct": 9.09, "severe_area_m2": 392.7013, "total_area_m2": 10270.0123, "severity_score": 0.2109, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1012.4321, "severity_score": 0.0, "dominant_damage": "unclassified"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000110_1099", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000110", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 0.4433, "building_count": 3, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 1, "major_damage": 0, "destroyed": 1, "unclassified": 1}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 392.7013, "total_area_m2": 1210.1092, "severity_score": 0.4433, "dominant_damage": "minor-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 9059.9031, "severity_score": 0.1237, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 135.2356, "severity_score": 0.0, "dominant_damage": "unclassified"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000110_1100", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000110", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 1, "severe_footprint_m2": 392.7013, "major_damage_area_m2": 0.0, "destroyed_area_m2": 392.7013, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000110_1101", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000110", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 1, "max_severe_in_cell": 1, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 1, "Q1": 0, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000110_1102", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000110", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 14, "classified_buildings": 9, "counts": {"no-damage": 4, "minor-damage": 4, "major-damage": 0, "destroyed": 1, "unclassified": 5}, "severe_count": 1, "severe_pct": 11.11, "damaged_pct": 55.56}, "area_summary": {"total_footprint_m2": 11282.4444, "severe_footprint_m2": 392.7013, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 1}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_woolsey-fire_00000568_1103", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000568", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (woolsey-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 49, "minor-damage": 0, "major-damage": 0, "destroyed": 1, "unclassified": 1}, "total_buildings": 51, "classified_buildings": 50, "damaged_count": 1, "severe_count": 1, "damaged_pct": 2.0, "severe_pct": 2.0, "severe_footprint_m2": 6786.3652, "total_footprint_m2": 201732.1288, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_woolsey-fire_00000568_1104", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000568", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (woolsey-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 1, "severe_footprint_m2": 6786.3652, "severe_pct_in_quadrant": 9.09, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 11, "severe_count": 1, "severe_pct": 9.09, "severe_area_m2": 6786.3652, "total_area_m2": 48178.8222, "severity_score": 0.0909, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 53825.7668, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 13, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 70074.5556, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 15, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 53825.7668, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 11, "severe_count": 1, "severe_pct": 9.09, "severe_area_m2": 6786.3652, "total_area_m2": 48178.8222, "severity_score": 0.0909, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 13, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 70074.5556, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 29652.9842, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_woolsey-fire_00000568_1105", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000568", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G13", "severity_score": 0.25, "building_count": 4, "severe_count": 1, "class_histogram": {"no_damage": 3, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G13", "row": 1, "col": 3, "building_count": 4, "severe_count": 1, "severe_pct": 25.0, "severe_area_m2": 6786.3652, "total_area_m2": 16077.4641, "severity_score": 0.25, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 6, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 14090.5734, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 1002.7987, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_woolsey-fire_00000568_1106", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000568", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (woolsey-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 1, "severe_footprint_m2": 6786.3652, "major_damage_area_m2": 0.0, "destroyed_area_m2": 6786.3652, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_woolsey-fire_00000568_1107", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000568", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 1, "max_severe_in_cell": 1, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 1, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_woolsey-fire_00000568_1108", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000568", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (woolsey-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "woolsey-fire", "disaster_type": "fire", "capture_date": "2018-11-18T18:44:07.000Z", "gsd": 1.7726423000000002, "damage_summary": {"total_buildings": 51, "classified_buildings": 50, "counts": {"no-damage": 49, "minor-damage": 0, "major-damage": 0, "destroyed": 1, "unclassified": 1}, "severe_count": 1, "severe_pct": 2.0, "damaged_pct": 2.0}, "area_summary": {"total_footprint_m2": 201732.1288, "severe_footprint_m2": 6786.3652, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 1}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_palu-tsunami_00000040_1109", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000040", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a tsunami event (palu-tsunami) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 56, "minor-damage": 0, "major-damage": 0, "destroyed": 2, "unclassified": 0}, "total_buildings": 58, "classified_buildings": 58, "damaged_count": 2, "severe_count": 2, "damaged_pct": 3.45, "severe_pct": 3.45, "severe_footprint_m2": 2123.5703, "total_footprint_m2": 201525.0428, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_palu-tsunami_00000040_1110", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000040", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster tsunami scene (palu-tsunami), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 2, "severe_footprint_m2": 2123.5703, "severe_pct_in_quadrant": 8.7, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 23, "severe_count": 2, "severe_pct": 8.7, "severe_area_m2": 2123.5703, "total_area_m2": 58056.8416, "severity_score": 0.087, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 33, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 136231.1382, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 33, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 136231.1382, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 7237.063, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 23, "severe_count": 2, "severe_pct": 8.7, "severe_area_m2": 2123.5703, "total_area_m2": 58056.8416, "severity_score": 0.087, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_palu-tsunami_00000040_1111", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000040", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster tsunami satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G23", "severity_score": 0.1818, "building_count": 11, "severe_count": 2, "class_histogram": {"no_damage": 9, "minor_damage": 0, "major_damage": 0, "destroyed": 2, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G23", "row": 2, "col": 3, "building_count": 11, "severe_count": 2, "severe_pct": 18.18, "severe_area_m2": 2123.5703, "total_area_m2": 19859.986, "severity_score": 0.1818, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_palu-tsunami_00000040_1112", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000040", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster tsunami scene (palu-tsunami), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 2, "severe_footprint_m2": 2123.5703, "major_damage_area_m2": 0.0, "destroyed_area_m2": 2123.5703, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_palu-tsunami_00000040_1113", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000040", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster tsunami scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 2, "max_severe_in_cell": 2, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 0, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_palu-tsunami_00000040_1114", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000040", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster tsunami scene (palu-tsunami). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "palu-tsunami", "disaster_type": "tsunami", "capture_date": "2018-10-01T02:26:02.000Z", "gsd": 2.8247435, "damage_summary": {"total_buildings": 58, "classified_buildings": 58, "counts": {"no-damage": 56, "minor-damage": 0, "major-damage": 0, "destroyed": 2, "unclassified": 0}, "severe_count": 2, "severe_pct": 3.45, "damaged_pct": 3.45}, "area_summary": {"total_footprint_m2": 201525.0428, "severe_footprint_m2": 2123.5703, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 310, "polygon_severe_count": 2, "pixel_histogram": {"0": 1021321, "1": 26945, "2": 0, "3": 0, "4": 310, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 98.8626, "2": 0.0, "3": 0.0, "4": 1.1374, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_santa-rosa-wildfire_00000078_1115", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000078", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (santa-rosa-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 5, "minor-damage": 0, "major-damage": 0, "destroyed": 7, "unclassified": 0}, "total_buildings": 12, "classified_buildings": 12, "damaged_count": 7, "severe_count": 7, "damaged_pct": 58.33, "severe_pct": 58.33, "severe_footprint_m2": 25336.4472, "total_footprint_m2": 35116.5053, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_santa-rosa-wildfire_00000078_1116", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000078", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (santa-rosa-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 5, "severe_footprint_m2": 20814.0515, "severe_pct_in_quadrant": 100.0, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 5, "severe_count": 5, "severe_pct": 100.0, "severe_area_m2": 20814.0515, "total_area_m2": 20814.0515, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 7, "severe_count": 2, "severe_pct": 28.57, "severe_area_m2": 4522.3957, "total_area_m2": 14302.4538, "severity_score": 0.2857, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 5, "severe_count": 5, "severe_pct": 100.0, "severe_area_m2": 20814.0515, "total_area_m2": 20814.0515, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 7, "severe_count": 2, "severe_pct": 28.57, "severe_area_m2": 4522.3957, "total_area_m2": 14302.4538, "severity_score": 0.2857, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_santa-rosa-wildfire_00000078_1117", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000078", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 1.0, "building_count": 4, "severe_count": 4, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 4, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 18646.5486, "total_area_m2": 18646.5486, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 4522.3957, "total_area_m2": 4522.3957, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 2167.5029, "total_area_m2": 2167.5029, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_santa-rosa-wildfire_00000078_1118", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000078", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (santa-rosa-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 7, "severe_footprint_m2": 25336.4472, "major_damage_area_m2": 0.0, "destroyed_area_m2": 25336.4472, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_santa-rosa-wildfire_00000078_1119", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000078", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 7, "max_severe_in_cell": 5, "concentration_score": 0.7143, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 5, "Q1": 2, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_santa-rosa-wildfire_00000078_1120", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000078", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (santa-rosa-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "santa-rosa-wildfire", "disaster_type": "fire", "capture_date": "2017-10-11T19:19:41.000Z", "gsd": 1.8769937, "damage_summary": {"total_buildings": 12, "classified_buildings": 12, "counts": {"no-damage": 5, "minor-damage": 0, "major-damage": 0, "destroyed": 7, "unclassified": 0}, "severe_count": 7, "severe_pct": 58.33, "damaged_pct": 58.33}, "area_summary": {"total_footprint_m2": 35116.5053, "severe_footprint_m2": 25336.4472, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.7143, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 7652, "polygon_severe_count": 7, "pixel_histogram": {"0": 1038016, "1": 2908, "2": 0, "3": 0, "4": 7652, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 27.5379, "2": 0.0, "3": 0.0, "4": 72.4621, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_joplin-tornado_00000145_1121", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000145", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a wind event (joplin-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 0, "major-damage": 15, "destroyed": 163, "unclassified": 0}, "total_buildings": 178, "classified_buildings": 178, "damaged_count": 178, "severe_count": 178, "damaged_pct": 100.0, "severe_pct": 100.0, "severe_footprint_m2": 820920.0382, "total_footprint_m2": 820920.0382, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_joplin-tornado_00000145_1122", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000145", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster wind scene (joplin-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 71, "severe_footprint_m2": 238915.8436, "severe_pct_in_quadrant": 100.0, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 71, "severe_count": 71, "severe_pct": 100.0, "severe_area_m2": 238915.8436, "total_area_m2": 238915.8436, "severity_score": 0.9814, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 45, "severe_count": 45, "severe_pct": 100.0, "severe_area_m2": 233311.6671, "total_area_m2": 233311.6671, "severity_score": 0.9193, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 23, "severe_count": 23, "severe_pct": 100.0, "severe_area_m2": 217099.9699, "total_area_m2": 217099.9699, "severity_score": 1.0, "dominant_damage": "destroyed"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 71, "severe_count": 71, "severe_pct": 100.0, "severe_area_m2": 238915.8436, "total_area_m2": 238915.8436, "severity_score": 0.9814, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 45, "severe_count": 45, "severe_pct": 100.0, "severe_area_m2": 233311.6671, "total_area_m2": 233311.6671, "severity_score": 0.9193, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 39, "severe_count": 39, "severe_pct": 100.0, "severe_area_m2": 131592.5576, "total_area_m2": 131592.5576, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 23, "severe_count": 23, "severe_pct": 100.0, "severe_area_m2": 217099.9699, "total_area_m2": 217099.9699, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_joplin-tornado_00000145_1123", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000145", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 1.0, "building_count": 12, "severe_count": 12, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 12, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 12, "severe_count": 12, "severe_pct": 100.0, "severe_area_m2": 42638.8294, "total_area_m2": 42638.8294, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 13, "severe_count": 13, "severe_pct": 100.0, "severe_area_m2": 54273.123, "total_area_m2": 54273.123, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 11, "severe_count": 11, "severe_pct": 100.0, "severe_area_m2": 74437.1239, "total_area_m2": 74437.1239, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_joplin-tornado_00000145_1124", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000145", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster wind scene (joplin-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 178, "severe_footprint_m2": 820920.0382, "major_damage_area_m2": 115210.3643, "destroyed_area_m2": 705709.6739, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_joplin-tornado_00000145_1125", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000145", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 178, "max_severe_in_cell": 71, "concentration_score": 0.3989, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 71, "Q1": 45, "Q2": 39, "Q3": 23}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_joplin-tornado_00000145_1126", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000145", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (joplin-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "joplin-tornado", "disaster_type": "wind", "capture_date": "2011-05-29T17:29:25.433Z", "gsd": 2.35255861282349, "damage_summary": {"total_buildings": 178, "classified_buildings": 178, "counts": {"no-damage": 0, "minor-damage": 0, "major-damage": 15, "destroyed": 163, "unclassified": 0}, "severe_count": 178, "severe_pct": 100.0, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 820920.0382, "severe_footprint_m2": 820920.0382, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.3989, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 178}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_pinery-bushfire_00001719_1133", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001719", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (pinery-bushfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 20, "minor-damage": 0, "major-damage": 0, "destroyed": 6, "unclassified": 2}, "total_buildings": 28, "classified_buildings": 26, "damaged_count": 6, "severe_count": 6, "damaged_pct": 23.08, "severe_pct": 23.08, "severe_footprint_m2": 2090.2815, "total_footprint_m2": 19362.7098, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_pinery-bushfire_00001719_1134", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001719", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (pinery-bushfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 6, "severe_footprint_m2": 2090.2815, "severe_pct_in_quadrant": 21.43, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 28, "severe_count": 6, "severe_pct": 21.43, "severe_area_m2": 2090.2815, "total_area_m2": 19362.7098, "severity_score": 0.2143, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 28, "severe_count": 6, "severe_pct": 21.43, "severe_area_m2": 2090.2815, "total_area_m2": 19362.7098, "severity_score": 0.2143, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_pinery-bushfire_00001719_1135", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001719", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G12", "severity_score": 0.3, "building_count": 20, "severe_count": 6, "class_histogram": {"no_damage": 14, "minor_damage": 0, "major_damage": 0, "destroyed": 6, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G12", "row": 1, "col": 2, "building_count": 20, "severe_count": 6, "severe_pct": 30.0, "severe_area_m2": 2090.2815, "total_area_m2": 10822.0023, "severity_score": 0.3, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_pinery-bushfire_00001719_1136", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001719", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (pinery-bushfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 6, "severe_footprint_m2": 2090.2815, "major_damage_area_m2": 0.0, "destroyed_area_m2": 2090.2815, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_pinery-bushfire_00001719_1137", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001719", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 6, "max_severe_in_cell": 6, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 6, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_pinery-bushfire_00001719_1138", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001719", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (pinery-bushfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "pinery-bushfire", "disaster_type": "fire", "capture_date": "2015-11-27T00:39:41.429Z", "gsd": 2.03157305717468, "damage_summary": {"total_buildings": 28, "classified_buildings": 26, "counts": {"no-damage": 20, "minor-damage": 0, "major-damage": 0, "destroyed": 6, "unclassified": 2}, "severe_count": 6, "severe_pct": 23.08, "damaged_pct": 23.08}, "area_summary": {"total_footprint_m2": 19362.7098, "severe_footprint_m2": 2090.2815, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 6}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_santa-rosa-wildfire_00000277_1139", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000277", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (santa-rosa-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 1, "minor-damage": 1, "major-damage": 2, "destroyed": 10, "unclassified": 0}, "total_buildings": 14, "classified_buildings": 14, "damaged_count": 13, "severe_count": 12, "damaged_pct": 92.86, "severe_pct": 85.71, "severe_footprint_m2": 59386.307, "total_footprint_m2": 70303.8693, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_santa-rosa-wildfire_00000277_1140", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000277", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (santa-rosa-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 4, "severe_footprint_m2": 20686.6097, "severe_pct_in_quadrant": 100.0, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 20686.6097, "total_area_m2": 20686.6097, "severity_score": 0.9175, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 5, "severe_count": 5, "severe_pct": 100.0, "severe_area_m2": 20042.9331, "total_area_m2": 20042.9331, "severity_score": 0.934, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 11533.2136, "total_area_m2": 11533.2136, "severity_score": 1.0, "dominant_damage": "destroyed"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 5, "severe_count": 5, "severe_pct": 100.0, "severe_area_m2": 20042.9331, "total_area_m2": 20042.9331, "severity_score": 0.934, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 20686.6097, "total_area_m2": 20686.6097, "severity_score": 0.9175, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 11533.2136, "total_area_m2": 11533.2136, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 7123.5506, "total_area_m2": 18041.1129, "severity_score": 0.4433, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_santa-rosa-wildfire_00000277_1141", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000277", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1692.6364, "total_area_m2": 1692.6364, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 9689.4327, "total_area_m2": 9689.4327, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 9937.7451, "total_area_m2": 9937.7451, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_santa-rosa-wildfire_00000277_1142", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000277", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (santa-rosa-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 12, "severe_footprint_m2": 59386.307, "major_damage_area_m2": 3554.6346, "destroyed_area_m2": 55831.6724, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_santa-rosa-wildfire_00000277_1143", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000277", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 12, "max_severe_in_cell": 5, "concentration_score": 0.4167, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 5, "Q1": 4, "Q2": 2, "Q3": 1}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_santa-rosa-wildfire_00000277_1144", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000277", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (santa-rosa-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "santa-rosa-wildfire", "disaster_type": "fire", "capture_date": "2017-10-11T19:19:41.000Z", "gsd": 1.8769937, "damage_summary": {"total_buildings": 14, "classified_buildings": 14, "counts": {"no-damage": 1, "minor-damage": 1, "major-damage": 2, "destroyed": 10, "unclassified": 0}, "severe_count": 12, "severe_pct": 85.71, "damaged_pct": 92.86}, "area_summary": {"total_footprint_m2": 70303.8693, "severe_footprint_m2": 59386.307, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.4167, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 17832, "polygon_severe_count": 12, "pixel_histogram": {"0": 1027467, "1": 2885, "2": 392, "3": 1077, "4": 16755, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 13.6672, "2": 1.857, "3": 5.1021, "4": 79.3737, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_tuscaloosa-tornado_00000169_1145", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000169", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a wind event (tuscaloosa-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 1, "major-damage": 0, "destroyed": 55, "unclassified": 1}, "total_buildings": 57, "classified_buildings": 56, "damaged_count": 56, "severe_count": 55, "damaged_pct": 100.0, "severe_pct": 98.21, "severe_footprint_m2": 69127.1762, "total_footprint_m2": 71687.3909, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_tuscaloosa-tornado_00000169_1146", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000169", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster wind scene (tuscaloosa-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 29, "severe_footprint_m2": 36618.1131, "severe_pct_in_quadrant": 100.0, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 29, "severe_count": 29, "severe_pct": 100.0, "severe_area_m2": 36618.1131, "total_area_m2": 36618.1131, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 22, "severe_count": 22, "severe_pct": 100.0, "severe_area_m2": 30194.8639, "total_area_m2": 30194.8639, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 5, "severe_count": 3, "severe_pct": 60.0, "severe_area_m2": 1902.2212, "total_area_m2": 4462.4359, "severity_score": 0.666, "dominant_damage": "destroyed"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 22, "severe_count": 22, "severe_pct": 100.0, "severe_area_m2": 30194.8639, "total_area_m2": 30194.8639, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 29, "severe_count": 29, "severe_pct": 100.0, "severe_area_m2": 36618.1131, "total_area_m2": 36618.1131, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 411.978, "total_area_m2": 411.978, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 5, "severe_count": 3, "severe_pct": 60.0, "severe_area_m2": 1902.2212, "total_area_m2": 4462.4359, "severity_score": 0.666, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_tuscaloosa-tornado_00000169_1147", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000169", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 1.0, "building_count": 6, "severe_count": 6, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 6, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 8127.7302, "total_area_m2": 8127.7302, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 11, "severe_count": 11, "severe_pct": 100.0, "severe_area_m2": 16443.6645, "total_area_m2": 16443.6645, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 14, "severe_count": 14, "severe_pct": 100.0, "severe_area_m2": 16768.7671, "total_area_m2": 16768.7671, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_tuscaloosa-tornado_00000169_1148", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000169", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster wind scene (tuscaloosa-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 55, "severe_footprint_m2": 69127.1762, "major_damage_area_m2": 0.0, "destroyed_area_m2": 69127.1762, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_tuscaloosa-tornado_00000169_1149", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000169", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 55, "max_severe_in_cell": 29, "concentration_score": 0.5273, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 22, "Q1": 29, "Q2": 1, "Q3": 3}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_tuscaloosa-tornado_00000169_1150", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000169", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (tuscaloosa-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "tuscaloosa-tornado", "disaster_type": "wind", "capture_date": "2011-05-19T16:48:01.085Z", "gsd": 1.85690224170685, "damage_summary": {"total_buildings": 57, "classified_buildings": 56, "counts": {"no-damage": 0, "minor-damage": 1, "major-damage": 0, "destroyed": 55, "unclassified": 1}, "severe_count": 55, "severe_pct": 98.21, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 71687.3909, "severe_footprint_m2": 69127.1762, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5273, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 55}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_woolsey-fire_00000262_1151", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000262", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (woolsey-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 19, "minor-damage": 1, "major-damage": 0, "destroyed": 11, "unclassified": 0}, "total_buildings": 31, "classified_buildings": 31, "damaged_count": 12, "severe_count": 11, "damaged_pct": 38.71, "severe_pct": 35.48, "severe_footprint_m2": 13626.9703, "total_footprint_m2": 28582.1308, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_woolsey-fire_00000262_1152", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000262", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (woolsey-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 9, "severe_footprint_m2": 11891.9811, "severe_pct_in_quadrant": 33.33, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 27, "severe_count": 9, "severe_pct": 33.33, "severe_area_m2": 11891.9811, "total_area_m2": 26544.9445, "severity_score": 0.3456, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1515.2207, "total_area_m2": 1515.2207, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 219.7685, "total_area_m2": 521.9656, "severity_score": 0.3333, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 219.7685, "total_area_m2": 521.9656, "severity_score": 0.3333, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1515.2207, "total_area_m2": 1515.2207, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 27, "severe_count": 9, "severe_pct": 33.33, "severe_area_m2": 11891.9811, "total_area_m2": 26544.9445, "severity_score": 0.3456, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_woolsey-fire_00000262_1153", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000262", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G31", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G31", "row": 3, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1515.2207, "total_area_m2": 1515.2207, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G22", "row": 2, "col": 2, "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 2235.6299, "total_area_m2": 2548.7889, "severity_score": 0.6667, "dominant_damage": "destroyed"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 7, "severe_count": 4, "severe_pct": 57.14, "severe_area_m2": 8174.5807, "total_area_m2": 9703.599, "severity_score": 0.6186, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_woolsey-fire_00000262_1154", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000262", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (woolsey-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 11, "severe_footprint_m2": 13626.9703, "major_damage_area_m2": 0.0, "destroyed_area_m2": 13626.9703, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_woolsey-fire_00000262_1155", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000262", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 11, "max_severe_in_cell": 9, "concentration_score": 0.8182, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 1, "Q2": 1, "Q3": 9}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_woolsey-fire_00000262_1156", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000262", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (woolsey-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "woolsey-fire", "disaster_type": "fire", "capture_date": "2018-11-18T18:44:07.000Z", "gsd": 1.7726423000000002, "damage_summary": {"total_buildings": 31, "classified_buildings": 31, "counts": {"no-damage": 19, "minor-damage": 1, "major-damage": 0, "destroyed": 11, "unclassified": 0}, "severe_count": 11, "severe_pct": 35.48, "damaged_pct": 38.71}, "area_summary": {"total_footprint_m2": 28582.1308, "severe_footprint_m2": 13626.9703, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.8182, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 11}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000337_1157", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000337", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 17, "minor-damage": 3, "major-damage": 0, "destroyed": 2, "unclassified": 1}, "total_buildings": 23, "classified_buildings": 22, "damaged_count": 5, "severe_count": 2, "damaged_pct": 22.73, "severe_pct": 9.09, "severe_footprint_m2": 1947.5006, "total_footprint_m2": 81171.3052, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000337_1158", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000337", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 2, "severe_footprint_m2": 1947.5006, "severe_pct_in_quadrant": 33.33, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 6, "severe_count": 2, "severe_pct": 33.33, "severe_area_m2": 1947.5006, "total_area_m2": 19450.835, "severity_score": 0.3883, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 15180.7089, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 8505.1208, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 15180.7089, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 8505.1208, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 10, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 38034.6405, "severity_score": 0.066, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 6, "severe_count": 2, "severe_pct": 33.33, "severe_area_m2": 1947.5006, "total_area_m2": 19450.835, "severity_score": 0.3883, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000337_1159", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000337", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G32", "severity_score": 0.7767, "building_count": 3, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 1, "major_damage": 0, "destroyed": 2, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G32", "row": 3, "col": 2, "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 1947.5006, "total_area_m2": 6658.123, "severity_score": 0.7767, "dominant_damage": "destroyed"}, {"cell_id": "G30", "row": 3, "col": 0, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 4941.323, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "G31", "row": 3, "col": 1, "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 18093.2908, "severity_score": 0.0825, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-michael_00000337_1160", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000337", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster wind scene (hurricane-michael), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 2, "severe_footprint_m2": 1947.5006, "major_damage_area_m2": 0.0, "destroyed_area_m2": 1947.5006, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-michael_00000337_1161", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000337", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 2, "max_severe_in_cell": 2, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 0, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000337_1162", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000337", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 23, "classified_buildings": 22, "counts": {"no-damage": 17, "minor-damage": 3, "major-damage": 0, "destroyed": 2, "unclassified": 1}, "severe_count": 2, "severe_pct": 9.09, "damaged_pct": 22.73}, "area_summary": {"total_footprint_m2": 81171.3052, "severe_footprint_m2": 1947.5006, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 518, "polygon_severe_count": 2, "pixel_histogram": {"0": 1028924, "1": 14545, "2": 4589, "3": 0, "4": 518, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 74.0128, "2": 23.3513, "3": 0.0, "4": 2.6359, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_socal-fire_00000494_1163", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "socal-fire_00000494", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (socal-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 9, "minor-damage": 0, "major-damage": 1, "destroyed": 11, "unclassified": 0}, "total_buildings": 21, "classified_buildings": 21, "damaged_count": 12, "severe_count": 12, "damaged_pct": 57.14, "severe_pct": 57.14, "severe_footprint_m2": 65513.695, "total_footprint_m2": 174923.582, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_socal-fire_00000494_1164", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "socal-fire_00000494", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (socal-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 4, "severe_footprint_m2": 25364.3569, "severe_pct_in_quadrant": 50.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 8, "severe_count": 4, "severe_pct": 50.0, "severe_area_m2": 25364.3569, "total_area_m2": 29915.9549, "severity_score": 0.4587, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 8, "severe_count": 5, "severe_pct": 62.5, "severe_area_m2": 24328.9196, "total_area_m2": 122480.3771, "severity_score": 0.625, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 4, "severe_count": 3, "severe_pct": 75.0, "severe_area_m2": 15820.4185, "total_area_m2": 18187.5086, "severity_score": 0.75, "dominant_damage": "destroyed"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 3, "severe_pct": 75.0, "severe_area_m2": 15820.4185, "total_area_m2": 18187.5086, "severity_score": 0.75, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 8, "severe_count": 5, "severe_pct": 62.5, "severe_area_m2": 24328.9196, "total_area_m2": 122480.3771, "severity_score": 0.625, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 4, "severe_pct": 50.0, "severe_area_m2": 25364.3569, "total_area_m2": 29915.9549, "severity_score": 0.4587, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 4339.7414, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_socal-fire_00000494_1165", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "socal-fire_00000494", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G03", "severity_score": 1.0, "building_count": 5, "severe_count": 5, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 5, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G03", "row": 0, "col": 3, "building_count": 5, "severe_count": 5, "severe_pct": 100.0, "severe_area_m2": 24328.9196, "total_area_m2": 24328.9196, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G30", "row": 3, "col": 0, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 25364.3569, "total_area_m2": 25364.3569, "severity_score": 0.9175, "dominant_damage": "destroyed"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 4, "severe_count": 3, "severe_pct": 75.0, "severe_area_m2": 15820.4185, "total_area_m2": 18187.5086, "severity_score": 0.75, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_socal-fire_00000494_1166", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "socal-fire_00000494", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (socal-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 12, "severe_footprint_m2": 65513.695, "major_damage_area_m2": 9644.007, "destroyed_area_m2": 55869.688, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_socal-fire_00000494_1167", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "socal-fire_00000494", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 12, "max_severe_in_cell": 5, "concentration_score": 0.4167, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 3, "Q1": 5, "Q2": 4, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_socal-fire_00000494_1168", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "socal-fire_00000494", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (socal-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "socal-fire", "disaster_type": "fire", "capture_date": "2018-11-14T18:42:58.000Z", "gsd": 2.5700748, "damage_summary": {"total_buildings": 21, "classified_buildings": 21, "counts": {"no-damage": 9, "minor-damage": 0, "major-damage": 1, "destroyed": 11, "unclassified": 0}, "severe_count": 12, "severe_pct": 57.14, "damaged_pct": 57.14}, "area_summary": {"total_footprint_m2": 174923.582, "severe_footprint_m2": 65513.695, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.4167, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 10485, "polygon_severe_count": 12, "pixel_histogram": {"0": 1020941, "1": 17150, "2": 0, "3": 1534, "4": 8951, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 62.059, "2": 0.0, "3": 5.5509, "4": 32.3901, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_woolsey-fire_00000557_1169", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000557", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (woolsey-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 17, "minor-damage": 0, "major-damage": 0, "destroyed": 5, "unclassified": 0}, "total_buildings": 22, "classified_buildings": 22, "damaged_count": 5, "severe_count": 5, "damaged_pct": 22.73, "severe_pct": 22.73, "severe_footprint_m2": 20555.6206, "total_footprint_m2": 88680.3233, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_woolsey-fire_00000557_1170", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000557", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (woolsey-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 2, "severe_footprint_m2": 12534.1335, "severe_pct_in_quadrant": 22.22, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 9, "severe_count": 2, "severe_pct": 22.22, "severe_area_m2": 12534.1335, "total_area_m2": 34384.8661, "severity_score": 0.2222, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 1, "severe_pct": 12.5, "severe_area_m2": 5721.7257, "total_area_m2": 39355.7893, "severity_score": 0.125, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 4, "severe_count": 2, "severe_pct": 50.0, "severe_area_m2": 2299.7614, "total_area_m2": 10326.395, "severity_score": 0.5, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 2, "severe_pct": 50.0, "severe_area_m2": 2299.7614, "total_area_m2": 10326.395, "severity_score": 0.5, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 9, "severe_count": 2, "severe_pct": 22.22, "severe_area_m2": 12534.1335, "total_area_m2": 34384.8661, "severity_score": 0.2222, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 1, "severe_pct": 12.5, "severe_area_m2": 5721.7257, "total_area_m2": 39355.7893, "severity_score": 0.125, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 4613.2729, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_woolsey-fire_00000557_1171", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000557", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G10", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G10", "row": 1, "col": 0, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 737.6066, "total_area_m2": 737.6066, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 4, "severe_count": 2, "severe_pct": 50.0, "severe_area_m2": 12534.1335, "total_area_m2": 17826.6206, "severity_score": 0.5, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 1562.1548, "total_area_m2": 9588.7884, "severity_score": 0.3333, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_woolsey-fire_00000557_1172", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000557", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (woolsey-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 5, "severe_footprint_m2": 20555.6206, "major_damage_area_m2": 0.0, "destroyed_area_m2": 20555.6206, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_woolsey-fire_00000557_1173", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000557", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 5, "max_severe_in_cell": 2, "concentration_score": 0.4, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 2, "Q1": 2, "Q2": 1, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_woolsey-fire_00000557_1174", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000557", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (woolsey-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "woolsey-fire", "disaster_type": "fire", "capture_date": "2018-11-18T18:44:07.000Z", "gsd": 1.7726423000000002, "damage_summary": {"total_buildings": 22, "classified_buildings": 22, "counts": {"no-damage": 17, "minor-damage": 0, "major-damage": 0, "destroyed": 5, "unclassified": 0}, "severe_count": 5, "severe_pct": 22.73, "damaged_pct": 22.73}, "area_summary": {"total_footprint_m2": 88680.3233, "severe_footprint_m2": 20555.6206, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.4, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 5}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_woolsey-fire_00000722_1175", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000722", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (woolsey-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 19, "minor-damage": 1, "major-damage": 0, "destroyed": 8, "unclassified": 1}, "total_buildings": 29, "classified_buildings": 28, "damaged_count": 9, "severe_count": 8, "damaged_pct": 32.14, "severe_pct": 28.57, "severe_footprint_m2": 15295.7128, "total_footprint_m2": 36921.2648, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_woolsey-fire_00000722_1176", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000722", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (woolsey-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 4, "severe_footprint_m2": 8384.0849, "severe_pct_in_quadrant": 40.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 10, "severe_count": 4, "severe_pct": 40.0, "severe_area_m2": 8384.0849, "total_area_m2": 12352.4813, "severity_score": 0.4, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 9, "severe_count": 2, "severe_pct": 22.22, "severe_area_m2": 4691.0517, "total_area_m2": 11085.7243, "severity_score": 0.2222, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 7, "severe_count": 2, "severe_pct": 28.57, "severe_area_m2": 2220.5762, "total_area_m2": 10924.1719, "severity_score": 0.3329, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 10, "severe_count": 4, "severe_pct": 40.0, "severe_area_m2": 8384.0849, "total_area_m2": 12352.4813, "severity_score": 0.4, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 3, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 2558.8873, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 9, "severe_count": 2, "severe_pct": 22.22, "severe_area_m2": 4691.0517, "total_area_m2": 11085.7243, "severity_score": 0.2222, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 7, "severe_count": 2, "severe_pct": 28.57, "severe_area_m2": 2220.5762, "total_area_m2": 10924.1719, "severity_score": 0.3329, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_woolsey-fire_00000722_1177", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000722", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G32", "severity_score": 1.0, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 2, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G32", "row": 3, "col": 2, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 2220.5762, "total_area_m2": 2220.5762, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G21", "row": 2, "col": 1, "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 4691.0517, "total_area_m2": 6331.8292, "severity_score": 0.6667, "dominant_damage": "destroyed"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 4, "severe_count": 2, "severe_pct": 50.0, "severe_area_m2": 7332.3834, "total_area_m2": 10227.551, "severity_score": 0.5, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_woolsey-fire_00000722_1178", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000722", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (woolsey-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 8, "severe_footprint_m2": 15295.7128, "major_damage_area_m2": 0.0, "destroyed_area_m2": 15295.7128, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_woolsey-fire_00000722_1179", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000722", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 8, "max_severe_in_cell": 4, "concentration_score": 0.5, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 4, "Q1": 0, "Q2": 2, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_woolsey-fire_00000722_1180", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000722", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (woolsey-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "woolsey-fire", "disaster_type": "fire", "capture_date": "2018-11-18T18:44:07.000Z", "gsd": 1.7726423000000002, "damage_summary": {"total_buildings": 29, "classified_buildings": 28, "counts": {"no-damage": 19, "minor-damage": 1, "major-damage": 0, "destroyed": 8, "unclassified": 1}, "severe_count": 8, "severe_pct": 28.57, "damaged_pct": 32.14}, "area_summary": {"total_footprint_m2": 36921.2648, "severe_footprint_m2": 15295.7128, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 8}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_socal-fire_00000387_1187", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "socal-fire_00000387", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (socal-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 0, "major-damage": 1, "destroyed": 21, "unclassified": 0}, "total_buildings": 22, "classified_buildings": 22, "damaged_count": 22, "severe_count": 22, "damaged_pct": 100.0, "severe_pct": 100.0, "severe_footprint_m2": 94777.3648, "total_footprint_m2": 94777.3648, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_socal-fire_00000387_1188", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "socal-fire_00000387", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (socal-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 19, "severe_footprint_m2": 83149.9074, "severe_pct_in_quadrant": 100.0, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 19, "severe_count": 19, "severe_pct": 100.0, "severe_area_m2": 83149.9074, "total_area_m2": 83149.9074, "severity_score": 0.9826, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 11627.4574, "total_area_m2": 11627.4574, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 19, "severe_count": 19, "severe_pct": 100.0, "severe_area_m2": 83149.9074, "total_area_m2": 83149.9074, "severity_score": 0.9826, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 11627.4574, "total_area_m2": 11627.4574, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_socal-fire_00000387_1189", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "socal-fire_00000387", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 1.0, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 2, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 3655.3595, "total_area_m2": 3655.3595, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G03", "row": 0, "col": 3, "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 21918.403, "total_area_m2": 21918.403, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 19264.6092, "total_area_m2": 19264.6092, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_socal-fire_00000387_1190", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "socal-fire_00000387", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (socal-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 22, "severe_footprint_m2": 94777.3648, "major_damage_area_m2": 36341.9587, "destroyed_area_m2": 58435.4061, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_socal-fire_00000387_1191", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "socal-fire_00000387", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 22, "max_severe_in_cell": 19, "concentration_score": 0.8636, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 19, "Q2": 0, "Q3": 3}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_socal-fire_00000387_1192", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "socal-fire_00000387", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (socal-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "socal-fire", "disaster_type": "fire", "capture_date": "2018-11-14T18:42:58.000Z", "gsd": 2.5700748, "damage_summary": {"total_buildings": 22, "classified_buildings": 22, "counts": {"no-damage": 0, "minor-damage": 0, "major-damage": 1, "destroyed": 21, "unclassified": 0}, "severe_count": 22, "severe_pct": 100.0, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 94777.3648, "severe_footprint_m2": 94777.3648, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.8636, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 15198, "polygon_severe_count": 22, "pixel_histogram": {"0": 1033378, "1": 0, "2": 0, "3": 5662, "4": 9536, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 0.0, "3": 37.2549, "4": 62.7451, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_lower-puna-volcano_00000224_1193", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000224", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a volcano event (lower-puna-volcano) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 24, "minor-damage": 4, "major-damage": 0, "destroyed": 15, "unclassified": 0}, "total_buildings": 43, "classified_buildings": 43, "damaged_count": 19, "severe_count": 15, "damaged_pct": 44.19, "severe_pct": 34.88, "severe_footprint_m2": 31371.0606, "total_footprint_m2": 113543.3677, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_lower-puna-volcano_00000224_1194", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000224", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster volcano scene (lower-puna-volcano), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 11, "severe_footprint_m2": 23793.8737, "severe_pct_in_quadrant": 73.33, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 15, "severe_count": 11, "severe_pct": 73.33, "severe_area_m2": 23793.8737, "total_area_m2": 32889.4479, "severity_score": 0.7333, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 7, "severe_count": 4, "severe_pct": 57.14, "severe_area_m2": 7577.1869, "total_area_m2": 24516.2221, "severity_score": 0.7129, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 9036.5871, "severity_score": 0.0471, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 7, "severe_count": 4, "severe_pct": 57.14, "severe_area_m2": 7577.1869, "total_area_m2": 24516.2221, "severity_score": 0.7129, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 9036.5871, "severity_score": 0.0471, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 15, "severe_count": 11, "severe_pct": 73.33, "severe_area_m2": 23793.8737, "total_area_m2": 32889.4479, "severity_score": 0.7333, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 14, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 47101.1106, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_lower-puna-volcano_00000224_1195", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000224", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster volcano satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G20", "severity_score": 1.0, "building_count": 10, "severe_count": 10, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 10, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G20", "row": 2, "col": 0, "building_count": 10, "severe_count": 10, "severe_pct": 100.0, "severe_area_m2": 16282.8273, "total_area_m2": 16282.8273, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G21", "row": 2, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 7511.0464, "total_area_m2": 7511.0464, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 6, "severe_count": 4, "severe_pct": 66.67, "severe_area_m2": 7577.1869, "total_area_m2": 15037.9546, "severity_score": 0.7767, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_lower-puna-volcano_00000224_1196", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000224", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster volcano scene (lower-puna-volcano), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 15, "severe_footprint_m2": 31371.0606, "major_damage_area_m2": 0.0, "destroyed_area_m2": 31371.0606, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_lower-puna-volcano_00000224_1197", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000224", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster volcano scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 15, "max_severe_in_cell": 11, "concentration_score": 0.7333, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 4, "Q1": 0, "Q2": 11, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_lower-puna-volcano_00000224_1198", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000224", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster volcano scene (lower-puna-volcano). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "lower-puna-volcano", "disaster_type": "volcano", "capture_date": "2018-05-23T20:59:21.000Z", "gsd": 2.2423599, "damage_summary": {"total_buildings": 43, "classified_buildings": 43, "counts": {"no-damage": 24, "minor-damage": 4, "major-damage": 0, "destroyed": 15, "unclassified": 0}, "severe_count": 15, "severe_pct": 34.88, "damaged_pct": 44.19}, "area_summary": {"total_footprint_m2": 113543.3677, "severe_footprint_m2": 31371.0606, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.7333, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 15}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_woolsey-fire_00000126_1199", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000126", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (woolsey-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 27, "minor-damage": 3, "major-damage": 0, "destroyed": 2, "unclassified": 4}, "total_buildings": 36, "classified_buildings": 32, "damaged_count": 5, "severe_count": 2, "damaged_pct": 15.62, "severe_pct": 6.25, "severe_footprint_m2": 2106.797, "total_footprint_m2": 112428.2426, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_woolsey-fire_00000126_1200", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000126", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (woolsey-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 2, "severe_footprint_m2": 2106.797, "severe_pct_in_quadrant": 25.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 8, "severe_count": 2, "severe_pct": 25.0, "severe_area_m2": 2106.797, "total_area_m2": 18295.8988, "severity_score": 0.3325, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 16303.5858, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 31711.5576, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 7, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 16303.5858, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 31711.5576, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 46117.2004, "severity_score": 0.0275, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 8, "severe_count": 2, "severe_pct": 25.0, "severe_area_m2": 2106.797, "total_area_m2": 18295.8988, "severity_score": 0.3325, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_woolsey-fire_00000126_1201", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000126", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G32", "severity_score": 0.665, "building_count": 2, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 1, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G32", "row": 3, "col": 2, "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 1316.8709, "total_area_m2": 3735.4368, "severity_score": 0.665, "dominant_damage": "minor-damage"}, {"cell_id": "G22", "row": 2, "col": 2, "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 789.9261, "total_area_m2": 4855.8471, "severity_score": 0.4433, "dominant_damage": "no-damage"}, {"cell_id": "G31", "row": 3, "col": 1, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 5415.1695, "severity_score": 0.33, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_woolsey-fire_00000126_1202", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000126", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (woolsey-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 2, "severe_footprint_m2": 2106.797, "major_damage_area_m2": 0.0, "destroyed_area_m2": 2106.797, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_woolsey-fire_00000126_1203", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000126", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 2, "max_severe_in_cell": 2, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 0, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_woolsey-fire_00000126_1204", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000126", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (woolsey-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "woolsey-fire", "disaster_type": "fire", "capture_date": "2018-11-18T18:44:07.000Z", "gsd": 1.7726423000000002, "damage_summary": {"total_buildings": 36, "classified_buildings": 32, "counts": {"no-damage": 27, "minor-damage": 3, "major-damage": 0, "destroyed": 2, "unclassified": 4}, "severe_count": 2, "severe_pct": 6.25, "damaged_pct": 15.62}, "area_summary": {"total_footprint_m2": 112428.2426, "severe_footprint_m2": 2106.797, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 2}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-florence_00000374_1205", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000374", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a flooding event (hurricane-florence) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 109, "minor-damage": 2, "major-damage": 0, "destroyed": 1, "unclassified": 0}, "total_buildings": 112, "classified_buildings": 112, "damaged_count": 3, "severe_count": 1, "damaged_pct": 2.68, "severe_pct": 0.89, "severe_footprint_m2": 34357.6007, "total_footprint_m2": 1199945.8569, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-florence_00000374_1206", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000374", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster flooding scene (hurricane-florence), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 1, "severe_footprint_m2": 34357.6007, "severe_pct_in_quadrant": 2.38, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 42, "severe_count": 1, "severe_pct": 2.38, "severe_area_m2": 34357.6007, "total_area_m2": 380500.5422, "severity_score": 0.0317, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 18, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 181706.9858, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 20, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 233981.5156, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 18, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 181706.9858, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 20, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 233981.5156, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 32, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 403756.8133, "severity_score": 0.0103, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 42, "severe_count": 1, "severe_pct": 2.38, "severe_area_m2": 34357.6007, "total_area_m2": 380500.5422, "severity_score": 0.0317, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-florence_00000374_1207", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000374", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G22", "severity_score": 0.125, "building_count": 8, "severe_count": 1, "class_histogram": {"no_damage": 7, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G22", "row": 2, "col": 2, "building_count": 8, "severe_count": 1, "severe_pct": 12.5, "severe_area_m2": 34357.6007, "total_area_m2": 130638.4302, "severity_score": 0.125, "dominant_damage": "no-damage"}, {"cell_id": "G32", "row": 3, "col": 2, "building_count": 12, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 149209.1257, "severity_score": 0.0275, "dominant_damage": "no-damage"}, {"cell_id": "G31", "row": 3, "col": 1, "building_count": 13, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 102845.0474, "severity_score": 0.0254, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-florence_00000374_1208", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000374", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster flooding scene (hurricane-florence), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 1, "severe_footprint_m2": 34357.6007, "major_damage_area_m2": 0.0, "destroyed_area_m2": 34357.6007, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-florence_00000374_1209", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000374", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 1, "max_severe_in_cell": 1, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 0, "Q3": 1}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-florence_00000374_1210", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000374", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-florence). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-florence", "disaster_type": "flooding", "capture_date": "2018-09-20T16:04:41.000Z", "gsd": 2.0916247, "damage_summary": {"total_buildings": 112, "classified_buildings": 112, "counts": {"no-damage": 109, "minor-damage": 2, "major-damage": 0, "destroyed": 1, "unclassified": 0}, "severe_count": 1, "severe_pct": 0.89, "damaged_pct": 2.68}, "area_summary": {"total_footprint_m2": 1199945.8569, "severe_footprint_m2": 34357.6007, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 8025, "polygon_severe_count": 1, "pixel_histogram": {"0": 764969, "1": 266104, "2": 9478, "3": 0, "4": 8025, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 93.8284, "2": 3.3419, "3": 0.0, "4": 2.8296, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00001682_1211", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001682", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 65, "minor-damage": 0, "major-damage": 3, "destroyed": 21, "unclassified": 1}, "total_buildings": 90, "classified_buildings": 89, "damaged_count": 24, "severe_count": 24, "damaged_pct": 26.97, "severe_pct": 26.97, "severe_footprint_m2": 59843.6872, "total_footprint_m2": 254374.0879, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00001682_1212", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001682", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 11, "severe_footprint_m2": 31825.3177, "severe_pct_in_quadrant": 20.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 55, "severe_count": 11, "severe_pct": 20.0, "severe_area_m2": 31825.3177, "total_area_m2": 173268.2031, "severity_score": 0.188, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 34, "severe_count": 12, "severe_pct": 35.29, "severe_area_m2": 26937.8469, "total_area_m2": 80025.3622, "severity_score": 0.3432, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1080.5226, "total_area_m2": 1080.5226, "severity_score": 1.0, "dominant_damage": "destroyed"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 55, "severe_count": 11, "severe_pct": 20.0, "severe_area_m2": 31825.3177, "total_area_m2": 173268.2031, "severity_score": 0.188, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1080.5226, "total_area_m2": 1080.5226, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 34, "severe_count": 12, "severe_pct": 35.29, "severe_area_m2": 26937.8469, "total_area_m2": 80025.3622, "severity_score": 0.3432, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00001682_1213", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001682", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G12", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G12", "row": 1, "col": 2, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1080.5226, "total_area_m2": 1080.5226, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G31", "row": 3, "col": 1, "building_count": 5, "severe_count": 4, "severe_pct": 80.0, "severe_area_m2": 7209.2281, "total_area_m2": 8069.8965, "severity_score": 0.8, "dominant_damage": "destroyed"}, {"cell_id": "G21", "row": 2, "col": 1, "building_count": 10, "severe_count": 4, "severe_pct": 40.0, "severe_area_m2": 8198.7186, "total_area_m2": 25085.0316, "severity_score": 0.4, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_portugal-wildfire_00001682_1214", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001682", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (portugal-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 24, "severe_footprint_m2": 59843.6872, "major_damage_area_m2": 12596.4978, "destroyed_area_m2": 47247.1894, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_portugal-wildfire_00001682_1215", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001682", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 24, "max_severe_in_cell": 12, "concentration_score": 0.5, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 11, "Q1": 1, "Q2": 12, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00001682_1216", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001682", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 90, "classified_buildings": 89, "counts": {"no-damage": 65, "minor-damage": 0, "major-damage": 3, "destroyed": 21, "unclassified": 1}, "severe_count": 24, "severe_pct": 26.97, "damaged_pct": 26.97}, "area_summary": {"total_footprint_m2": 254374.0879, "severe_footprint_m2": 59843.6872, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 24}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_santa-rosa-wildfire_00000170_1217", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000170", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a fire event (santa-rosa-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 2, "minor-damage": 8, "major-damage": 1, "destroyed": 17, "unclassified": 0}, "total_buildings": 28, "classified_buildings": 28, "damaged_count": 26, "severe_count": 18, "damaged_pct": 92.86, "severe_pct": 64.29, "severe_footprint_m2": 63339.5145, "total_footprint_m2": 89096.4747, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_santa-rosa-wildfire_00000170_1218", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000170", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster fire scene (santa-rosa-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 5, "severe_footprint_m2": 22362.9511, "severe_pct_in_quadrant": 62.5, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 8, "severe_count": 5, "severe_pct": 62.5, "severe_area_m2": 22362.9511, "total_area_m2": 27097.4502, "severity_score": 0.6663, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 9, "severe_count": 6, "severe_pct": 66.67, "severe_area_m2": 16336.6579, "total_area_m2": 24318.4837, "severity_score": 0.7767, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 14561.5814, "total_area_m2": 14561.5814, "severity_score": 0.9175, "dominant_damage": "destroyed"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 9, "severe_count": 6, "severe_pct": 66.67, "severe_area_m2": 16336.6579, "total_area_m2": 24318.4837, "severity_score": 0.7767, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 7, "severe_count": 3, "severe_pct": 42.86, "severe_area_m2": 10078.3241, "total_area_m2": 23118.9594, "severity_score": 0.6171, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 5, "severe_pct": 62.5, "severe_area_m2": 22362.9511, "total_area_m2": 27097.4502, "severity_score": 0.6663, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 14561.5814, "total_area_m2": 14561.5814, "severity_score": 0.9175, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_santa-rosa-wildfire_00000170_1219", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000170", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 1.0, "building_count": 5, "severe_count": 5, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 5, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 5, "severe_count": 5, "severe_pct": 100.0, "severe_area_m2": 9786.1985, "total_area_m2": 9786.1985, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1006.3544, "total_area_m2": 1006.3544, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 6550.4594, "total_area_m2": 6550.4594, "severity_score": 1.0, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_santa-rosa-wildfire_00000170_1220", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000170", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster fire scene (santa-rosa-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 18, "severe_footprint_m2": 63339.5145, "major_damage_area_m2": 436.2956, "destroyed_area_m2": 62903.2189, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_santa-rosa-wildfire_00000170_1221", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000170", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 18, "max_severe_in_cell": 6, "concentration_score": 0.3333, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 6, "Q1": 3, "Q2": 5, "Q3": 4}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_santa-rosa-wildfire_00000170_1222", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000170", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (santa-rosa-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "santa-rosa-wildfire", "disaster_type": "fire", "capture_date": "2017-10-11T19:19:41.000Z", "gsd": 1.8769937, "damage_summary": {"total_buildings": 28, "classified_buildings": 28, "counts": {"no-damage": 2, "minor-damage": 8, "major-damage": 1, "destroyed": 17, "unclassified": 0}, "severe_count": 18, "severe_pct": 64.29, "damaged_pct": 92.86}, "area_summary": {"total_footprint_m2": 89096.4747, "severe_footprint_m2": 63339.5145, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.3333, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 19126, "polygon_severe_count": 18, "pixel_histogram": {"0": 1021626, "1": 967, "2": 6857, "3": 157, "4": 18969, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 3.5881, "2": 25.4434, "3": 0.5826, "4": 70.3859, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-harvey_00000244_1223", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000244", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "A post-disaster satellite scene from a flooding event (hurricane-harvey) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 5, "minor-damage": 2, "major-damage": 0, "destroyed": 7, "unclassified": 0}, "total_buildings": 14, "classified_buildings": 14, "damaged_count": 9, "severe_count": 7, "damaged_pct": 64.29, "severe_pct": 50.0, "severe_footprint_m2": 4724.2055, "total_footprint_m2": 68796.5251, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-harvey_00000244_1224", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000244", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "For a post-disaster flooding scene (hurricane-harvey), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 4, "severe_footprint_m2": 2892.2193, "severe_pct_in_quadrant": 66.67, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 6, "severe_count": 4, "severe_pct": 66.67, "severe_area_m2": 2892.2193, "total_area_m2": 54912.5207, "severity_score": 0.7767, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 8, "severe_count": 3, "severe_pct": 37.5, "severe_area_m2": 1831.9862, "total_area_m2": 13884.0044, "severity_score": 0.375, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 8, "severe_count": 3, "severe_pct": 37.5, "severe_area_m2": 1831.9862, "total_area_m2": 13884.0044, "severity_score": 0.375, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 6, "severe_count": 4, "severe_pct": 66.67, "severe_area_m2": 2892.2193, "total_area_m2": 54912.5207, "severity_score": 0.7767, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-harvey_00000244_1225", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000244", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G22", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G22", "row": 2, "col": 2, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 964.8629, "total_area_m2": 964.8629, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 5, "severe_count": 3, "severe_pct": 60.0, "severe_area_m2": 1927.3564, "total_area_m2": 53947.6578, "severity_score": 0.732, "dominant_damage": "destroyed"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 8, "severe_count": 3, "severe_pct": 37.5, "severe_area_m2": 1831.9862, "total_area_m2": 13884.0044, "severity_score": 0.375, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-harvey_00000244_1226", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000244", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "From the post-disaster flooding scene (hurricane-harvey), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 7, "severe_footprint_m2": 4724.2055, "major_damage_area_m2": 0.0, "destroyed_area_m2": 4724.2055, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-harvey_00000244_1227", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000244", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 7, "max_severe_in_cell": 4, "concentration_score": 0.5714, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 3, "Q2": 0, "Q3": 4}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-harvey_00000244_1228", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000244", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "destroyed-heavy", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-harvey). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-harvey", "disaster_type": "flooding", "capture_date": "2017-08-31T17:38:50.685Z", "gsd": 3.024613, "damage_summary": {"total_buildings": 14, "classified_buildings": 14, "counts": {"no-damage": 5, "minor-damage": 2, "major-damage": 0, "destroyed": 7, "unclassified": 0}, "severe_count": 7, "severe_pct": 50.0, "damaged_pct": 64.29}, "area_summary": {"total_footprint_m2": 68796.5251, "severe_footprint_m2": 4724.2055, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5714, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 657, "polygon_severe_count": 7, "pixel_histogram": {"0": 1040577, "1": 1478, "2": 5864, "3": 0, "4": 657, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 18.4773, "2": 73.3092, "3": 0.0, "4": 8.2135, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00000232_1241", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000232", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 47, "minor-damage": 1, "major-damage": 1, "destroyed": 2, "unclassified": 1}, "total_buildings": 52, "classified_buildings": 51, "damaged_count": 4, "severe_count": 3, "damaged_pct": 7.84, "severe_pct": 5.88, "severe_footprint_m2": 7318.8021, "total_footprint_m2": 153598.9737, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00000232_1242", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000232", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 1, "severe_footprint_m2": 2961.9622, "severe_pct_in_quadrant": 14.29, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 7, "severe_count": 1, "severe_pct": 14.29, "severe_area_m2": 2961.9622, "total_area_m2": 12481.0547, "severity_score": 0.0957, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 1, "severe_pct": 20.0, "severe_area_m2": 2743.1613, "total_area_m2": 15673.3245, "severity_score": 0.2, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 26, "severe_count": 1, "severe_pct": 3.85, "severe_area_m2": 1613.6786, "total_area_m2": 69374.2782, "severity_score": 0.0385, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 26, "severe_count": 1, "severe_pct": 3.85, "severe_area_m2": 1613.6786, "total_area_m2": 69374.2782, "severity_score": 0.0385, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 5, "severe_count": 1, "severe_pct": 20.0, "severe_area_m2": 2743.1613, "total_area_m2": 15673.3245, "severity_score": 0.2, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 14, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 56070.3163, "severity_score": 0.0236, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 7, "severe_count": 1, "severe_pct": 14.29, "severe_area_m2": 2961.9622, "total_area_m2": 12481.0547, "severity_score": 0.0957, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00000232_1243", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000232", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 0.25, "building_count": 4, "severe_count": 1, "class_histogram": {"no_damage": 3, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 4, "severe_count": 1, "severe_pct": 25.0, "severe_area_m2": 2743.1613, "total_area_m2": 14749.1326, "severity_score": 0.25, "dominant_damage": "no-damage"}, {"cell_id": "G22", "row": 2, "col": 2, "building_count": 7, "severe_count": 1, "severe_pct": 14.29, "severe_area_m2": 2961.9622, "total_area_m2": 12481.0547, "severity_score": 0.0957, "dominant_damage": "no-damage"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 17, "severe_count": 1, "severe_pct": 5.88, "severe_area_m2": 1613.6786, "total_area_m2": 44510.2259, "severity_score": 0.0588, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_portugal-wildfire_00000232_1244", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000232", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "From the post-disaster fire scene (portugal-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 3, "severe_footprint_m2": 7318.8021, "major_damage_area_m2": 2961.9622, "destroyed_area_m2": 4356.8399, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_portugal-wildfire_00000232_1245", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000232", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 3, "max_severe_in_cell": 1, "concentration_score": 0.3333, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 1, "Q1": 1, "Q2": 0, "Q3": 1}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00000232_1246", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000232", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 52, "classified_buildings": 51, "counts": {"no-damage": 47, "minor-damage": 1, "major-damage": 1, "destroyed": 2, "unclassified": 1}, "severe_count": 3, "severe_pct": 5.88, "damaged_pct": 7.84}, "area_summary": {"total_footprint_m2": 153598.9737, "severe_footprint_m2": 7318.8021, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.3333, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 3}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_tuscaloosa-tornado_00000112_1247", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000112", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (tuscaloosa-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 32, "minor-damage": 31, "major-damage": 21, "destroyed": 23, "unclassified": 5}, "total_buildings": 112, "classified_buildings": 107, "damaged_count": 75, "severe_count": 44, "damaged_pct": 70.09, "severe_pct": 41.12, "severe_footprint_m2": 198723.5842, "total_footprint_m2": 451946.7938, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_tuscaloosa-tornado_00000112_1248", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000112", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (tuscaloosa-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 21, "severe_footprint_m2": 102952.7581, "severe_pct_in_quadrant": 72.41, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 29, "severe_count": 21, "severe_pct": 72.41, "severe_area_m2": 102952.7581, "total_area_m2": 136269.0204, "severity_score": 0.69, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 30, "severe_count": 19, "severe_pct": 63.33, "severe_area_m2": 65892.9933, "total_area_m2": 103049.9987, "severity_score": 0.6223, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 18, "severe_count": 3, "severe_pct": 16.67, "severe_area_m2": 29031.0526, "total_area_m2": 96440.7544, "severity_score": 0.2583, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 29, "severe_count": 21, "severe_pct": 72.41, "severe_area_m2": 102952.7581, "total_area_m2": 136269.0204, "severity_score": 0.69, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 30, "severe_count": 19, "severe_pct": 63.33, "severe_area_m2": 65892.9933, "total_area_m2": 103049.9987, "severity_score": 0.6223, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 18, "severe_count": 3, "severe_pct": 16.67, "severe_area_m2": 29031.0526, "total_area_m2": 96440.7544, "severity_score": 0.2583, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 35, "severe_count": 1, "severe_pct": 2.86, "severe_area_m2": 846.7802, "total_area_m2": 116187.0203, "severity_score": 0.1134, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_tuscaloosa-tornado_00000112_1249", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000112", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 0.88, "building_count": 11, "severe_count": 11, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 4, "destroyed": 7, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 11, "severe_count": 11, "severe_pct": 100.0, "severe_area_m2": 39720.7239, "total_area_m2": 39720.7239, "severity_score": 0.88, "dominant_damage": "destroyed"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 8, "severe_count": 7, "severe_pct": 87.5, "severe_area_m2": 28901.6811, "total_area_m2": 33021.6532, "severity_score": 0.7925, "dominant_damage": "destroyed"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 6, "severe_count": 5, "severe_pct": 83.33, "severe_area_m2": 34332.7652, "total_area_m2": 35153.0612, "severity_score": 0.7783, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_tuscaloosa-tornado_00000112_1250", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000112", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (tuscaloosa-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 44, "severe_footprint_m2": 198723.5842, "major_damage_area_m2": 128112.9022, "destroyed_area_m2": 70610.682, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_tuscaloosa-tornado_00000112_1251", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000112", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 44, "max_severe_in_cell": 21, "concentration_score": 0.4773, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 21, "Q1": 19, "Q2": 3, "Q3": 1}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_tuscaloosa-tornado_00000112_1252", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000112", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (tuscaloosa-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "tuscaloosa-tornado", "disaster_type": "wind", "capture_date": "2011-05-19T16:48:01.085Z", "gsd": 1.85690224170685, "damage_summary": {"total_buildings": 112, "classified_buildings": 107, "counts": {"no-damage": 32, "minor-damage": 31, "major-damage": 21, "destroyed": 23, "unclassified": 5}, "severe_count": 44, "severe_pct": 41.12, "damaged_pct": 70.09}, "area_summary": {"total_footprint_m2": 451946.7938, "severe_footprint_m2": 198723.5842, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.4773, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 44}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_moore-tornado_00000140_1259", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "moore-tornado_00000140", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (moore-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 99, "minor-damage": 55, "major-damage": 24, "destroyed": 30, "unclassified": 4}, "total_buildings": 212, "classified_buildings": 208, "damaged_count": 109, "severe_count": 54, "damaged_pct": 52.4, "severe_pct": 25.96, "severe_footprint_m2": 169207.4905, "total_footprint_m2": 649919.9369, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_moore-tornado_00000140_1260", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "moore-tornado_00000140", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (moore-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 33, "severe_footprint_m2": 115491.4004, "severe_pct_in_quadrant": 62.26, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 53, "severe_count": 33, "severe_pct": 62.26, "severe_area_m2": 115491.4004, "total_area_m2": 184896.9745, "severity_score": 0.6289, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 54, "severe_count": 19, "severe_pct": 35.19, "severe_area_m2": 53372.6844, "total_area_m2": 152166.7667, "severity_score": 0.413, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 57, "severe_count": 2, "severe_pct": 3.51, "severe_area_m2": 343.4057, "total_area_m2": 161883.3531, "severity_score": 0.093, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 53, "severe_count": 33, "severe_pct": 62.26, "severe_area_m2": 115491.4004, "total_area_m2": 184896.9745, "severity_score": 0.6289, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 54, "severe_count": 19, "severe_pct": 35.19, "severe_area_m2": 53372.6844, "total_area_m2": 152166.7667, "severity_score": 0.413, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 48, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 150972.8426, "severity_score": 0.0688, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 57, "severe_count": 2, "severe_pct": 3.51, "severe_area_m2": 343.4057, "total_area_m2": 161883.3531, "severity_score": 0.093, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_moore-tornado_00000140_1261", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "moore-tornado_00000140", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 1.0, "building_count": 7, "severe_count": 7, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 7, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 7, "severe_count": 7, "severe_pct": 100.0, "severe_area_m2": 17138.0916, "total_area_m2": 17138.0916, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 7, "severe_count": 6, "severe_pct": 85.71, "severe_area_m2": 25845.8762, "total_area_m2": 30251.5327, "severity_score": 0.8571, "dominant_damage": "destroyed"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 9, "severe_count": 8, "severe_pct": 88.89, "severe_area_m2": 20054.0275, "total_area_m2": 23484.3514, "severity_score": 0.8522, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_moore-tornado_00000140_1262", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "moore-tornado_00000140", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (moore-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 54, "severe_footprint_m2": 169207.4905, "major_damage_area_m2": 93797.074, "destroyed_area_m2": 75410.4165, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_moore-tornado_00000140_1263", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "moore-tornado_00000140", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 54, "max_severe_in_cell": 33, "concentration_score": 0.6111, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 33, "Q1": 19, "Q2": 0, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_moore-tornado_00000140_1264", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "moore-tornado_00000140", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (moore-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "moore-tornado", "disaster_type": "wind", "capture_date": "2013-05-22T17:26:30.085Z", "gsd": 1.72984850406647, "damage_summary": {"total_buildings": 212, "classified_buildings": 208, "counts": {"no-damage": 99, "minor-damage": 55, "major-damage": 24, "destroyed": 30, "unclassified": 4}, "severe_count": 54, "severe_pct": 25.96, "damaged_pct": 52.4}, "area_summary": {"total_footprint_m2": 649919.9369, "severe_footprint_m2": 169207.4905, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6111, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 54}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-matthew_00000063_1265", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000063", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (hurricane-matthew) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 9, "major-damage": 2, "destroyed": 3, "unclassified": 0}, "total_buildings": 14, "classified_buildings": 14, "damaged_count": 14, "severe_count": 5, "damaged_pct": 100.0, "severe_pct": 35.71, "severe_footprint_m2": 17692.1073, "total_footprint_m2": 59762.7155, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-matthew_00000063_1266", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000063", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (hurricane-matthew), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 5, "severe_footprint_m2": 17692.1073, "severe_pct_in_quadrant": 41.67, "dominant_damage": "minor-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 12, "severe_count": 5, "severe_pct": 41.67, "severe_area_m2": 17692.1073, "total_area_m2": 45881.8913, "severity_score": 0.5542, "dominant_damage": "minor-damage"}, {"cell_id": "Q0", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 13880.8242, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 13880.8242, "severity_score": 0.33, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 12, "severe_count": 5, "severe_pct": 41.67, "severe_area_m2": 17692.1073, "total_area_m2": 45881.8913, "severity_score": 0.5542, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-matthew_00000063_1267", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000063", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G13", "severity_score": 0.665, "building_count": 4, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 2, "major_damage": 0, "destroyed": 2, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G13", "row": 1, "col": 3, "building_count": 4, "severe_count": 2, "severe_pct": 50.0, "severe_area_m2": 3579.3264, "total_area_m2": 9359.9581, "severity_score": 0.665, "dominant_damage": "minor-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 5, "severe_count": 2, "severe_pct": 40.0, "severe_area_m2": 11794.7157, "total_area_m2": 28215.8726, "severity_score": 0.532, "dominant_damage": "minor-damage"}, {"cell_id": "G03", "row": 0, "col": 3, "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 2318.0652, "total_area_m2": 8306.0606, "severity_score": 0.4433, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-matthew_00000063_1268", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000063", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (hurricane-matthew), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 5, "severe_footprint_m2": 17692.1073, "major_damage_area_m2": 6486.2416, "destroyed_area_m2": 11205.8657, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-matthew_00000063_1269", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000063", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 5, "max_severe_in_cell": 5, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 5, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-matthew_00000063_1270", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000063", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-matthew). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-matthew", "disaster_type": "wind", "capture_date": "2016-10-09T15:32:03.000Z", "gsd": 2.77336645, "damage_summary": {"total_buildings": 14, "classified_buildings": 14, "counts": {"no-damage": 0, "minor-damage": 9, "major-damage": 2, "destroyed": 3, "unclassified": 0}, "severe_count": 5, "severe_pct": 35.71, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 59762.7155, "severe_footprint_m2": 17692.1073, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 2485, "polygon_severe_count": 5, "pixel_histogram": {"0": 1040230, "1": 0, "2": 5861, "3": 890, "4": 1595, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 70.2253, "3": 10.6638, "4": 19.111, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000162_1277", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000162", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 47, "minor-damage": 18, "major-damage": 21, "destroyed": 11, "unclassified": 0}, "total_buildings": 97, "classified_buildings": 97, "damaged_count": 50, "severe_count": 32, "damaged_pct": 51.55, "severe_pct": 32.99, "severe_footprint_m2": 160816.4999, "total_footprint_m2": 369687.4153, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000162_1278", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000162", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 10, "severe_footprint_m2": 92049.3522, "severe_pct_in_quadrant": 47.62, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 21, "severe_count": 10, "severe_pct": 47.62, "severe_area_m2": 92049.3522, "total_area_m2": 123089.0598, "severity_score": 0.4919, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 33, "severe_count": 9, "severe_pct": 27.27, "severe_area_m2": 31060.8114, "total_area_m2": 140932.3976, "severity_score": 0.2227, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 20, "severe_count": 4, "severe_pct": 20.0, "severe_area_m2": 19785.8251, "total_area_m2": 53393.5663, "severity_score": 0.2, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 21, "severe_count": 10, "severe_pct": 47.62, "severe_area_m2": 92049.3522, "total_area_m2": 123089.0598, "severity_score": 0.4919, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 33, "severe_count": 9, "severe_pct": 27.27, "severe_area_m2": 31060.8114, "total_area_m2": 140932.3976, "severity_score": 0.2227, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 23, "severe_count": 9, "severe_pct": 39.13, "severe_area_m2": 17920.5112, "total_area_m2": 52272.3916, "severity_score": 0.4057, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 20, "severe_count": 4, "severe_pct": 20.0, "severe_area_m2": 19785.8251, "total_area_m2": 53393.5663, "severity_score": 0.2, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000162_1279", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000162", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G30", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G30", "row": 3, "col": 0, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1977.5289, "total_area_m2": 1977.5289, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 82813.7844, "total_area_m2": 82813.7844, "severity_score": 0.835, "dominant_damage": "major-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 6, "severe_count": 3, "severe_pct": 50.0, "severe_area_m2": 5705.0373, "total_area_m2": 14467.9632, "severity_score": 0.5, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-michael_00000162_1280", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000162", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (hurricane-michael), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 32, "severe_footprint_m2": 160816.4999, "major_damage_area_m2": 144393.2519, "destroyed_area_m2": 16423.248, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-michael_00000162_1281", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000162", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 32, "max_severe_in_cell": 10, "concentration_score": 0.3125, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 10, "Q1": 9, "Q2": 9, "Q3": 4}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000162_1282", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000162", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 97, "classified_buildings": 97, "counts": {"no-damage": 47, "minor-damage": 18, "major-damage": 21, "destroyed": 11, "unclassified": 0}, "severe_count": 32, "severe_pct": 32.99, "damaged_pct": 51.55}, "area_summary": {"total_footprint_m2": 369687.4153, "severe_footprint_m2": 160816.4999, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.3125, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 39572, "polygon_severe_count": 32, "pixel_histogram": {"0": 956849, "1": 37087, "2": 15068, "3": 35319, "4": 4253, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 40.4319, "2": 16.427, "3": 38.5045, "4": 4.6366, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-matthew_00000097_1283", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000097", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (hurricane-matthew) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 83, "major-damage": 8, "destroyed": 7, "unclassified": 1}, "total_buildings": 99, "classified_buildings": 98, "damaged_count": 98, "severe_count": 15, "damaged_pct": 100.0, "severe_pct": 15.31, "severe_footprint_m2": 18939.8828, "total_footprint_m2": 150475.173, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-matthew_00000097_1284", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000097", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (hurricane-matthew), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 2, "severe_footprint_m2": 9381.9521, "severe_pct_in_quadrant": 20.0, "dominant_damage": "minor-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 10, "severe_count": 2, "severe_pct": 20.0, "severe_area_m2": 9381.9521, "total_area_m2": 37088.1379, "severity_score": 0.398, "dominant_damage": "minor-damage"}, {"cell_id": "Q0", "building_count": 83, "severe_count": 11, "severe_pct": 13.25, "severe_area_m2": 8342.2532, "total_area_m2": 105124.7544, "severity_score": 0.3989, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 6, "severe_count": 2, "severe_pct": 33.33, "severe_area_m2": 1215.6775, "total_area_m2": 8262.2807, "severity_score": 0.4433, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 83, "severe_count": 11, "severe_pct": 13.25, "severe_area_m2": 8342.2532, "total_area_m2": 105124.7544, "severity_score": 0.3989, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 6, "severe_count": 2, "severe_pct": 33.33, "severe_area_m2": 1215.6775, "total_area_m2": 8262.2807, "severity_score": 0.4433, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 10, "severe_count": 2, "severe_pct": 20.0, "severe_area_m2": 9381.9521, "total_area_m2": 37088.1379, "severity_score": 0.398, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-matthew_00000097_1285", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000097", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G10", "severity_score": 0.4482, "building_count": 17, "severe_count": 3, "class_histogram": {"no_damage": 0, "minor_damage": 14, "major_damage": 0, "destroyed": 3, "unclassified": 0}, "dominant_damage": "minor-damage", "top_k_cells": [{"cell_id": "G10", "row": 1, "col": 0, "building_count": 17, "severe_count": 3, "severe_pct": 17.65, "severe_area_m2": 1613.6776, "total_area_m2": 24848.7229, "severity_score": 0.4482, "dominant_damage": "minor-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 6, "severe_count": 2, "severe_pct": 33.33, "severe_area_m2": 1215.6775, "total_area_m2": 8262.2807, "severity_score": 0.4433, "dominant_damage": "minor-damage"}, {"cell_id": "G20", "row": 2, "col": 0, "building_count": 8, "severe_count": 2, "severe_pct": 25.0, "severe_area_m2": 9381.9521, "total_area_m2": 35922.7898, "severity_score": 0.415, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-matthew_00000097_1286", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000097", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (hurricane-matthew), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 15, "severe_footprint_m2": 18939.8828, "major_damage_area_m2": 14058.3067, "destroyed_area_m2": 4881.5761, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-matthew_00000097_1287", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000097", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 15, "max_severe_in_cell": 11, "concentration_score": 0.7333, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 11, "Q1": 2, "Q2": 2, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-matthew_00000097_1288", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000097", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-matthew). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-matthew", "disaster_type": "wind", "capture_date": "2016-10-09T15:32:03.000Z", "gsd": 2.77336645, "damage_summary": {"total_buildings": 99, "classified_buildings": 98, "counts": {"no-damage": 0, "minor-damage": 83, "major-damage": 8, "destroyed": 7, "unclassified": 1}, "severe_count": 15, "severe_pct": 15.31, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 150475.173, "severe_footprint_m2": 18939.8828, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.7333, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 2813, "polygon_severe_count": 15, "pixel_histogram": {"0": 1026563, "1": 0, "2": 19200, "3": 2060, "4": 753, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 87.2212, "3": 9.3581, "4": 3.4207, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_tuscaloosa-tornado_00000266_1295", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000266", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (tuscaloosa-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 22, "minor-damage": 12, "major-damage": 3, "destroyed": 3, "unclassified": 15}, "total_buildings": 55, "classified_buildings": 40, "damaged_count": 18, "severe_count": 6, "damaged_pct": 45.0, "severe_pct": 15.0, "severe_footprint_m2": 12779.672, "total_footprint_m2": 79636.9439, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_tuscaloosa-tornado_00000266_1296", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000266", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (tuscaloosa-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 4, "severe_footprint_m2": 6095.9373, "severe_pct_in_quadrant": 16.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 25, "severe_count": 4, "severe_pct": 16.0, "severe_area_m2": 6095.9373, "total_area_m2": 32952.7501, "severity_score": 0.1732, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 4, "severe_count": 1, "severe_pct": 25.0, "severe_area_m2": 5102.0831, "total_area_m2": 9203.581, "severity_score": 0.3325, "dominant_damage": "minor-damage"}, {"cell_id": "Q0", "building_count": 17, "severe_count": 1, "severe_pct": 5.88, "severe_area_m2": 1581.6516, "total_area_m2": 24302.8372, "severity_score": 0.1753, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 17, "severe_count": 1, "severe_pct": 5.88, "severe_area_m2": 1581.6516, "total_area_m2": 24302.8372, "severity_score": 0.1753, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 25, "severe_count": 4, "severe_pct": 16.0, "severe_area_m2": 6095.9373, "total_area_m2": 32952.7501, "severity_score": 0.1732, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 4, "severe_count": 1, "severe_pct": 25.0, "severe_area_m2": 5102.0831, "total_area_m2": 9203.581, "severity_score": 0.3325, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 13177.7756, "severity_score": 0.0367, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_tuscaloosa-tornado_00000266_1297", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000266", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G31", "severity_score": 0.67, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G31", "row": 3, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 5102.0831, "total_area_m2": 5102.0831, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 8, "severe_count": 1, "severe_pct": 12.5, "severe_area_m2": 1581.6516, "total_area_m2": 13576.5278, "severity_score": 0.3725, "dominant_damage": "minor-damage"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 7, "severe_count": 2, "severe_pct": 28.57, "severe_area_m2": 1856.2267, "total_area_m2": 5154.6988, "severity_score": 0.2857, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_tuscaloosa-tornado_00000266_1298", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000266", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (tuscaloosa-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 6, "severe_footprint_m2": 12779.672, "major_damage_area_m2": 7747.2749, "destroyed_area_m2": 5032.3971, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_tuscaloosa-tornado_00000266_1299", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000266", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 6, "max_severe_in_cell": 4, "concentration_score": 0.6667, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 1, "Q1": 4, "Q2": 1, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_tuscaloosa-tornado_00000266_1300", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000266", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (tuscaloosa-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "tuscaloosa-tornado", "disaster_type": "wind", "capture_date": "2011-05-19T16:48:01.085Z", "gsd": 1.85690224170685, "damage_summary": {"total_buildings": 55, "classified_buildings": 40, "counts": {"no-damage": 22, "minor-damage": 12, "major-damage": 3, "destroyed": 3, "unclassified": 15}, "severe_count": 6, "severe_pct": 15.0, "damaged_pct": 45.0}, "area_summary": {"total_footprint_m2": 79636.9439, "severe_footprint_m2": 12779.672, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6667, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 6}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00001816_1301", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001816", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 31, "minor-damage": 0, "major-damage": 3, "destroyed": 9, "unclassified": 0}, "total_buildings": 43, "classified_buildings": 43, "damaged_count": 12, "severe_count": 12, "damaged_pct": 27.91, "severe_pct": 27.91, "severe_footprint_m2": 36858.9051, "total_footprint_m2": 104090.7023, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00001816_1302", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001816", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 8, "severe_footprint_m2": 29468.6046, "severe_pct_in_quadrant": 27.59, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 29, "severe_count": 8, "severe_pct": 27.59, "severe_area_m2": 29468.6046, "total_area_m2": 77567.497, "severity_score": 0.2417, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 12, "severe_count": 4, "severe_pct": 33.33, "severe_area_m2": 7390.3005, "total_area_m2": 22835.1721, "severity_score": 0.3333, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 3688.0332, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 3688.0332, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 29, "severe_count": 8, "severe_pct": 27.59, "severe_area_m2": 29468.6046, "total_area_m2": 77567.497, "severity_score": 0.2417, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 12, "severe_count": 4, "severe_pct": 33.33, "severe_area_m2": 7390.3005, "total_area_m2": 22835.1721, "severity_score": 0.3333, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00001816_1303", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001816", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G30", "severity_score": 0.445, "building_count": 6, "severe_count": 3, "class_histogram": {"no_damage": 3, "minor_damage": 0, "major_damage": 1, "destroyed": 2, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G30", "row": 3, "col": 0, "building_count": 6, "severe_count": 3, "severe_pct": 50.0, "severe_area_m2": 8726.1475, "total_area_m2": 17507.9306, "severity_score": 0.445, "dominant_damage": "no-damage"}, {"cell_id": "G20", "row": 2, "col": 0, "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 1008.7569, "total_area_m2": 6329.3831, "severity_score": 0.3333, "dominant_damage": "no-damage"}, {"cell_id": "G32", "row": 3, "col": 2, "building_count": 12, "severe_count": 4, "severe_pct": 33.33, "severe_area_m2": 7390.3005, "total_area_m2": 22835.1721, "severity_score": 0.3333, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_portugal-wildfire_00001816_1304", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001816", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "From the post-disaster fire scene (portugal-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 12, "severe_footprint_m2": 36858.9051, "major_damage_area_m2": 19751.4601, "destroyed_area_m2": 17107.445, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_portugal-wildfire_00001816_1305", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001816", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 12, "max_severe_in_cell": 8, "concentration_score": 0.6667, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 8, "Q3": 4}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00001816_1306", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001816", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 43, "classified_buildings": 43, "counts": {"no-damage": 31, "minor-damage": 0, "major-damage": 3, "destroyed": 9, "unclassified": 0}, "severe_count": 12, "severe_pct": 27.91, "damaged_pct": 27.91}, "area_summary": {"total_footprint_m2": 104090.7023, "severe_footprint_m2": 36858.9051, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6667, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 12}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-matthew_00000202_1307", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000202", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (hurricane-matthew) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 2, "major-damage": 5, "destroyed": 5, "unclassified": 6}, "total_buildings": 18, "classified_buildings": 12, "damaged_count": 12, "severe_count": 10, "damaged_pct": 100.0, "severe_pct": 83.33, "severe_footprint_m2": 9249.3087, "total_footprint_m2": 12597.4789, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-matthew_00000202_1308", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000202", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (hurricane-matthew), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 3, "severe_footprint_m2": 5160.0631, "severe_pct_in_quadrant": 50.0, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 6, "severe_count": 3, "severe_pct": 50.0, "severe_area_m2": 5160.0631, "total_area_m2": 6425.4602, "severity_score": 0.5, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 5, "severe_pct": 62.5, "severe_area_m2": 2523.5148, "total_area_m2": 3731.8787, "severity_score": 0.5012, "dominant_damage": "major-damage"}, {"cell_id": "Q0", "building_count": 4, "severe_count": 2, "severe_pct": 50.0, "severe_area_m2": 1565.7308, "total_area_m2": 2440.14, "severity_score": 0.5, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 2, "severe_pct": 50.0, "severe_area_m2": 1565.7308, "total_area_m2": 2440.14, "severity_score": 0.5, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 5, "severe_pct": 62.5, "severe_area_m2": 2523.5148, "total_area_m2": 3731.8787, "severity_score": 0.5012, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 6, "severe_count": 3, "severe_pct": 50.0, "severe_area_m2": 5160.0631, "total_area_m2": 6425.4602, "severity_score": 0.5, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-matthew_00000202_1309", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000202", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G20", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G20", "row": 2, "col": 0, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 480.442, "total_area_m2": 480.442, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 839.509, "total_area_m2": 839.509, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G32", "row": 3, "col": 2, "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 4991.703, "total_area_m2": 5573.928, "severity_score": 0.6667, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-matthew_00000202_1310", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000202", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (hurricane-matthew), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 10, "severe_footprint_m2": 9249.3087, "major_damage_area_m2": 2948.9224, "destroyed_area_m2": 6300.3863, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-matthew_00000202_1311", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000202", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 10, "max_severe_in_cell": 5, "concentration_score": 0.5, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 2, "Q1": 0, "Q2": 5, "Q3": 3}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-matthew_00000202_1312", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000202", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-matthew). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-matthew", "disaster_type": "wind", "capture_date": "2016-10-10T16:04:56.646Z", "gsd": 1.59234964847565, "damage_summary": {"total_buildings": 18, "classified_buildings": 12, "counts": {"no-damage": 0, "minor-damage": 2, "major-damage": 5, "destroyed": 5, "unclassified": 6}, "severe_count": 10, "severe_pct": 83.33, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 12597.4789, "severe_footprint_m2": 9249.3087, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 3980, "polygon_severe_count": 10, "pixel_histogram": {"0": 1044210, "1": 0, "2": 386, "3": 1292, "4": 2688, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 8.841, "3": 29.5923, "4": 61.5667, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_joplin-tornado_00000131_1313", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000131", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (joplin-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 10, "major-damage": 8, "destroyed": 17, "unclassified": 1}, "total_buildings": 36, "classified_buildings": 35, "damaged_count": 35, "severe_count": 25, "damaged_pct": 100.0, "severe_pct": 71.43, "severe_footprint_m2": 136829.9616, "total_footprint_m2": 187666.1629, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_joplin-tornado_00000131_1314", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000131", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (joplin-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 12, "severe_footprint_m2": 71763.5019, "severe_pct_in_quadrant": 52.17, "dominant_damage": "minor-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 23, "severe_count": 12, "severe_pct": 52.17, "severe_area_m2": 71763.5019, "total_area_m2": 122599.7032, "severity_score": 0.5648, "dominant_damage": "minor-damage"}, {"cell_id": "Q0", "building_count": 13, "severe_count": 13, "severe_pct": 100.0, "severe_area_m2": 65066.4597, "total_area_m2": 65066.4597, "severity_score": 0.9746, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 13, "severe_count": 13, "severe_pct": 100.0, "severe_area_m2": 65066.4597, "total_area_m2": 65066.4597, "severity_score": 0.9746, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 23, "severe_count": 12, "severe_pct": 52.17, "severe_area_m2": 71763.5019, "total_area_m2": 122599.7032, "severity_score": 0.5648, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_joplin-tornado_00000131_1315", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000131", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G10", "severity_score": 1.0, "building_count": 10, "severe_count": 10, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 10, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G10", "row": 1, "col": 0, "building_count": 10, "severe_count": 10, "severe_pct": 100.0, "severe_area_m2": 51869.3327, "total_area_m2": 51869.3327, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 13197.127, "total_area_m2": 13197.127, "severity_score": 0.89, "dominant_damage": "destroyed"}, {"cell_id": "G20", "row": 2, "col": 0, "building_count": 9, "severe_count": 7, "severe_pct": 77.78, "severe_area_m2": 51826.3663, "total_area_m2": 65627.6605, "severity_score": 0.7778, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_joplin-tornado_00000131_1316", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000131", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (joplin-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 25, "severe_footprint_m2": 136829.9616, "major_damage_area_m2": 41365.2392, "destroyed_area_m2": 95464.7224, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_joplin-tornado_00000131_1317", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000131", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 25, "max_severe_in_cell": 13, "concentration_score": 0.52, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 13, "Q1": 0, "Q2": 12, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_joplin-tornado_00000131_1318", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000131", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (joplin-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "joplin-tornado", "disaster_type": "wind", "capture_date": "2011-05-29T17:29:25.433Z", "gsd": 2.35255861282349, "damage_summary": {"total_buildings": 36, "classified_buildings": 35, "counts": {"no-damage": 0, "minor-damage": 10, "major-damage": 8, "destroyed": 17, "unclassified": 1}, "severe_count": 25, "severe_pct": 71.43, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 187666.1629, "severe_footprint_m2": 136829.9616, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.52, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 25}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_pinery-bushfire_00000103_1331", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000103", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a fire event (pinery-bushfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 11, "minor-damage": 0, "major-damage": 1, "destroyed": 1, "unclassified": 2}, "total_buildings": 15, "classified_buildings": 13, "damaged_count": 2, "severe_count": 2, "damaged_pct": 15.38, "severe_pct": 15.38, "severe_footprint_m2": 961.3134, "total_footprint_m2": 8547.3975, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_pinery-bushfire_00000103_1332", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000103", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "For a post-disaster fire scene (pinery-bushfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 2, "severe_footprint_m2": 961.3134, "severe_pct_in_quadrant": 13.33, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 15, "severe_count": 2, "severe_pct": 13.33, "severe_area_m2": 961.3134, "total_area_m2": 8547.3975, "severity_score": 0.1113, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 15, "severe_count": 2, "severe_pct": 13.33, "severe_area_m2": 961.3134, "total_area_m2": 8547.3975, "severity_score": 0.1113, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_pinery-bushfire_00000103_1333", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000103", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G30", "severity_score": 0.2087, "building_count": 8, "severe_count": 2, "class_histogram": {"no_damage": 4, "minor_damage": 0, "major_damage": 1, "destroyed": 1, "unclassified": 2}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G30", "row": 3, "col": 0, "building_count": 8, "severe_count": 2, "severe_pct": 25.0, "severe_area_m2": 961.3134, "total_area_m2": 3017.6883, "severity_score": 0.2087, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_pinery-bushfire_00000103_1334", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000103", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "From the post-disaster fire scene (pinery-bushfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 2, "severe_footprint_m2": 961.3134, "major_damage_area_m2": 671.4675, "destroyed_area_m2": 289.8459, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_pinery-bushfire_00000103_1335", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000103", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 2, "max_severe_in_cell": 2, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 2, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_pinery-bushfire_00000103_1336", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000103", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (pinery-bushfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "pinery-bushfire", "disaster_type": "fire", "capture_date": "2015-11-29T01:09:22.998Z", "gsd": 1.65928483009338, "damage_summary": {"total_buildings": 15, "classified_buildings": 13, "counts": {"no-damage": 11, "minor-damage": 0, "major-damage": 1, "destroyed": 1, "unclassified": 2}, "severe_count": 2, "severe_pct": 15.38, "damaged_pct": 15.38}, "area_summary": {"total_footprint_m2": 8547.3975, "severe_footprint_m2": 961.3134, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 2}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_pinery-bushfire_00000008_1349", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000008", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a fire event (pinery-bushfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 14, "minor-damage": 0, "major-damage": 2, "destroyed": 5, "unclassified": 0}, "total_buildings": 21, "classified_buildings": 21, "damaged_count": 7, "severe_count": 7, "damaged_pct": 33.33, "severe_pct": 33.33, "severe_footprint_m2": 13605.901, "total_footprint_m2": 27635.3608, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_pinery-bushfire_00000008_1350", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000008", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "For a post-disaster fire scene (pinery-bushfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 5, "severe_footprint_m2": 9455.465, "severe_pct_in_quadrant": 27.78, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 18, "severe_count": 5, "severe_pct": 27.78, "severe_area_m2": 9455.465, "total_area_m2": 20607.2297, "severity_score": 0.2594, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 4150.436, "total_area_m2": 7028.1311, "severity_score": 0.5567, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 4150.436, "total_area_m2": 7028.1311, "severity_score": 0.5567, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 18, "severe_count": 5, "severe_pct": 27.78, "severe_area_m2": 9455.465, "total_area_m2": 20607.2297, "severity_score": 0.2594, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_pinery-bushfire_00000008_1351", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000008", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G31", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G31", "row": 3, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 3603.5541, "total_area_m2": 3603.5541, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 3835.2428, "total_area_m2": 3835.2428, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 315.1932, "total_area_m2": 3192.8883, "severity_score": 0.5, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_pinery-bushfire_00000008_1352", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000008", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "From the post-disaster fire scene (pinery-bushfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 7, "severe_footprint_m2": 13605.901, "major_damage_area_m2": 6501.578, "destroyed_area_m2": 7104.323, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_pinery-bushfire_00000008_1353", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000008", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 7, "max_severe_in_cell": 5, "concentration_score": 0.7143, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 2, "Q1": 0, "Q2": 5, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_pinery-bushfire_00000008_1354", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000008", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (pinery-bushfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "pinery-bushfire", "disaster_type": "fire", "capture_date": "2015-11-29T01:09:22.998Z", "gsd": 1.65928483009338, "damage_summary": {"total_buildings": 21, "classified_buildings": 21, "counts": {"no-damage": 14, "minor-damage": 0, "major-damage": 2, "destroyed": 5, "unclassified": 0}, "severe_count": 7, "severe_pct": 33.33, "damaged_pct": 33.33}, "area_summary": {"total_footprint_m2": 27635.3608, "severe_footprint_m2": 13605.901, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.7143, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 7}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-matthew_00000090_1373", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000090", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (hurricane-matthew) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 40, "major-damage": 7, "destroyed": 4, "unclassified": 0}, "total_buildings": 51, "classified_buildings": 51, "damaged_count": 51, "severe_count": 11, "damaged_pct": 100.0, "severe_pct": 21.57, "severe_footprint_m2": 36193.6012, "total_footprint_m2": 151101.0507, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-matthew_00000090_1374", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000090", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (hurricane-matthew), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 8, "severe_footprint_m2": 26682.96, "severe_pct_in_quadrant": 42.11, "dominant_damage": "minor-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 19, "severe_count": 8, "severe_pct": 42.11, "severe_area_m2": 26682.96, "total_area_m2": 53447.8343, "severity_score": 0.5253, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 15, "severe_count": 2, "severe_pct": 13.33, "severe_area_m2": 7287.2304, "total_area_m2": 46615.5194, "severity_score": 0.3753, "dominant_damage": "minor-damage"}, {"cell_id": "Q0", "building_count": 9, "severe_count": 1, "severe_pct": 11.11, "severe_area_m2": 2223.4108, "total_area_m2": 31930.1954, "severity_score": 0.4044, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 9, "severe_count": 1, "severe_pct": 11.11, "severe_area_m2": 2223.4108, "total_area_m2": 31930.1954, "severity_score": 0.4044, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 15, "severe_count": 2, "severe_pct": 13.33, "severe_area_m2": 7287.2304, "total_area_m2": 46615.5194, "severity_score": 0.3753, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 19, "severe_count": 8, "severe_pct": 42.11, "severe_area_m2": 26682.96, "total_area_m2": 53447.8343, "severity_score": 0.5253, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 8, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 19107.5016, "severity_score": 0.33, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-matthew_00000090_1375", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000090", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G21", "severity_score": 0.667, "building_count": 10, "severe_count": 7, "class_histogram": {"no_damage": 0, "minor_damage": 3, "major_damage": 4, "destroyed": 3, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G21", "row": 2, "col": 1, "building_count": 10, "severe_count": 7, "severe_pct": 70.0, "severe_area_m2": 18523.172, "total_area_m2": 23970.0776, "severity_score": 0.667, "dominant_damage": "major-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 2223.4108, "total_area_m2": 7852.5341, "severity_score": 0.665, "dominant_damage": "minor-damage"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 8, "severe_count": 2, "severe_pct": 25.0, "severe_area_m2": 7287.2304, "total_area_m2": 35808.2178, "severity_score": 0.415, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-matthew_00000090_1376", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000090", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (hurricane-matthew), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 11, "severe_footprint_m2": 36193.6012, "major_damage_area_m2": 29604.7933, "destroyed_area_m2": 6588.8079, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-matthew_00000090_1377", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000090", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 11, "max_severe_in_cell": 8, "concentration_score": 0.7273, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 1, "Q1": 2, "Q2": 8, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-matthew_00000090_1378", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000090", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-matthew). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-matthew", "disaster_type": "wind", "capture_date": "2016-10-09T15:32:03.000Z", "gsd": 2.77336645, "damage_summary": {"total_buildings": 51, "classified_buildings": 51, "counts": {"no-damage": 0, "minor-damage": 40, "major-damage": 7, "destroyed": 4, "unclassified": 0}, "severe_count": 11, "severe_pct": 21.57, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 151101.0507, "severe_footprint_m2": 36193.6012, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.7273, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 5130, "polygon_severe_count": 11, "pixel_histogram": {"0": 1027040, "1": 0, "2": 16406, "3": 4172, "4": 958, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 76.1794, "3": 19.3722, "4": 4.4484, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000083_1379", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000083", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 31, "minor-damage": 7, "major-damage": 4, "destroyed": 2, "unclassified": 7}, "total_buildings": 51, "classified_buildings": 44, "damaged_count": 13, "severe_count": 6, "damaged_pct": 29.55, "severe_pct": 13.64, "severe_footprint_m2": 19991.2191, "total_footprint_m2": 120349.3372, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000083_1380", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000083", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 5, "severe_footprint_m2": 12573.9419, "severe_pct_in_quadrant": 11.9, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 42, "severe_count": 5, "severe_pct": 11.9, "severe_area_m2": 12573.9419, "total_area_m2": 95748.8326, "severity_score": 0.1348, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 1, "severe_pct": 11.11, "severe_area_m2": 7417.2772, "total_area_m2": 24600.5046, "severity_score": 0.1478, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 42, "severe_count": 5, "severe_pct": 11.9, "severe_area_m2": 12573.9419, "total_area_m2": 95748.8326, "severity_score": 0.1348, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 1, "severe_pct": 11.11, "severe_area_m2": 7417.2772, "total_area_m2": 24600.5046, "severity_score": 0.1478, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000083_1381", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000083", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G21", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G21", "row": 2, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 672.6017, "total_area_m2": 672.6017, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G32", "row": 3, "col": 2, "building_count": 9, "severe_count": 1, "severe_pct": 11.11, "severe_area_m2": 7417.2772, "total_area_m2": 24600.5046, "severity_score": 0.1478, "dominant_damage": "no-damage"}, {"cell_id": "G31", "row": 3, "col": 1, "building_count": 41, "severe_count": 4, "severe_pct": 9.76, "severe_area_m2": 11901.3402, "total_area_m2": 95076.2309, "severity_score": 0.1137, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000083_1382", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000083", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 6, "severe_footprint_m2": 19991.2191, "major_damage_area_m2": 18683.7642, "destroyed_area_m2": 1307.4549, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000083_1383", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000083", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 6, "max_severe_in_cell": 5, "concentration_score": 0.8333, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 5, "Q3": 1}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000083_1384", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000083", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 51, "classified_buildings": 44, "counts": {"no-damage": 31, "minor-damage": 7, "major-damage": 4, "destroyed": 2, "unclassified": 7}, "severe_count": 6, "severe_pct": 13.64, "damaged_pct": 29.55}, "area_summary": {"total_footprint_m2": 120349.3372, "severe_footprint_m2": 19991.2191, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.8333, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 6}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00001711_1391", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001711", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 104, "minor-damage": 0, "major-damage": 4, "destroyed": 8, "unclassified": 7}, "total_buildings": 123, "classified_buildings": 116, "damaged_count": 12, "severe_count": 12, "damaged_pct": 10.34, "severe_pct": 10.34, "severe_footprint_m2": 24960.4688, "total_footprint_m2": 356894.3569, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00001711_1392", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001711", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 8, "severe_footprint_m2": 12553.1407, "severe_pct_in_quadrant": 32.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 25, "severe_count": 8, "severe_pct": 32.0, "severe_area_m2": 12553.1407, "total_area_m2": 68493.6373, "severity_score": 0.2672, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 38, "severe_count": 3, "severe_pct": 7.89, "severe_area_m2": 11275.4097, "total_area_m2": 112636.6171, "severity_score": 0.0789, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 4, "severe_count": 1, "severe_pct": 25.0, "severe_area_m2": 1131.9184, "total_area_m2": 9053.2444, "severity_score": 0.25, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 25, "severe_count": 8, "severe_pct": 32.0, "severe_area_m2": 12553.1407, "total_area_m2": 68493.6373, "severity_score": 0.2672, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 4, "severe_count": 1, "severe_pct": 25.0, "severe_area_m2": 1131.9184, "total_area_m2": 9053.2444, "severity_score": 0.25, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 56, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 166710.8581, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 38, "severe_count": 3, "severe_pct": 7.89, "severe_area_m2": 11275.4097, "total_area_m2": 112636.6171, "severity_score": 0.0789, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00001711_1393", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001711", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1131.9184, "total_area_m2": 1131.9184, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 5, "severe_count": 2, "severe_pct": 40.0, "severe_area_m2": 1426.4316, "total_area_m2": 23868.5538, "severity_score": 0.4, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 16, "severe_count": 6, "severe_pct": 37.5, "severe_area_m2": 11126.7091, "total_area_m2": 34508.336, "severity_score": 0.2925, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_portugal-wildfire_00001711_1394", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001711", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "From the post-disaster fire scene (portugal-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 12, "severe_footprint_m2": 24960.4688, "major_damage_area_m2": 4604.8183, "destroyed_area_m2": 20355.6505, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_portugal-wildfire_00001711_1395", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001711", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 12, "max_severe_in_cell": 8, "concentration_score": 0.6667, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 8, "Q1": 1, "Q2": 0, "Q3": 3}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00001711_1396", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001711", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 123, "classified_buildings": 116, "counts": {"no-damage": 104, "minor-damage": 0, "major-damage": 4, "destroyed": 8, "unclassified": 7}, "severe_count": 12, "severe_pct": 10.34, "damaged_pct": 10.34}, "area_summary": {"total_footprint_m2": 356894.3569, "severe_footprint_m2": 24960.4688, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6667, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 12}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_joplin-tornado_00000046_1397", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000046", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (joplin-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 162, "minor-damage": 60, "major-damage": 7, "destroyed": 6, "unclassified": 10}, "total_buildings": 245, "classified_buildings": 235, "damaged_count": 73, "severe_count": 13, "damaged_pct": 31.06, "severe_pct": 5.53, "severe_footprint_m2": 43945.5391, "total_footprint_m2": 820588.257, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_joplin-tornado_00000046_1398", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000046", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (joplin-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 7, "severe_footprint_m2": 28760.6437, "severe_pct_in_quadrant": 12.73, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 55, "severe_count": 7, "severe_pct": 12.73, "severe_area_m2": 28760.6437, "total_area_m2": 200296.9967, "severity_score": 0.2233, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 56, "severe_count": 6, "severe_pct": 10.71, "severe_area_m2": 15184.8954, "total_area_m2": 152475.7026, "severity_score": 0.2073, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 61, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 214487.1772, "severity_score": 0.0379, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 61, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 214487.1772, "severity_score": 0.0379, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 73, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 253328.3805, "severity_score": 0.0588, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 56, "severe_count": 6, "severe_pct": 10.71, "severe_area_m2": 15184.8954, "total_area_m2": 152475.7026, "severity_score": 0.2073, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 55, "severe_count": 7, "severe_pct": 12.73, "severe_area_m2": 28760.6437, "total_area_m2": 200296.9967, "severity_score": 0.2233, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_joplin-tornado_00000046_1399", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000046", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G33", "severity_score": 0.4245, "building_count": 11, "severe_count": 5, "class_histogram": {"no_damage": 2, "minor_damage": 3, "major_damage": 4, "destroyed": 1, "unclassified": 1}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G33", "row": 3, "col": 3, "building_count": 11, "severe_count": 5, "severe_pct": 45.45, "severe_area_m2": 15590.7101, "total_area_m2": 27918.9381, "severity_score": 0.4245, "dominant_damage": "major-damage"}, {"cell_id": "G30", "row": 3, "col": 0, "building_count": 21, "severe_count": 3, "severe_pct": 14.29, "severe_area_m2": 3397.8064, "total_area_m2": 59782.7798, "severity_score": 0.2843, "dominant_damage": "minor-damage"}, {"cell_id": "G31", "row": 3, "col": 1, "building_count": 5, "severe_count": 1, "severe_pct": 20.0, "severe_area_m2": 2149.4817, "total_area_m2": 8741.0617, "severity_score": 0.266, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_joplin-tornado_00000046_1400", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000046", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (joplin-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 13, "severe_footprint_m2": 43945.5391, "major_damage_area_m2": 36098.4595, "destroyed_area_m2": 7847.0796, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_joplin-tornado_00000046_1401", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000046", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 13, "max_severe_in_cell": 7, "concentration_score": 0.5385, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 6, "Q3": 7}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_joplin-tornado_00000046_1402", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000046", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (joplin-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "joplin-tornado", "disaster_type": "wind", "capture_date": "2011-05-29T17:29:25.433Z", "gsd": 2.35255861282349, "damage_summary": {"total_buildings": 245, "classified_buildings": 235, "counts": {"no-damage": 162, "minor-damage": 60, "major-damage": 7, "destroyed": 6, "unclassified": 10}, "severe_count": 13, "severe_pct": 5.53, "damaged_pct": 31.06}, "area_summary": {"total_footprint_m2": 820588.257, "severe_footprint_m2": 43945.5391, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5385, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 13}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00000588_1403", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000588", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 189, "minor-damage": 1, "major-damage": 5, "destroyed": 3, "unclassified": 10}, "total_buildings": 208, "classified_buildings": 198, "damaged_count": 9, "severe_count": 8, "damaged_pct": 4.55, "severe_pct": 4.04, "severe_footprint_m2": 11252.7199, "total_footprint_m2": 324314.0649, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00000588_1404", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000588", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 4, "severe_footprint_m2": 7159.7092, "severe_pct_in_quadrant": 10.81, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 37, "severe_count": 4, "severe_pct": 10.81, "severe_area_m2": 7159.7092, "total_area_m2": 64169.2931, "severity_score": 0.0992, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 84, "severe_count": 2, "severe_pct": 2.38, "severe_area_m2": 2546.1611, "total_area_m2": 131669.303, "severity_score": 0.016, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 41, "severe_count": 2, "severe_pct": 4.88, "severe_area_m2": 1546.8496, "total_area_m2": 63554.5982, "severity_score": 0.0407, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 41, "severe_count": 2, "severe_pct": 4.88, "severe_area_m2": 1546.8496, "total_area_m2": 63554.5982, "severity_score": 0.0407, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 37, "severe_count": 4, "severe_pct": 10.81, "severe_area_m2": 7159.7092, "total_area_m2": 64169.2931, "severity_score": 0.0992, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 84, "severe_count": 2, "severe_pct": 2.38, "severe_area_m2": 2546.1611, "total_area_m2": 131669.303, "severity_score": 0.016, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 46, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 64920.8706, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00000588_1405", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000588", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.5567, "building_count": 3, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 1, "major_damage": 2, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 1546.8496, "total_area_m2": 2763.9867, "severity_score": 0.5567, "dominant_damage": "major-damage"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 11, "severe_count": 4, "severe_pct": 36.36, "severe_area_m2": 7159.7092, "total_area_m2": 24849.3429, "severity_score": 0.3336, "dominant_damage": "no-damage"}, {"cell_id": "G31", "row": 3, "col": 1, "building_count": 18, "severe_count": 2, "severe_pct": 11.11, "severe_area_m2": 2546.1611, "total_area_m2": 21309.0213, "severity_score": 0.0744, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_portugal-wildfire_00000588_1406", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000588", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "From the post-disaster fire scene (portugal-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 8, "severe_footprint_m2": 11252.7199, "major_damage_area_m2": 6829.2413, "destroyed_area_m2": 4423.4786, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_portugal-wildfire_00000588_1407", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000588", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 8, "max_severe_in_cell": 4, "concentration_score": 0.5, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 2, "Q1": 4, "Q2": 2, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00000588_1408", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000588", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 208, "classified_buildings": 198, "counts": {"no-damage": 189, "minor-damage": 1, "major-damage": 5, "destroyed": 3, "unclassified": 10}, "severe_count": 8, "severe_pct": 4.04, "damaged_pct": 4.55}, "area_summary": {"total_footprint_m2": 324314.0649, "severe_footprint_m2": 11252.7199, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 8}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_midwest-flooding_00000340_1415", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000340", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a flooding event (midwest-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 5, "major-damage": 13, "destroyed": 6, "unclassified": 0}, "total_buildings": 24, "classified_buildings": 24, "damaged_count": 24, "severe_count": 19, "damaged_pct": 100.0, "severe_pct": 79.17, "severe_footprint_m2": 120702.1714, "total_footprint_m2": 129757.3627, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_midwest-flooding_00000340_1416", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000340", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "For a post-disaster flooding scene (midwest-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 13, "severe_footprint_m2": 84354.0205, "severe_pct_in_quadrant": 100.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 13, "severe_count": 13, "severe_pct": 100.0, "severe_area_m2": 84354.0205, "total_area_m2": 84354.0205, "severity_score": 0.8223, "dominant_damage": "major-damage"}, {"cell_id": "Q0", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 33931.6759, "total_area_m2": 33931.6759, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 4, "severe_pct": 44.44, "severe_area_m2": 2416.475, "total_area_m2": 11471.6663, "severity_score": 0.4811, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 33931.6759, "total_area_m2": 33931.6759, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 13, "severe_count": 13, "severe_pct": 100.0, "severe_area_m2": 84354.0205, "total_area_m2": 84354.0205, "severity_score": 0.8223, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 4, "severe_pct": 44.44, "severe_area_m2": 2416.475, "total_area_m2": 11471.6663, "severity_score": 0.4811, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_midwest-flooding_00000340_1417", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000340", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G20", "severity_score": 1.0, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 2, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G20", "row": 2, "col": 0, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 22277.4949, "total_area_m2": 22277.4949, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G21", "row": 2, "col": 1, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 3394.4952, "total_area_m2": 3394.4952, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 33931.6759, "total_area_m2": 33931.6759, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_midwest-flooding_00000340_1418", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000340", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "From the post-disaster flooding scene (midwest-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 19, "severe_footprint_m2": 120702.1714, "major_damage_area_m2": 95030.1813, "destroyed_area_m2": 25671.9901, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_midwest-flooding_00000340_1419", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000340", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 19, "max_severe_in_cell": 13, "concentration_score": 0.6842, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 2, "Q1": 0, "Q2": 13, "Q3": 4}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_midwest-flooding_00000340_1420", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000340", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (midwest-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "midwest-flooding", "disaster_type": "flooding", "capture_date": "2019-05-30T17:35:04.000Z", "gsd": 1.2524601, "damage_summary": {"total_buildings": 24, "classified_buildings": 24, "counts": {"no-damage": 0, "minor-damage": 5, "major-damage": 13, "destroyed": 6, "unclassified": 0}, "severe_count": 19, "severe_pct": 79.17, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 129757.3627, "severe_footprint_m2": 120702.1714, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6842, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 78802, "polygon_severe_count": 19, "pixel_histogram": {"0": 963732, "1": 0, "2": 6042, "3": 62033, "4": 16769, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 7.1213, "3": 73.1142, "4": 19.7645, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_tuscaloosa-tornado_00000095_1421", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000095", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (tuscaloosa-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 29, "minor-damage": 56, "major-damage": 15, "destroyed": 14, "unclassified": 6}, "total_buildings": 120, "classified_buildings": 114, "damaged_count": 85, "severe_count": 29, "damaged_pct": 74.56, "severe_pct": 25.44, "severe_footprint_m2": 116151.4108, "total_footprint_m2": 638267.5722, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_tuscaloosa-tornado_00000095_1422", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000095", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (tuscaloosa-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 18, "severe_footprint_m2": 83470.8871, "severe_pct_in_quadrant": 54.55, "dominant_damage": "minor-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 33, "severe_count": 18, "severe_pct": 54.55, "severe_area_m2": 83470.8871, "total_area_m2": 148982.7896, "severity_score": 0.5555, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 27, "severe_count": 4, "severe_pct": 14.81, "severe_area_m2": 17133.698, "total_area_m2": 154630.6495, "severity_score": 0.3193, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 40, "severe_count": 7, "severe_pct": 17.5, "severe_area_m2": 15546.8257, "total_area_m2": 136833.4577, "severity_score": 0.3648, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 20, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 197820.6754, "severity_score": 0.0495, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 40, "severe_count": 7, "severe_pct": 17.5, "severe_area_m2": 15546.8257, "total_area_m2": 136833.4577, "severity_score": 0.3648, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 27, "severe_count": 4, "severe_pct": 14.81, "severe_area_m2": 17133.698, "total_area_m2": 154630.6495, "severity_score": 0.3193, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 33, "severe_count": 18, "severe_pct": 54.55, "severe_area_m2": 83470.8871, "total_area_m2": 148982.7896, "severity_score": 0.5555, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_tuscaloosa-tornado_00000095_1423", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000095", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G33", "severity_score": 0.945, "building_count": 6, "severe_count": 6, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 1, "destroyed": 5, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G33", "row": 3, "col": 3, "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 42282.9464, "total_area_m2": 42282.9464, "severity_score": 0.945, "dominant_damage": "destroyed"}, {"cell_id": "G30", "row": 3, "col": 0, "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 1319.5309, "total_area_m2": 10001.7074, "severity_score": 0.665, "dominant_damage": "minor-damage"}, {"cell_id": "G32", "row": 3, "col": 2, "building_count": 9, "severe_count": 6, "severe_pct": 66.67, "severe_area_m2": 20038.9464, "total_area_m2": 25032.3745, "severity_score": 0.63, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_tuscaloosa-tornado_00000095_1424", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000095", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (tuscaloosa-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 29, "severe_footprint_m2": 116151.4108, "major_damage_area_m2": 64949.131, "destroyed_area_m2": 51202.2798, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_tuscaloosa-tornado_00000095_1425", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000095", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 29, "max_severe_in_cell": 18, "concentration_score": 0.6207, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 7, "Q2": 4, "Q3": 18}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_tuscaloosa-tornado_00000095_1426", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000095", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (tuscaloosa-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "tuscaloosa-tornado", "disaster_type": "wind", "capture_date": "2011-05-19T16:48:01.085Z", "gsd": 1.85690224170685, "damage_summary": {"total_buildings": 120, "classified_buildings": 114, "counts": {"no-damage": 29, "minor-damage": 56, "major-damage": 15, "destroyed": 14, "unclassified": 6}, "severe_count": 29, "severe_pct": 25.44, "damaged_pct": 74.56}, "area_summary": {"total_footprint_m2": 638267.5722, "severe_footprint_m2": 116151.4108, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6207, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 29}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_tuscaloosa-tornado_00000030_1427", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000030", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (tuscaloosa-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 52, "minor-damage": 48, "major-damage": 9, "destroyed": 12, "unclassified": 3}, "total_buildings": 124, "classified_buildings": 121, "damaged_count": 69, "severe_count": 21, "damaged_pct": 57.02, "severe_pct": 17.36, "severe_footprint_m2": 115258.9714, "total_footprint_m2": 426384.8559, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_tuscaloosa-tornado_00000030_1428", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000030", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (tuscaloosa-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 2, "severe_footprint_m2": 58530.9733, "severe_pct_in_quadrant": 5.41, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 37, "severe_count": 2, "severe_pct": 5.41, "severe_area_m2": 58530.9733, "total_area_m2": 175380.0805, "severity_score": 0.1343, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 57, "severe_count": 18, "severe_pct": 31.58, "severe_area_m2": 51195.2203, "total_area_m2": 145538.1548, "severity_score": 0.4489, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 21, "severe_count": 1, "severe_pct": 4.76, "severe_area_m2": 5532.7778, "total_area_m2": 64205.4897, "severity_score": 0.1262, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 57, "severe_count": 18, "severe_pct": 31.58, "severe_area_m2": 51195.2203, "total_area_m2": 145538.1548, "severity_score": 0.4489, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 37, "severe_count": 2, "severe_pct": 5.41, "severe_area_m2": 58530.9733, "total_area_m2": 175380.0805, "severity_score": 0.1343, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 21, "severe_count": 1, "severe_pct": 4.76, "severe_area_m2": 5532.7778, "total_area_m2": 64205.4897, "severity_score": 0.1262, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 41261.1309, "severity_score": 0.0733, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_tuscaloosa-tornado_00000030_1429", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000030", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G00", "severity_score": 0.8343, "building_count": 14, "severe_count": 13, "class_histogram": {"no_damage": 0, "minor_damage": 1, "major_damage": 5, "destroyed": 8, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G00", "row": 0, "col": 0, "building_count": 14, "severe_count": 13, "severe_pct": 92.86, "severe_area_m2": 36305.3988, "total_area_m2": 40955.1452, "severity_score": 0.8343, "dominant_damage": "destroyed"}, {"cell_id": "G30", "row": 3, "col": 0, "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 5532.7778, "total_area_m2": 6834.0386, "severity_score": 0.665, "dominant_damage": "minor-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 12, "severe_count": 4, "severe_pct": 33.33, "severe_area_m2": 12269.1441, "total_area_m2": 27051.2912, "severity_score": 0.4708, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_tuscaloosa-tornado_00000030_1430", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000030", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (tuscaloosa-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 21, "severe_footprint_m2": 115258.9714, "major_damage_area_m2": 75099.7915, "destroyed_area_m2": 40159.1799, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_tuscaloosa-tornado_00000030_1431", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000030", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 21, "max_severe_in_cell": 18, "concentration_score": 0.8571, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 18, "Q1": 2, "Q2": 1, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_tuscaloosa-tornado_00000030_1432", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000030", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (tuscaloosa-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "tuscaloosa-tornado", "disaster_type": "wind", "capture_date": "2011-05-19T16:48:01.085Z", "gsd": 1.85690224170685, "damage_summary": {"total_buildings": 124, "classified_buildings": 121, "counts": {"no-damage": 52, "minor-damage": 48, "major-damage": 9, "destroyed": 12, "unclassified": 3}, "severe_count": 21, "severe_pct": 17.36, "damaged_pct": 57.02}, "area_summary": {"total_footprint_m2": 426384.8559, "severe_footprint_m2": 115258.9714, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.8571, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 21}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000303_1439", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000303", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 57, "minor-damage": 25, "major-damage": 19, "destroyed": 25, "unclassified": 1}, "total_buildings": 127, "classified_buildings": 126, "damaged_count": 69, "severe_count": 44, "damaged_pct": 54.76, "severe_pct": 34.92, "severe_footprint_m2": 149991.3422, "total_footprint_m2": 492332.2619, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000303_1440", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000303", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 17, "severe_footprint_m2": 54982.9559, "severe_pct_in_quadrant": 38.64, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 44, "severe_count": 17, "severe_pct": 38.64, "severe_area_m2": 54982.9559, "total_area_m2": 143487.9916, "severity_score": 0.4089, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 29, "severe_count": 11, "severe_pct": 37.93, "severe_area_m2": 34916.9563, "total_area_m2": 83618.9904, "severity_score": 0.3793, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 17, "severe_count": 7, "severe_pct": 41.18, "severe_area_m2": 32644.5211, "total_area_m2": 138185.11, "severity_score": 0.4118, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 17, "severe_count": 7, "severe_pct": 41.18, "severe_area_m2": 32644.5211, "total_area_m2": 138185.11, "severity_score": 0.4118, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 44, "severe_count": 17, "severe_pct": 38.64, "severe_area_m2": 54982.9559, "total_area_m2": 143487.9916, "severity_score": 0.4089, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 29, "severe_count": 11, "severe_pct": 37.93, "severe_area_m2": 34916.9563, "total_area_m2": 83618.9904, "severity_score": 0.3793, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 37, "severe_count": 9, "severe_pct": 24.32, "severe_area_m2": 27446.9089, "total_area_m2": 127040.1699, "severity_score": 0.27, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000303_1441", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000303", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G11", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G11", "row": 1, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 2951.4034, "total_area_m2": 2951.4034, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 4, "severe_count": 3, "severe_pct": 75.0, "severe_area_m2": 16339.9165, "total_area_m2": 29631.9563, "severity_score": 0.75, "dominant_damage": "destroyed"}, {"cell_id": "G21", "row": 2, "col": 1, "building_count": 6, "severe_count": 4, "severe_pct": 66.67, "severe_area_m2": 11685.2927, "total_area_m2": 20258.3865, "severity_score": 0.6117, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-michael_00000303_1442", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000303", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (hurricane-michael), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 44, "severe_footprint_m2": 149991.3422, "major_damage_area_m2": 75675.2144, "destroyed_area_m2": 74316.1278, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-michael_00000303_1443", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000303", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 44, "max_severe_in_cell": 17, "concentration_score": 0.3864, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 7, "Q1": 17, "Q2": 11, "Q3": 9}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000303_1444", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000303", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 127, "classified_buildings": 126, "counts": {"no-damage": 57, "minor-damage": 25, "major-damage": 19, "destroyed": 25, "unclassified": 1}, "severe_count": 44, "severe_pct": 34.92, "damaged_pct": 54.76}, "area_summary": {"total_footprint_m2": 492332.2619, "severe_footprint_m2": 149991.3422, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.3864, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 37373, "polygon_severe_count": 44, "pixel_histogram": {"0": 926163, "1": 55629, "2": 29411, "3": 18776, "4": 18597, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 45.4437, "2": 24.026, "3": 15.3382, "4": 15.192, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000219_1445", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000219", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 82, "minor-damage": 27, "major-damage": 1, "destroyed": 3, "unclassified": 0}, "total_buildings": 113, "classified_buildings": 113, "damaged_count": 31, "severe_count": 4, "damaged_pct": 27.43, "severe_pct": 3.54, "severe_footprint_m2": 15026.4051, "total_footprint_m2": 571268.0699, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000219_1446", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000219", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 3, "severe_footprint_m2": 10827.4615, "severe_pct_in_quadrant": 7.89, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 38, "severe_count": 3, "severe_pct": 7.89, "severe_area_m2": 10827.4615, "total_area_m2": 185057.5487, "severity_score": 0.1484, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 21, "severe_count": 1, "severe_pct": 4.76, "severe_area_m2": 4198.9436, "total_area_m2": 113404.2217, "severity_score": 0.1576, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 29, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 131407.1121, "severity_score": 0.0228, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 21, "severe_count": 1, "severe_pct": 4.76, "severe_area_m2": 4198.9436, "total_area_m2": 113404.2217, "severity_score": 0.1576, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 38, "severe_count": 3, "severe_pct": 7.89, "severe_area_m2": 10827.4615, "total_area_m2": 185057.5487, "severity_score": 0.1484, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 29, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 131407.1121, "severity_score": 0.0228, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 25, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 141399.1874, "severity_score": 0.1188, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000219_1447", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000219", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G11", "severity_score": 0.2843, "building_count": 7, "severe_count": 1, "class_histogram": {"no_damage": 3, "minor_damage": 3, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G11", "row": 1, "col": 1, "building_count": 7, "severe_count": 1, "severe_pct": 14.29, "severe_area_m2": 4198.9436, "total_area_m2": 35978.3364, "severity_score": 0.2843, "dominant_damage": "no-damage"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 10, "severe_count": 1, "severe_pct": 10.0, "severe_area_m2": 2573.3062, "total_area_m2": 43665.2699, "severity_score": 0.232, "dominant_damage": "no-damage"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 8, "severe_count": 2, "severe_pct": 25.0, "severe_area_m2": 8254.1553, "total_area_m2": 36064.7257, "severity_score": 0.2087, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-michael_00000219_1448", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000219", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (hurricane-michael), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 4, "severe_footprint_m2": 15026.4051, "major_damage_area_m2": 5947.2224, "destroyed_area_m2": 9079.1827, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-michael_00000219_1449", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000219", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 4, "max_severe_in_cell": 3, "concentration_score": 0.75, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 1, "Q1": 3, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000219_1450", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000219", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 113, "classified_buildings": 113, "counts": {"no-damage": 82, "minor-damage": 27, "major-damage": 1, "destroyed": 3, "unclassified": 0}, "severe_count": 4, "severe_pct": 3.54, "damaged_pct": 27.43}, "area_summary": {"total_footprint_m2": 571268.0699, "severe_footprint_m2": 15026.4051, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.75, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 3723, "polygon_severe_count": 4, "pixel_histogram": {"0": 907246, "1": 97192, "2": 40415, "3": 1463, "4": 2260, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 68.7695, "2": 28.5962, "3": 1.0352, "4": 1.5991, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000167_1457", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000167", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 0, "major-damage": 20, "destroyed": 6, "unclassified": 0}, "total_buildings": 26, "classified_buildings": 26, "damaged_count": 26, "severe_count": 26, "damaged_pct": 100.0, "severe_pct": 100.0, "severe_footprint_m2": 48508.0547, "total_footprint_m2": 48508.0547, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000167_1458", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000167", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 14, "severe_footprint_m2": 26494.3948, "severe_pct_in_quadrant": 100.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 14, "severe_count": 14, "severe_pct": 100.0, "severe_area_m2": 26494.3948, "total_area_m2": 26494.3948, "severity_score": 0.7171, "dominant_damage": "major-damage"}, {"cell_id": "Q0", "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 14980.4829, "total_area_m2": 14980.4829, "severity_score": 0.835, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 7033.177, "total_area_m2": 7033.177, "severity_score": 0.67, "dominant_damage": "major-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 8, "severe_count": 8, "severe_pct": 100.0, "severe_area_m2": 14980.4829, "total_area_m2": 14980.4829, "severity_score": 0.835, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 14, "severe_count": 14, "severe_pct": 100.0, "severe_area_m2": 26494.3948, "total_area_m2": 26494.3948, "severity_score": 0.7171, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 7033.177, "total_area_m2": 7033.177, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000167_1459", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000167", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G11", "severity_score": 0.89, "building_count": 6, "severe_count": 6, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 2, "destroyed": 4, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G11", "row": 1, "col": 1, "building_count": 6, "severe_count": 6, "severe_pct": 100.0, "severe_area_m2": 10343.8333, "total_area_m2": 10343.8333, "severity_score": 0.89, "dominant_damage": "destroyed"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 3670.9579, "total_area_m2": 3670.9579, "severity_score": 0.78, "dominant_damage": "major-damage"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 11, "severe_count": 11, "severe_pct": 100.0, "severe_area_m2": 22823.4369, "total_area_m2": 22823.4369, "severity_score": 0.7, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000167_1460", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000167", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 26, "severe_footprint_m2": 48508.0547, "major_damage_area_m2": 39185.2112, "destroyed_area_m2": 9322.8435, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000167_1461", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000167", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 26, "max_severe_in_cell": 14, "concentration_score": 0.5385, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 8, "Q1": 14, "Q2": 0, "Q3": 4}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000167_1462", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000167", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:17:02.085Z", "gsd": 2.9729018, "damage_summary": {"total_buildings": 26, "classified_buildings": 26, "counts": {"no-damage": 0, "minor-damage": 0, "major-damage": 20, "destroyed": 6, "unclassified": 0}, "severe_count": 26, "severe_pct": 100.0, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 48508.0547, "severe_footprint_m2": 48508.0547, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5385, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 26}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_sunda-tsunami_00000005_1475", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000005", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a tsunami event (sunda-tsunami) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 43, "minor-damage": 0, "major-damage": 2, "destroyed": 6, "unclassified": 0}, "total_buildings": 51, "classified_buildings": 51, "damaged_count": 8, "severe_count": 8, "damaged_pct": 15.69, "severe_pct": 15.69, "severe_footprint_m2": 3945.1524, "total_footprint_m2": 61683.5799, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_sunda-tsunami_00000005_1476", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000005", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "mixed-severe", "question": "For a post-disaster tsunami scene (sunda-tsunami), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 8, "severe_footprint_m2": 3945.1524, "severe_pct_in_quadrant": 15.69, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 51, "severe_count": 8, "severe_pct": 15.69, "severe_area_m2": 3945.1524, "total_area_m2": 61683.5799, "severity_score": 0.1439, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 51, "severe_count": 8, "severe_pct": 15.69, "severe_area_m2": 3945.1524, "total_area_m2": 61683.5799, "severity_score": 0.1439, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_sunda-tsunami_00000005_1477", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000005", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster tsunami satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 0.4893, "building_count": 15, "severe_count": 8, "class_histogram": {"no_damage": 7, "minor_damage": 0, "major_damage": 2, "destroyed": 6, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 15, "severe_count": 8, "severe_pct": 53.33, "severe_area_m2": 3945.1524, "total_area_m2": 12400.3051, "severity_score": 0.4893, "dominant_damage": "no-damage"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_sunda-tsunami_00000005_1478", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000005", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "mixed-severe", "question": "From the post-disaster tsunami scene (sunda-tsunami), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 8, "severe_footprint_m2": 3945.1524, "major_damage_area_m2": 1487.2372, "destroyed_area_m2": 2457.9152, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_sunda-tsunami_00000005_1479", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000005", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster tsunami scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 8, "max_severe_in_cell": 8, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 8, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_sunda-tsunami_00000005_1480", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000005", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster tsunami scene (sunda-tsunami). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "sunda-tsunami", "disaster_type": "tsunami", "capture_date": "2019-01-02T03:17:07.000Z", "gsd": 1.7746705, "damage_summary": {"total_buildings": 51, "classified_buildings": 51, "counts": {"no-damage": 43, "minor-damage": 0, "major-damage": 2, "destroyed": 6, "unclassified": 0}, "severe_count": 8, "severe_pct": 15.69, "damaged_pct": 15.69}, "area_summary": {"total_footprint_m2": 61683.5799, "severe_footprint_m2": 3945.1524, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 8}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000373_1481", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000373", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 88, "minor-damage": 26, "major-damage": 9, "destroyed": 4, "unclassified": 0}, "total_buildings": 127, "classified_buildings": 127, "damaged_count": 39, "severe_count": 13, "damaged_pct": 30.71, "severe_pct": 10.24, "severe_footprint_m2": 28674.5262, "total_footprint_m2": 248980.5378, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000373_1482", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000373", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 4, "severe_footprint_m2": 13529.8954, "severe_pct_in_quadrant": 16.67, "dominant_damage": "minor-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 24, "severe_count": 4, "severe_pct": 16.67, "severe_area_m2": 13529.8954, "total_area_m2": 64962.7303, "severity_score": 0.3454, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 7, "severe_pct": 100.0, "severe_area_m2": 9818.0978, "total_area_m2": 9818.0978, "severity_score": 0.8114, "dominant_damage": "major-damage"}, {"cell_id": "Q0", "building_count": 80, "severe_count": 2, "severe_pct": 2.5, "severe_area_m2": 5326.533, "total_area_m2": 134767.8234, "severity_score": 0.0291, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 80, "severe_count": 2, "severe_pct": 2.5, "severe_area_m2": 5326.533, "total_area_m2": 134767.8234, "severity_score": 0.0291, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 16, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 39431.8863, "severity_score": 0.1444, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 7, "severe_count": 7, "severe_pct": 100.0, "severe_area_m2": 9818.0978, "total_area_m2": 9818.0978, "severity_score": 0.8114, "dominant_damage": "major-damage"}, {"cell_id": "Q3", "building_count": 24, "severe_count": 4, "severe_pct": 16.67, "severe_area_m2": 13529.8954, "total_area_m2": 64962.7303, "severity_score": 0.3454, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000373_1483", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000373", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G31", "severity_score": 1.0, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 2, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G31", "row": 3, "col": 1, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 4820.9872, "total_area_m2": 4820.9872, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G21", "row": 2, "col": 1, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 846.3244, "total_area_m2": 846.3244, "severity_score": 0.835, "dominant_damage": "major-damage"}, {"cell_id": "G20", "row": 2, "col": 0, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 4150.7862, "total_area_m2": 4150.7862, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000373_1484", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000373", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 13, "severe_footprint_m2": 28674.5262, "major_damage_area_m2": 21980.3651, "destroyed_area_m2": 6694.1611, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000373_1485", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000373", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 13, "max_severe_in_cell": 7, "concentration_score": 0.5385, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 2, "Q1": 0, "Q2": 7, "Q3": 4}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000373_1486", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000373", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 127, "classified_buildings": 127, "counts": {"no-damage": 88, "minor-damage": 26, "major-damage": 9, "destroyed": 4, "unclassified": 0}, "severe_count": 13, "severe_pct": 10.24, "damaged_pct": 30.71}, "area_summary": {"total_footprint_m2": 248980.5378, "severe_footprint_m2": 28674.5262, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5385, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 13}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_joplin-tornado_00000037_1487", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000037", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (joplin-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 88, "minor-damage": 29, "major-damage": 3, "destroyed": 4, "unclassified": 21}, "total_buildings": 145, "classified_buildings": 124, "damaged_count": 36, "severe_count": 7, "damaged_pct": 29.03, "severe_pct": 5.65, "severe_footprint_m2": 22827.5758, "total_footprint_m2": 529535.8885, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_joplin-tornado_00000037_1488", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000037", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (joplin-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 6, "severe_footprint_m2": 17686.1379, "severe_pct_in_quadrant": 10.17, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 59, "severe_count": 6, "severe_pct": 10.17, "severe_area_m2": 17686.1379, "total_area_m2": 208208.8063, "severity_score": 0.1856, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 18, "severe_count": 1, "severe_pct": 5.56, "severe_area_m2": 5141.4379, "total_area_m2": 66016.974, "severity_score": 0.1472, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 25, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 62792.7021, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 25, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 62792.7021, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 43, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 192517.4061, "severity_score": 0.046, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 18, "severe_count": 1, "severe_pct": 5.56, "severe_area_m2": 5141.4379, "total_area_m2": 66016.974, "severity_score": 0.1472, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 59, "severe_count": 6, "severe_pct": 10.17, "severe_area_m2": 17686.1379, "total_area_m2": 208208.8063, "severity_score": 0.1856, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_joplin-tornado_00000037_1489", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000037", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G33", "severity_score": 0.2935, "building_count": 17, "severe_count": 4, "class_histogram": {"no_damage": 4, "minor_damage": 3, "major_damage": 0, "destroyed": 4, "unclassified": 6}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G33", "row": 3, "col": 3, "building_count": 17, "severe_count": 4, "severe_pct": 23.53, "severe_area_m2": 5329.378, "total_area_m2": 60569.4911, "severity_score": 0.2935, "dominant_damage": "no-damage"}, {"cell_id": "G31", "row": 3, "col": 1, "building_count": 7, "severe_count": 1, "severe_pct": 14.29, "severe_area_m2": 5141.4379, "total_area_m2": 27602.1286, "severity_score": 0.2843, "dominant_damage": "minor-damage"}, {"cell_id": "G32", "row": 3, "col": 2, "building_count": 14, "severe_count": 1, "severe_pct": 7.14, "severe_area_m2": 9329.0283, "total_area_m2": 52792.678, "severity_score": 0.1893, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_joplin-tornado_00000037_1490", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000037", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (joplin-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 7, "severe_footprint_m2": 22827.5758, "major_damage_area_m2": 17498.1978, "destroyed_area_m2": 5329.378, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_joplin-tornado_00000037_1491", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000037", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 7, "max_severe_in_cell": 6, "concentration_score": 0.8571, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 1, "Q3": 6}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_joplin-tornado_00000037_1492", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000037", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (joplin-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "joplin-tornado", "disaster_type": "wind", "capture_date": "2011-05-29T17:29:25.433Z", "gsd": 2.35255861282349, "damage_summary": {"total_buildings": 145, "classified_buildings": 124, "counts": {"no-damage": 88, "minor-damage": 29, "major-damage": 3, "destroyed": 4, "unclassified": 21}, "severe_count": 7, "severe_pct": 5.65, "damaged_pct": 29.03}, "area_summary": {"total_footprint_m2": 529535.8885, "severe_footprint_m2": 22827.5758, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.8571, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 7}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_mexico-earthquake_00000142_1493", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000142", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a earthquake event (mexico-earthquake) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 163, "minor-damage": 1, "major-damage": 1, "destroyed": 1, "unclassified": 11}, "total_buildings": 177, "classified_buildings": 166, "damaged_count": 3, "severe_count": 2, "damaged_pct": 1.81, "severe_pct": 1.2, "severe_footprint_m2": 6777.0004, "total_footprint_m2": 1329353.3963, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_mexico-earthquake_00000142_1494", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000142", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "mixed-severe", "question": "For a post-disaster earthquake scene (mexico-earthquake), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 2, "severe_footprint_m2": 6777.0004, "severe_pct_in_quadrant": 2.15, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 93, "severe_count": 2, "severe_pct": 2.15, "severe_area_m2": 6777.0004, "total_area_m2": 269992.4151, "severity_score": 0.018, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 41, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 343532.4558, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 25, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 334910.0762, "severity_score": 0.0132, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 41, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 343532.4558, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 25, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 334910.0762, "severity_score": 0.0132, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 18, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 380918.4492, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 93, "severe_count": 2, "severe_pct": 2.15, "severe_area_m2": 6777.0004, "total_area_m2": 269992.4151, "severity_score": 0.018, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_mexico-earthquake_00000142_1495", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000142", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster earthquake satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G13", "severity_score": 0.066, "building_count": 5, "severe_count": 0, "class_histogram": {"no_damage": 4, "minor_damage": 1, "major_damage": 0, "destroyed": 0, "unclassified": 0}, "dominant_damage": "no-damage", "top_k_cells": [{"cell_id": "G13", "row": 1, "col": 3, "building_count": 5, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 35365.4512, "severity_score": 0.066, "dominant_damage": "no-damage"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 28, "severe_count": 1, "severe_pct": 3.57, "severe_area_m2": 1891.5404, "total_area_m2": 60474.6138, "severity_score": 0.0357, "dominant_damage": "no-damage"}, {"cell_id": "G22", "row": 2, "col": 2, "building_count": 24, "severe_count": 1, "severe_pct": 4.17, "severe_area_m2": 4885.46, "total_area_m2": 62731.7618, "severity_score": 0.0279, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_mexico-earthquake_00000142_1496", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000142", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "mixed-severe", "question": "From the post-disaster earthquake scene (mexico-earthquake), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 2, "severe_footprint_m2": 6777.0004, "major_damage_area_m2": 4885.46, "destroyed_area_m2": 1891.5404, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_mexico-earthquake_00000142_1497", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000142", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster earthquake scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 2, "max_severe_in_cell": 2, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 0, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_mexico-earthquake_00000142_1498", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000142", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster earthquake scene (mexico-earthquake). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "mexico-earthquake", "disaster_type": "earthquake", "capture_date": "2017-09-20T17:46:11.000Z", "gsd": 2.6503215, "damage_summary": {"total_buildings": 177, "classified_buildings": 166, "counts": {"no-damage": 163, "minor-damage": 1, "major-damage": 1, "destroyed": 1, "unclassified": 11}, "severe_count": 2, "severe_pct": 1.2, "damaged_pct": 1.81}, "area_summary": {"total_footprint_m2": 1329353.3963, "severe_footprint_m2": 6777.0004, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 1044, "polygon_severe_count": 2, "pixel_histogram": {"0": 852846, "1": 193312, "2": 1374, "3": 755, "4": 289, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 98.7646, "2": 0.702, "3": 0.3857, "4": 0.1477, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_moore-tornado_00000003_1499", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "moore-tornado_00000003", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (moore-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 151, "minor-damage": 15, "major-damage": 10, "destroyed": 18, "unclassified": 18}, "total_buildings": 212, "classified_buildings": 194, "damaged_count": 43, "severe_count": 28, "damaged_pct": 22.16, "severe_pct": 14.43, "severe_footprint_m2": 51209.6375, "total_footprint_m2": 435274.9689, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_moore-tornado_00000003_1500", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "moore-tornado_00000003", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (moore-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 20, "severe_footprint_m2": 37534.0284, "severe_pct_in_quadrant": 60.61, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 33, "severe_count": 20, "severe_pct": 60.61, "severe_area_m2": 37534.0284, "total_area_m2": 71958.2554, "severity_score": 0.5461, "dominant_damage": "major-damage"}, {"cell_id": "Q0", "building_count": 9, "severe_count": 8, "severe_pct": 88.89, "severe_area_m2": 13675.6091, "total_area_m2": 15848.3719, "severity_score": 0.9256, "dominant_damage": "destroyed"}, {"cell_id": "Q2", "building_count": 78, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 163112.714, "severity_score": 0.0085, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 9, "severe_count": 8, "severe_pct": 88.89, "severe_area_m2": 13675.6091, "total_area_m2": 15848.3719, "severity_score": 0.9256, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 33, "severe_count": 20, "severe_pct": 60.61, "severe_area_m2": 37534.0284, "total_area_m2": 71958.2554, "severity_score": 0.5461, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 78, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 163112.714, "severity_score": 0.0085, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 92, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 184355.6276, "severity_score": 0.0287, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_moore-tornado_00000003_1501", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "moore-tornado_00000003", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G10", "severity_score": 0.9256, "building_count": 9, "severe_count": 8, "class_histogram": {"no_damage": 0, "minor_damage": 1, "major_damage": 0, "destroyed": 8, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G10", "row": 1, "col": 0, "building_count": 9, "severe_count": 8, "severe_pct": 88.89, "severe_area_m2": 13675.6091, "total_area_m2": 15848.3719, "severity_score": 0.9256, "dominant_damage": "destroyed"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 12, "severe_count": 6, "severe_pct": 50.0, "severe_area_m2": 12588.7972, "total_area_m2": 30799.1331, "severity_score": 0.555, "dominant_damage": "destroyed"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 21, "severe_count": 14, "severe_pct": 66.67, "severe_area_m2": 24945.2312, "total_area_m2": 41159.1223, "severity_score": 0.541, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_moore-tornado_00000003_1502", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "moore-tornado_00000003", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (moore-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 28, "severe_footprint_m2": 51209.6375, "major_damage_area_m2": 26566.3095, "destroyed_area_m2": 24643.328, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_moore-tornado_00000003_1503", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "moore-tornado_00000003", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 28, "max_severe_in_cell": 20, "concentration_score": 0.7143, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 8, "Q1": 20, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_moore-tornado_00000003_1504", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "moore-tornado_00000003", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (moore-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "moore-tornado", "disaster_type": "wind", "capture_date": "2013-05-22T17:26:30.085Z", "gsd": 1.72984850406647, "damage_summary": {"total_buildings": 212, "classified_buildings": 194, "counts": {"no-damage": 151, "minor-damage": 15, "major-damage": 10, "destroyed": 18, "unclassified": 18}, "severe_count": 28, "severe_pct": 14.43, "damaged_pct": 22.16}, "area_summary": {"total_footprint_m2": 435274.9689, "severe_footprint_m2": 51209.6375, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.7143, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 28}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_nepal-flooding_00000380_1505", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000380", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a flooding event (nepal-flooding) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 34, "minor-damage": 13, "major-damage": 2, "destroyed": 1, "unclassified": 4}, "total_buildings": 54, "classified_buildings": 50, "damaged_count": 16, "severe_count": 3, "damaged_pct": 32.0, "severe_pct": 6.0, "severe_footprint_m2": 6268.4598, "total_footprint_m2": 72942.0162, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_nepal-flooding_00000380_1506", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000380", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "For a post-disaster flooding scene (nepal-flooding), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 2, "severe_footprint_m2": 4141.4657, "severe_pct_in_quadrant": 3.85, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 52, "severe_count": 2, "severe_pct": 3.85, "severe_area_m2": 4141.4657, "total_area_m2": 70100.2215, "severity_score": 0.1083, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 2126.9941, "total_area_m2": 2841.7947, "severity_score": 0.5, "dominant_damage": "minor-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 52, "severe_count": 2, "severe_pct": 3.85, "severe_area_m2": 4141.4657, "total_area_m2": 70100.2215, "severity_score": 0.1083, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 2126.9941, "total_area_m2": 2841.7947, "severity_score": 0.5, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_nepal-flooding_00000380_1507", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000380", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G12", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G12", "row": 1, "col": 2, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1026.2338, "total_area_m2": 1026.2338, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 2126.9941, "total_area_m2": 2841.7947, "severity_score": 0.5, "dominant_damage": "minor-damage"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 23, "severe_count": 1, "severe_pct": 4.35, "severe_area_m2": 3115.2319, "total_area_m2": 27888.2868, "severity_score": 0.1009, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_nepal-flooding_00000380_1508", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000380", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "From the post-disaster flooding scene (nepal-flooding), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 3, "severe_footprint_m2": 6268.4598, "major_damage_area_m2": 5242.226, "destroyed_area_m2": 1026.2338, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_nepal-flooding_00000380_1509", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000380", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 3, "max_severe_in_cell": 2, "concentration_score": 0.6667, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 2, "Q2": 0, "Q3": 1}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_nepal-flooding_00000380_1510", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000380", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (nepal-flooding). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "nepal-flooding", "disaster_type": "flooding", "capture_date": "2017-08-24T05:16:36.085Z", "gsd": 2.3406696, "damage_summary": {"total_buildings": 54, "classified_buildings": 50, "counts": {"no-damage": 34, "minor-damage": 13, "major-damage": 2, "destroyed": 1, "unclassified": 4}, "severe_count": 3, "severe_pct": 6.0, "damaged_pct": 32.0}, "area_summary": {"total_footprint_m2": 72942.0162, "severe_footprint_m2": 6268.4598, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6667, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 3}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-harvey_00000224_1511", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000224", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a flooding event (hurricane-harvey) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 0, "minor-damage": 1, "major-damage": 10, "destroyed": 3, "unclassified": 2}, "total_buildings": 16, "classified_buildings": 14, "damaged_count": 14, "severe_count": 13, "damaged_pct": 100.0, "severe_pct": 92.86, "severe_footprint_m2": 66114.2844, "total_footprint_m2": 82190.8039, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-harvey_00000224_1512", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000224", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "For a post-disaster flooding scene (hurricane-harvey), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 7, "severe_footprint_m2": 26187.4067, "severe_pct_in_quadrant": 100.0, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 7, "severe_count": 7, "severe_pct": 100.0, "severe_area_m2": 26187.4067, "total_area_m2": 26187.4067, "severity_score": 0.7643, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 6, "severe_count": 5, "severe_pct": 83.33, "severe_area_m2": 23226.8846, "total_area_m2": 25598.3891, "severity_score": 0.6133, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 16699.9931, "total_area_m2": 30405.0081, "severity_score": 0.3333, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 7, "severe_count": 7, "severe_pct": 100.0, "severe_area_m2": 26187.4067, "total_area_m2": 26187.4067, "severity_score": 0.7643, "dominant_damage": "major-damage"}, {"cell_id": "Q1", "building_count": 6, "severe_count": 5, "severe_pct": 83.33, "severe_area_m2": 23226.8846, "total_area_m2": 25598.3891, "severity_score": 0.6133, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 3, "severe_count": 1, "severe_pct": 33.33, "severe_area_m2": 16699.9931, "total_area_m2": 30405.0081, "severity_score": 0.3333, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-harvey_00000224_1513", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000224", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 1.0, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 2, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 2775.68, "total_area_m2": 2775.68, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 4359.4066, "total_area_m2": 4359.4066, "severity_score": 0.78, "dominant_damage": "major-damage"}, {"cell_id": "G10", "row": 1, "col": 0, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 7725.9442, "total_area_m2": 7725.9442, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-harvey_00000224_1514", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000224", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "From the post-disaster flooding scene (hurricane-harvey), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 13, "severe_footprint_m2": 66114.2844, "major_damage_area_m2": 62597.2405, "destroyed_area_m2": 3517.0439, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-harvey_00000224_1515", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000224", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 13, "max_severe_in_cell": 7, "concentration_score": 0.5385, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 7, "Q1": 5, "Q2": 1, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-harvey_00000224_1516", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000224", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-harvey). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-harvey", "disaster_type": "flooding", "capture_date": "2017-08-31T17:38:50.685Z", "gsd": 3.024613, "damage_summary": {"total_buildings": 16, "classified_buildings": 14, "counts": {"no-damage": 0, "minor-damage": 1, "major-damage": 10, "destroyed": 3, "unclassified": 2}, "severe_count": 13, "severe_pct": 92.86, "damaged_pct": 100.0}, "area_summary": {"total_footprint_m2": 82190.8039, "severe_footprint_m2": 66114.2844, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.5385, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 7804, "polygon_severe_count": 13, "pixel_histogram": {"0": 1039628, "1": 0, "2": 1144, "3": 7364, "4": 440, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 0.0, "2": 12.785, "3": 82.2977, "4": 4.9173, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-harvey_00000512_1523", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000512", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a flooding event (hurricane-harvey) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 4, "minor-damage": 6, "major-damage": 1, "destroyed": 2, "unclassified": 0}, "total_buildings": 13, "classified_buildings": 13, "damaged_count": 9, "severe_count": 3, "damaged_pct": 69.23, "severe_pct": 23.08, "severe_footprint_m2": 33010.8365, "total_footprint_m2": 750511.3715, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-harvey_00000512_1524", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000512", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "For a post-disaster flooding scene (hurricane-harvey), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 3, "severe_footprint_m2": 33010.8365, "severe_pct_in_quadrant": 33.33, "dominant_damage": "minor-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 9, "severe_count": 3, "severe_pct": 33.33, "severe_area_m2": 33010.8365, "total_area_m2": 77106.4691, "severity_score": 0.5167, "dominant_damage": "minor-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q2", "building_count": 9, "severe_count": 3, "severe_pct": 33.33, "severe_area_m2": 33010.8365, "total_area_m2": 77106.4691, "severity_score": 0.5167, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 673404.9024, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-harvey_00000512_1525", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000512", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster flooding satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G31", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G31", "row": 3, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 7452.673, "total_area_m2": 7452.673, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G20", "row": 2, "col": 0, "building_count": 4, "severe_count": 2, "severe_pct": 50.0, "severe_area_m2": 25558.1635, "total_area_m2": 37128.1345, "severity_score": 0.5825, "dominant_damage": "minor-damage"}, {"cell_id": "G30", "row": 3, "col": 0, "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 32525.6616, "severity_score": 0.33, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-harvey_00000512_1526", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000512", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "From the post-disaster flooding scene (hurricane-harvey), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 3, "severe_footprint_m2": 33010.8365, "major_damage_area_m2": 19540.7819, "destroyed_area_m2": 13470.0546, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-harvey_00000512_1527", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000512", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster flooding scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 3, "max_severe_in_cell": 3, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 0, "Q2": 3, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-harvey_00000512_1528", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000512", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster flooding scene (hurricane-harvey). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-harvey", "disaster_type": "flooding", "capture_date": "2017-08-31T17:38:50.685Z", "gsd": 3.024613, "damage_summary": {"total_buildings": 13, "classified_buildings": 13, "counts": {"no-damage": 4, "minor-damage": 6, "major-damage": 1, "destroyed": 2, "unclassified": 0}, "severe_count": 3, "severe_pct": 23.08, "damaged_pct": 69.23}, "area_summary": {"total_footprint_m2": 750511.3715, "severe_footprint_m2": 33010.8365, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 3793, "polygon_severe_count": 3, "pixel_histogram": {"0": 965024, "1": 74604, "2": 5155, "3": 2218, "4": 1575, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 89.2905, "2": 6.1698, "3": 2.6546, "4": 1.8851, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_woolsey-fire_00000497_1529", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000497", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a fire event (woolsey-fire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 49, "minor-damage": 4, "major-damage": 4, "destroyed": 8, "unclassified": 4}, "total_buildings": 69, "classified_buildings": 65, "damaged_count": 16, "severe_count": 12, "damaged_pct": 24.62, "severe_pct": 18.46, "severe_footprint_m2": 31863.5148, "total_footprint_m2": 148927.0126, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_woolsey-fire_00000497_1530", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000497", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "For a post-disaster fire scene (woolsey-fire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q2", "severe_count": 5, "severe_footprint_m2": 16458.9535, "severe_pct_in_quadrant": 27.78, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q2", "building_count": 18, "severe_count": 5, "severe_pct": 27.78, "severe_area_m2": 16458.9535, "total_area_m2": 80738.1558, "severity_score": 0.2778, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 19, "severe_count": 1, "severe_pct": 5.26, "severe_area_m2": 7270.8017, "total_area_m2": 29294.3339, "severity_score": 0.07, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 15, "severe_count": 2, "severe_pct": 13.33, "severe_area_m2": 4441.1274, "total_area_m2": 18604.8123, "severity_score": 0.1113, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 19, "severe_count": 1, "severe_pct": 5.26, "severe_area_m2": 7270.8017, "total_area_m2": 29294.3339, "severity_score": 0.07, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 15, "severe_count": 2, "severe_pct": 13.33, "severe_area_m2": 4441.1274, "total_area_m2": 18604.8123, "severity_score": 0.1113, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 18, "severe_count": 5, "severe_pct": 27.78, "severe_area_m2": 16458.9535, "total_area_m2": 80738.1558, "severity_score": 0.2778, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 17, "severe_count": 4, "severe_pct": 23.53, "severe_area_m2": 3692.6322, "total_area_m2": 20289.7106, "severity_score": 0.2353, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_woolsey-fire_00000497_1531", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000497", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G31", "severity_score": 1.0, "building_count": 4, "severe_count": 4, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 4, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G31", "row": 3, "col": 1, "building_count": 4, "severe_count": 4, "severe_pct": 100.0, "severe_area_m2": 7891.1294, "total_area_m2": 7891.1294, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G22", "row": 2, "col": 2, "building_count": 7, "severe_count": 4, "severe_pct": 57.14, "severe_area_m2": 3692.6322, "total_area_m2": 7184.7477, "severity_score": 0.5714, "dominant_damage": "destroyed"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 7, "severe_count": 2, "severe_pct": 28.57, "severe_area_m2": 4441.1274, "total_area_m2": 13406.3007, "severity_score": 0.2386, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_woolsey-fire_00000497_1532", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000497", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "From the post-disaster fire scene (woolsey-fire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 12, "severe_footprint_m2": 31863.5148, "major_damage_area_m2": 18746.3577, "destroyed_area_m2": 13117.1571, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_woolsey-fire_00000497_1533", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000497", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 12, "max_severe_in_cell": 5, "concentration_score": 0.4167, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 1, "Q1": 2, "Q2": 5, "Q3": 4}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_woolsey-fire_00000497_1534", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000497", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (woolsey-fire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "woolsey-fire", "disaster_type": "fire", "capture_date": "2018-11-18T18:44:07.000Z", "gsd": 1.7726423000000002, "damage_summary": {"total_buildings": 69, "classified_buildings": 65, "counts": {"no-damage": 49, "minor-damage": 4, "major-damage": 4, "destroyed": 8, "unclassified": 4}, "severe_count": 12, "severe_pct": 18.46, "damaged_pct": 24.62}, "area_summary": {"total_footprint_m2": 148927.0126, "severe_footprint_m2": 31863.5148, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.4167, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 12}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_joplin-tornado_00000090_1535", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000090", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (joplin-tornado) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 14, "minor-damage": 15, "major-damage": 12, "destroyed": 18, "unclassified": 5}, "total_buildings": 64, "classified_buildings": 59, "damaged_count": 45, "severe_count": 30, "damaged_pct": 76.27, "severe_pct": 50.85, "severe_footprint_m2": 139535.2385, "total_footprint_m2": 283732.0627, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_joplin-tornado_00000090_1536", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000090", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (joplin-tornado), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 14, "severe_footprint_m2": 71077.7448, "severe_pct_in_quadrant": 63.64, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 22, "severe_count": 14, "severe_pct": 63.64, "severe_area_m2": 71077.7448, "total_area_m2": 107932.611, "severity_score": 0.5914, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 5, "severe_pct": 62.5, "severe_area_m2": 34691.4015, "total_area_m2": 64132.7674, "severity_score": 0.625, "dominant_damage": "destroyed"}, {"cell_id": "Q0", "building_count": 15, "severe_count": 10, "severe_pct": 66.67, "severe_area_m2": 32928.5671, "total_area_m2": 61248.4665, "severity_score": 0.6667, "dominant_damage": "destroyed"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 15, "severe_count": 10, "severe_pct": 66.67, "severe_area_m2": 32928.5671, "total_area_m2": 61248.4665, "severity_score": 0.6667, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 19, "severe_count": 1, "severe_pct": 5.26, "severe_area_m2": 837.5251, "total_area_m2": 50418.2178, "severity_score": 0.1568, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 8, "severe_count": 5, "severe_pct": 62.5, "severe_area_m2": 34691.4015, "total_area_m2": 64132.7674, "severity_score": 0.625, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 22, "severe_count": 14, "severe_pct": 63.64, "severe_area_m2": 71077.7448, "total_area_m2": 107932.611, "severity_score": 0.5914, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_joplin-tornado_00000090_1537", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000090", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G21", "severity_score": 1.0, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 2, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G21", "row": 2, "col": 1, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 5193.7613, "total_area_m2": 5193.7613, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 3, "severe_count": 3, "severe_pct": 100.0, "severe_area_m2": 17674.3385, "total_area_m2": 17674.3385, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 7, "severe_count": 7, "severe_pct": 100.0, "severe_area_m2": 29235.5953, "total_area_m2": 29235.5953, "severity_score": 0.9057, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_joplin-tornado_00000090_1538", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000090", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (joplin-tornado), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 30, "severe_footprint_m2": 139535.2385, "major_damage_area_m2": 80068.6079, "destroyed_area_m2": 59466.6306, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_joplin-tornado_00000090_1539", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000090", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 30, "max_severe_in_cell": 14, "concentration_score": 0.4667, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 10, "Q1": 1, "Q2": 5, "Q3": 14}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_joplin-tornado_00000090_1540", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000090", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (joplin-tornado). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "joplin-tornado", "disaster_type": "wind", "capture_date": "2011-05-29T17:29:25.433Z", "gsd": 2.35255861282349, "damage_summary": {"total_buildings": 64, "classified_buildings": 59, "counts": {"no-damage": 14, "minor-damage": 15, "major-damage": 12, "destroyed": 18, "unclassified": 5}, "severe_count": 30, "severe_pct": 50.85, "damaged_pct": 76.27}, "area_summary": {"total_footprint_m2": 283732.0627, "severe_footprint_m2": 139535.2385, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.4667, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 30}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_santa-rosa-wildfire_00000005_1541", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000005", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a fire event (santa-rosa-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 17, "minor-damage": 0, "major-damage": 1, "destroyed": 2, "unclassified": 1}, "total_buildings": 21, "classified_buildings": 20, "damaged_count": 3, "severe_count": 3, "damaged_pct": 15.0, "severe_pct": 15.0, "severe_footprint_m2": 233039.2041, "total_footprint_m2": 492846.1598, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_santa-rosa-wildfire_00000005_1542", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000005", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "For a post-disaster fire scene (santa-rosa-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 3, "severe_footprint_m2": 233039.2041, "severe_pct_in_quadrant": 42.86, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 7, "severe_count": 3, "severe_pct": 42.86, "severe_area_m2": 233039.2041, "total_area_m2": 266174.9512, "severity_score": 0.3814, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 27005.7358, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 195621.7806, "severity_score": 0.0, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 7, "severe_count": 3, "severe_pct": 42.86, "severe_area_m2": 233039.2041, "total_area_m2": 266174.9512, "severity_score": 0.3814, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 27005.7358, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 11, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 195621.7806, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 2, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 4043.6922, "severity_score": 0.0, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_santa-rosa-wildfire_00000005_1543", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000005", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G01", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G01", "row": 0, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 4535.9819, "total_area_m2": 4535.9819, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 213169.4065, "total_area_m2": 213169.4065, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G00", "row": 0, "col": 0, "building_count": 4, "severe_count": 1, "severe_pct": 25.0, "severe_area_m2": 15333.8157, "total_area_m2": 47704.7796, "severity_score": 0.1675, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_santa-rosa-wildfire_00000005_1544", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000005", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "From the post-disaster fire scene (santa-rosa-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 3, "severe_footprint_m2": 233039.2041, "major_damage_area_m2": 15333.8157, "destroyed_area_m2": 217705.3884, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_santa-rosa-wildfire_00000005_1545", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000005", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 3, "max_severe_in_cell": 3, "concentration_score": 1.0, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 3, "Q1": 0, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_santa-rosa-wildfire_00000005_1546", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000005", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (santa-rosa-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "santa-rosa-wildfire", "disaster_type": "fire", "capture_date": "2017-10-11T19:19:41.000Z", "gsd": 1.8769937, "damage_summary": {"total_buildings": 21, "classified_buildings": 20, "counts": {"no-damage": 17, "minor-damage": 0, "major-damage": 1, "destroyed": 2, "unclassified": 1}, "severe_count": 3, "severe_pct": 15.0, "damaged_pct": 15.0}, "area_summary": {"total_footprint_m2": 492846.1598, "severe_footprint_m2": 233039.2041, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 1.0, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 66829, "polygon_severe_count": 3, "pixel_histogram": {"0": 906136, "1": 75611, "2": 0, "3": 4517, "4": 62312, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 53.0827, "2": 0.0, "3": 3.1712, "4": 43.7461, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000077_1547", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000077", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 65, "minor-damage": 30, "major-damage": 1, "destroyed": 3, "unclassified": 1}, "total_buildings": 100, "classified_buildings": 99, "damaged_count": 34, "severe_count": 4, "damaged_pct": 34.34, "severe_pct": 4.04, "severe_footprint_m2": 12763.6158, "total_footprint_m2": 389783.256, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000077_1548", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000077", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 3, "severe_footprint_m2": 11824.1586, "severe_pct_in_quadrant": 15.0, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 20, "severe_count": 3, "severe_pct": 15.0, "severe_area_m2": 11824.1586, "total_area_m2": 71836.3519, "severity_score": 0.249, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 9, "severe_count": 1, "severe_pct": 11.11, "severe_area_m2": 939.4572, "total_area_m2": 33560.9654, "severity_score": 0.1111, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 36, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 110307.3013, "severity_score": 0.0733, "dominant_damage": "no-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 9, "severe_count": 1, "severe_pct": 11.11, "severe_area_m2": 939.4572, "total_area_m2": 33560.9654, "severity_score": 0.1111, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 20, "severe_count": 3, "severe_pct": 15.0, "severe_area_m2": 11824.1586, "total_area_m2": 71836.3519, "severity_score": 0.249, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 36, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 110307.3013, "severity_score": 0.0733, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 35, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 174078.6374, "severity_score": 0.1414, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000077_1549", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000077", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 1.0, "building_count": 2, "severe_count": 2, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 2, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 1758.625, "total_area_m2": 1758.625, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G11", "row": 1, "col": 1, "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 939.4572, "total_area_m2": 1566.702, "severity_score": 0.5, "dominant_damage": "no-damage"}, {"cell_id": "G21", "row": 2, "col": 1, "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 3446.1018, "severity_score": 0.33, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-michael_00000077_1550", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000077", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (hurricane-michael), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 4, "severe_footprint_m2": 12763.6158, "major_damage_area_m2": 10065.5336, "destroyed_area_m2": 2698.0822, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-michael_00000077_1551", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000077", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 4, "max_severe_in_cell": 3, "concentration_score": 0.75, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 1, "Q1": 3, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000077_1552", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000077", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 100, "classified_buildings": 99, "counts": {"no-damage": 65, "minor-damage": 30, "major-damage": 1, "destroyed": 3, "unclassified": 1}, "severe_count": 4, "severe_pct": 4.04, "damaged_pct": 34.34}, "area_summary": {"total_footprint_m2": 389783.256, "severe_footprint_m2": 12763.6158, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.75, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 3210, "polygon_severe_count": 4, "pixel_histogram": {"0": 951752, "1": 59349, "2": 34265, "3": 2483, "4": 727, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 61.2958, "2": 35.389, "3": 2.5644, "4": 0.7508, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_portugal-wildfire_00001572_1553", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001572", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a fire event (portugal-wildfire) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 23, "minor-damage": 4, "major-damage": 10, "destroyed": 17, "unclassified": 1}, "total_buildings": 55, "classified_buildings": 54, "damaged_count": 31, "severe_count": 27, "damaged_pct": 57.41, "severe_pct": 50.0, "severe_footprint_m2": 61858.7255, "total_footprint_m2": 124527.3457, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_portugal-wildfire_00001572_1554", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001572", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "For a post-disaster fire scene (portugal-wildfire), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q3", "severe_count": 17, "severe_footprint_m2": 45222.4529, "severe_pct_in_quadrant": 54.84, "dominant_damage": "destroyed", "top_k_quadrants": [{"cell_id": "Q3", "building_count": 31, "severe_count": 17, "severe_pct": 54.84, "severe_area_m2": 45222.4529, "total_area_m2": 75882.8933, "severity_score": 0.5165, "dominant_damage": "destroyed"}, {"cell_id": "Q1", "building_count": 19, "severe_count": 9, "severe_pct": 47.37, "severe_area_m2": 15079.827, "total_area_m2": 39712.1648, "severity_score": 0.4216, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1556.4456, "total_area_m2": 1556.4456, "severity_score": 1.0, "dominant_damage": "destroyed"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 7375.842, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 19, "severe_count": 9, "severe_pct": 47.37, "severe_area_m2": 15079.827, "total_area_m2": 39712.1648, "severity_score": 0.4216, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1556.4456, "total_area_m2": 1556.4456, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "Q3", "building_count": 31, "severe_count": 17, "severe_pct": 54.84, "severe_area_m2": 45222.4529, "total_area_m2": 75882.8933, "severity_score": 0.5165, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_portugal-wildfire_00001572_1555", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001572", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster fire satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G21", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G21", "row": 2, "col": 1, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 1556.4456, "total_area_m2": 1556.4456, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G02", "row": 0, "col": 2, "building_count": 5, "severe_count": 4, "severe_pct": 80.0, "severe_area_m2": 8103.1991, "total_area_m2": 9506.8045, "severity_score": 0.734, "dominant_damage": "destroyed"}, {"cell_id": "G33", "row": 3, "col": 3, "building_count": 13, "severe_count": 10, "severe_pct": 76.92, "severe_area_m2": 24623.0034, "total_area_m2": 29885.9273, "severity_score": 0.6931, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_portugal-wildfire_00001572_1556", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001572", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "From the post-disaster fire scene (portugal-wildfire), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 27, "severe_footprint_m2": 61858.7255, "major_damage_area_m2": 27342.4749, "destroyed_area_m2": 34516.2506, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_portugal-wildfire_00001572_1557", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001572", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster fire scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 27, "max_severe_in_cell": 17, "concentration_score": 0.6296, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 9, "Q2": 1, "Q3": 17}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_portugal-wildfire_00001572_1558", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001572", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster fire scene (portugal-wildfire). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "portugal-wildfire", "disaster_type": "fire", "capture_date": "2017-06-21T11:52:35.000Z", "gsd": 2.2328367, "damage_summary": {"total_buildings": 55, "classified_buildings": 54, "counts": {"no-damage": 23, "minor-damage": 4, "major-damage": 10, "destroyed": 17, "unclassified": 1}, "severe_count": 27, "severe_pct": 50.0, "damaged_pct": 57.41}, "area_summary": {"total_footprint_m2": 124527.3457, "severe_footprint_m2": 61858.7255, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6296, "worst_quadrant": "Q3"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 27}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-matthew_00000293_1559", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000293", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (hurricane-matthew) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 12, "minor-damage": 21, "major-damage": 7, "destroyed": 11, "unclassified": 13}, "total_buildings": 64, "classified_buildings": 51, "damaged_count": 39, "severe_count": 18, "damaged_pct": 76.47, "severe_pct": 35.29, "severe_footprint_m2": 12745.4035, "total_footprint_m2": 57022.9623, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-matthew_00000293_1560", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000293", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (hurricane-matthew), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q0", "severe_count": 7, "severe_footprint_m2": 4885.4046, "severe_pct_in_quadrant": 53.85, "dominant_damage": "minor-damage", "top_k_quadrants": [{"cell_id": "Q0", "building_count": 13, "severe_count": 7, "severe_pct": 53.85, "severe_area_m2": 4885.4046, "total_area_m2": 11464.9348, "severity_score": 0.5638, "dominant_damage": "minor-damage"}, {"cell_id": "Q2", "building_count": 30, "severe_count": 5, "severe_pct": 16.67, "severe_area_m2": 4553.4621, "total_area_m2": 22193.8545, "severity_score": 0.2217, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 20, "severe_count": 6, "severe_pct": 30.0, "severe_area_m2": 3306.5368, "total_area_m2": 20914.0969, "severity_score": 0.432, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 13, "severe_count": 7, "severe_pct": 53.85, "severe_area_m2": 4885.4046, "total_area_m2": 11464.9348, "severity_score": 0.5638, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 1, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 2450.0761, "severity_score": 0.0, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 30, "severe_count": 5, "severe_pct": 16.67, "severe_area_m2": 4553.4621, "total_area_m2": 22193.8545, "severity_score": 0.2217, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 20, "severe_count": 6, "severe_pct": 30.0, "severe_area_m2": 3306.5368, "total_area_m2": 20914.0969, "severity_score": 0.432, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-matthew_00000293_1561", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000293", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G11", "severity_score": 0.668, "building_count": 5, "severe_count": 4, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 2, "destroyed": 2, "unclassified": 1}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G11", "row": 1, "col": 1, "building_count": 5, "severe_count": 4, "severe_pct": 80.0, "severe_area_m2": 3045.0699, "total_area_m2": 3614.7475, "severity_score": 0.668, "dominant_damage": "major-damage"}, {"cell_id": "G01", "row": 0, "col": 1, "building_count": 3, "severe_count": 2, "severe_pct": 66.67, "severe_area_m2": 1474.3645, "total_area_m2": 2602.4803, "severity_score": 0.6667, "dominant_damage": "minor-damage"}, {"cell_id": "G23", "row": 2, "col": 3, "building_count": 2, "severe_count": 1, "severe_pct": 50.0, "severe_area_m2": 1389.4801, "total_area_m2": 7641.8602, "severity_score": 0.665, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-matthew_00000293_1562", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000293", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (hurricane-matthew), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 18, "severe_footprint_m2": 12745.4035, "major_damage_area_m2": 6075.8272, "destroyed_area_m2": 6669.5763, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-matthew_00000293_1563", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000293", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 18, "max_severe_in_cell": 7, "concentration_score": 0.3889, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 7, "Q1": 0, "Q2": 5, "Q3": 6}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-matthew_00000293_1564", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000293", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-matthew). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-matthew", "disaster_type": "wind", "capture_date": "2016-10-08T15:33:12.799Z", "gsd": 1.90909719467163, "damage_summary": {"total_buildings": 64, "classified_buildings": 51, "counts": {"no-damage": 12, "minor-damage": 21, "major-damage": 7, "destroyed": 11, "unclassified": 13}, "severe_count": 18, "severe_pct": 35.29, "damaged_pct": 76.47}, "area_summary": {"total_footprint_m2": 57022.9623, "severe_footprint_m2": 12745.4035, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.3889, "worst_quadrant": "Q0"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 3968, "polygon_severe_count": 18, "pixel_histogram": {"0": 1033195, "1": 3061, "2": 8352, "3": 1871, "4": 2097, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 19.9012, "2": 54.3008, "3": 12.1644, "4": 13.6337, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000269_1565", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000269", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 13, "minor-damage": 6, "major-damage": 4, "destroyed": 2, "unclassified": 0}, "total_buildings": 25, "classified_buildings": 25, "damaged_count": 12, "severe_count": 6, "damaged_pct": 48.0, "severe_pct": 24.0, "severe_footprint_m2": 43802.4323, "total_footprint_m2": 188203.3724, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "moderately concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000269_1566", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000269", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 4, "severe_footprint_m2": 34408.5127, "severe_pct_in_quadrant": 57.14, "dominant_damage": "no-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 7, "severe_count": 4, "severe_pct": 57.14, "severe_area_m2": 34408.5127, "total_area_m2": 123575.2274, "severity_score": 0.5243, "dominant_damage": "no-damage"}, {"cell_id": "Q3", "building_count": 18, "severe_count": 2, "severe_pct": 11.11, "severe_area_m2": 9393.9196, "total_area_m2": 64628.145, "severity_score": 0.1661, "dominant_damage": "no-damage"}, {"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q1", "building_count": 7, "severe_count": 4, "severe_pct": 57.14, "severe_area_m2": 34408.5127, "total_area_m2": 123575.2274, "severity_score": 0.5243, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 0, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 0.0, "severity_score": 0.0, "dominant_damage": "none"}, {"cell_id": "Q3", "building_count": 18, "severe_count": 2, "severe_pct": 11.11, "severe_area_m2": 9393.9196, "total_area_m2": 64628.145, "severity_score": 0.1661, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000269_1567", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000269", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 0.67, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 1, "destroyed": 0, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 7680.5451, "total_area_m2": 7680.5451, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 6634.7423, "total_area_m2": 6634.7423, "severity_score": 0.67, "dominant_damage": "major-damage"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 4, "severe_count": 2, "severe_pct": 50.0, "severe_area_m2": 20093.2253, "total_area_m2": 60289.7889, "severity_score": 0.5825, "dominant_damage": "destroyed"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-michael_00000269_1568", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000269", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (hurricane-michael), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 6, "severe_footprint_m2": 43802.4323, "major_damage_area_m2": 23709.207, "destroyed_area_m2": 20093.2253, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-michael_00000269_1569", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000269", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 6, "max_severe_in_cell": 4, "concentration_score": 0.6667, "assessment": "moderately concentrated", "per_quadrant_severe": {"Q0": 0, "Q1": 4, "Q2": 0, "Q3": 2}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000269_1570", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000269", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 25, "classified_buildings": 25, "counts": {"no-damage": 13, "minor-damage": 6, "major-damage": 4, "destroyed": 2, "unclassified": 0}, "severe_count": 6, "severe_pct": 24.0, "damaged_pct": 48.0}, "area_summary": {"total_footprint_m2": 188203.3724, "severe_footprint_m2": 43802.4323, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "moderately concentrated", "concentration_score": 0.6667, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 10771, "polygon_severe_count": 6, "pixel_histogram": {"0": 1002797, "1": 28046, "2": 6962, "3": 5878, "4": 4893, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 61.2639, "2": 15.2078, "3": 12.8399, "4": 10.6883, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_hurricane-michael_00000247_1583", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000247", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a wind event (hurricane-michael) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 52, "minor-damage": 60, "major-damage": 45, "destroyed": 20, "unclassified": 1}, "total_buildings": 178, "classified_buildings": 177, "damaged_count": 125, "severe_count": 65, "damaged_pct": 70.62, "severe_pct": 36.72, "severe_footprint_m2": 169884.7287, "total_footprint_m2": 435412.078, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "spatially dispersed"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_hurricane-michael_00000247_1584", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000247", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "For a post-disaster wind scene (hurricane-michael), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 18, "severe_footprint_m2": 68108.1586, "severe_pct_in_quadrant": 54.55, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 33, "severe_count": 18, "severe_pct": 54.55, "severe_area_m2": 68108.1586, "total_area_m2": 106204.1719, "severity_score": 0.5055, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 47, "severe_count": 20, "severe_pct": 42.55, "severe_area_m2": 43614.4247, "total_area_m2": 109657.8424, "severity_score": 0.4466, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 36, "severe_count": 16, "severe_pct": 44.44, "severe_area_m2": 30635.9633, "total_area_m2": 77878.3247, "severity_score": 0.3986, "dominant_damage": "major-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 62, "severe_count": 11, "severe_pct": 17.74, "severe_area_m2": 27526.1821, "total_area_m2": 141671.739, "severity_score": 0.2892, "dominant_damage": "minor-damage"}, {"cell_id": "Q1", "building_count": 33, "severe_count": 18, "severe_pct": 54.55, "severe_area_m2": 68108.1586, "total_area_m2": 106204.1719, "severity_score": 0.5055, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 47, "severe_count": 20, "severe_pct": 42.55, "severe_area_m2": 43614.4247, "total_area_m2": 109657.8424, "severity_score": 0.4466, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 36, "severe_count": 16, "severe_pct": 44.44, "severe_area_m2": 30635.9633, "total_area_m2": 77878.3247, "severity_score": 0.3986, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_hurricane-michael_00000247_1585", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000247", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster wind satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 0.7291, "building_count": 11, "severe_count": 10, "class_histogram": {"no_damage": 1, "minor_damage": 0, "major_damage": 6, "destroyed": 4, "unclassified": 0}, "dominant_damage": "major-damage", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 11, "severe_count": 10, "severe_pct": 90.91, "severe_area_m2": 47790.0233, "total_area_m2": 48483.3781, "severity_score": 0.7291, "dominant_damage": "major-damage"}, {"cell_id": "G30", "row": 3, "col": 0, "building_count": 7, "severe_count": 5, "severe_pct": 71.43, "severe_area_m2": 11464.2333, "total_area_m2": 13637.3469, "severity_score": 0.62, "dominant_damage": "major-damage"}, {"cell_id": "G12", "row": 1, "col": 2, "building_count": 18, "severe_count": 8, "severe_pct": 44.44, "severe_area_m2": 20318.1353, "total_area_m2": 44831.7462, "severity_score": 0.4811, "dominant_damage": "minor-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_hurricane-michael_00000247_1586", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000247", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "From the post-disaster wind scene (hurricane-michael), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 65, "severe_footprint_m2": 169884.7287, "major_damage_area_m2": 119089.8996, "destroyed_area_m2": 50794.8291, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_hurricane-michael_00000247_1587", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000247", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster wind scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 65, "max_severe_in_cell": 20, "concentration_score": 0.3077, "assessment": "spatially dispersed", "per_quadrant_severe": {"Q0": 11, "Q1": 18, "Q2": 20, "Q3": 16}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_hurricane-michael_00000247_1588", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000247", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster wind scene (hurricane-michael). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "hurricane-michael", "disaster_type": "wind", "capture_date": "2018-10-13T16:48:15.000Z", "gsd": 2.0711906, "damage_summary": {"total_buildings": 178, "classified_buildings": 177, "counts": {"no-damage": 52, "minor-damage": 60, "major-damage": 45, "destroyed": 20, "unclassified": 1}, "severe_count": 65, "severe_pct": 36.72, "damaged_pct": 70.62}, "area_summary": {"total_footprint_m2": 435412.078, "severe_footprint_m2": 169884.7287, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "spatially dispersed", "concentration_score": 0.3077, "worst_quadrant": "Q2"}, "mask_validation": {"mask_available": true, "mask_severe_pixels": 42896, "polygon_severe_count": 65, "pixel_histogram": {"0": 938688, "1": 32420, "2": 34572, "3": 30149, "4": 12747, "5": 0}, "pixel_pct_non_background": {"0": 0.0, "1": 29.5028, "2": 31.4611, "3": 27.4361, "4": 11.6, "5": 0.0}}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q1_lower-puna-volcano_00000068_1595", "template": "XBD-Q1", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000068", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "mixed-severe", "question": "A post-disaster satellite scene from a volcano event (lower-puna-volcano) has been captured. Generate a complete damage inventory report including: the count of buildings per damage class, the percentage of damaged and severely damaged buildings over all classified buildings, and the total severe-damage footprint area in square metres.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "ComputeDamagePercentages", "ExportDamageReport", "Terminate"], "gt_answer": {"damage_distribution": {"no-damage": 18, "minor-damage": 4, "major-damage": 2, "destroyed": 3, "unclassified": 2}, "total_buildings": 29, "classified_buildings": 27, "damaged_count": 9, "severe_count": 5, "damaged_pct": 33.33, "severe_pct": 18.52, "severe_footprint_m2": 36084.6247, "total_footprint_m2": 123104.2482, "area_method": "xy_polygon_area_times_gsd_squared", "spatial_dispersion": "highly concentrated"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "percentage_calculation"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q2_lower-puna-volcano_00000068_1596", "template": "XBD-Q2", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000068", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "mixed-severe", "question": "For a post-disaster volcano scene (lower-puna-volcano), the 1024\u00d71024 image has been divided into four quadrants (Q0\u2013Q3). Which quadrant contains the highest severe-damage footprint area? Report the quadrant ID, the count of severely damaged buildings, and the footprint area in m\u00b2.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantDamageStats", "RankQuadrantsBySevereDamageArea", "Terminate"], "gt_answer": {"highest_severe_area_quadrant": "Q1", "severe_count": 4, "severe_footprint_m2": 28443.5029, "severe_pct_in_quadrant": 66.67, "dominant_damage": "major-damage", "top_k_quadrants": [{"cell_id": "Q1", "building_count": 6, "severe_count": 4, "severe_pct": 66.67, "severe_area_m2": 28443.5029, "total_area_m2": 43699.6944, "severity_score": 0.6117, "dominant_damage": "major-damage"}, {"cell_id": "Q0", "building_count": 10, "severe_count": 1, "severe_pct": 10.0, "severe_area_m2": 7641.1218, "total_area_m2": 32998.0702, "severity_score": 0.1, "dominant_damage": "no-damage"}, {"cell_id": "Q2", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 12569.632, "severity_score": 0.165, "dominant_damage": "minor-damage"}], "all_quadrants": [{"cell_id": "Q0", "building_count": 10, "severe_count": 1, "severe_pct": 10.0, "severe_area_m2": 7641.1218, "total_area_m2": 32998.0702, "severity_score": 0.1, "dominant_damage": "no-damage"}, {"cell_id": "Q1", "building_count": 6, "severe_count": 4, "severe_pct": 66.67, "severe_area_m2": 28443.5029, "total_area_m2": 43699.6944, "severity_score": 0.6117, "dominant_damage": "major-damage"}, {"cell_id": "Q2", "building_count": 4, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 12569.632, "severity_score": 0.165, "dominant_damage": "minor-damage"}, {"cell_id": "Q3", "building_count": 9, "severe_count": 0, "severe_pct": 0.0, "severe_area_m2": 0.0, "total_area_m2": 33836.8516, "severity_score": 0.0367, "dominant_damage": "no-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "area_computation", "ranking"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q3_lower-puna-volcano_00000068_1597", "template": "XBD-Q3", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000068", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "mixed-severe", "question": "A 4\u00d74 grid has been overlaid on a post-disaster volcano satellite scene, creating 16 cells (G00\u2013G33). Each cell is scored by normalised damage severity (0=no damage, 1=fully destroyed). Which cell has the highest severity score? Report the cell ID, score, and its building class histogram.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadGrid4x4", "ComputeCellSeverityScores", "RankCellsByNormalizedSeverity", "Terminate"], "gt_answer": {"highest_severity_cell": "G02", "severity_score": 1.0, "building_count": 1, "severe_count": 1, "class_histogram": {"no_damage": 0, "minor_damage": 0, "major_damage": 0, "destroyed": 1, "unclassified": 0}, "dominant_damage": "destroyed", "top_k_cells": [{"cell_id": "G02", "row": 0, "col": 2, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 7324.6905, "total_area_m2": 7324.6905, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G13", "row": 1, "col": 3, "building_count": 1, "severe_count": 1, "severe_pct": 100.0, "severe_area_m2": 7856.5871, "total_area_m2": 7856.5871, "severity_score": 1.0, "dominant_damage": "destroyed"}, {"cell_id": "G03", "row": 0, "col": 3, "building_count": 2, "severe_count": 2, "severe_pct": 100.0, "severe_area_m2": 13262.2253, "total_area_m2": 13262.2253, "severity_score": 0.67, "dominant_damage": "major-damage"}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "severity_scoring", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q5_lower-puna-volcano_00000068_1598", "template": "XBD-Q5", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000068", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "mixed-severe", "question": "From the post-disaster volcano scene (lower-puna-volcano), identify all buildings with severe damage (major-damage or destroyed). Report the total severe building count and their combined footprint area in m\u00b2, computed from xy polygon coordinates and the scene GSD.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "FilterSevereDamagePolygons", "ComputeSevereDamageAreaFromXYGSD", "ExportSeverePolygonGeoJSON", "Terminate"], "gt_answer": {"filter": "subtype in [major-damage, destroyed]", "polygon_source": "building_damage_xy.geojson", "severe_count": 5, "severe_footprint_m2": 36084.6247, "major_damage_area_m2": 13262.2253, "destroyed_area_m2": 22822.3994, "area_method": "xy_polygon_area_times_gsd_squared"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["filtering", "area_computation"], "difficulty": "easy", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q7_lower-puna-volcano_00000068_1599", "template": "XBD-Q7", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000068", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "mixed-severe", "question": "Analyze the spatial distribution of severe building damage in this post-disaster volcano scene using a 2\u00d72 quadrant grid. Is the damage highly concentrated in one quadrant or spatially dispersed across the scene? Report the concentration score (0\u20131) and classification.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadQuadrantGrid", "ComputeQuadrantSevereCounts", "ComputeSpatialConcentrationScore", "ClassifySpatialDispersion", "Terminate"], "gt_answer": {"total_severe": 5, "max_severe_in_cell": 4, "concentration_score": 0.8, "assessment": "highly concentrated", "per_quadrant_severe": {"Q0": 1, "Q1": 4, "Q2": 0, "Q3": 0}, "grid_used": "quadrant_2x2"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["spatial_localization", "concentration_scoring", "classification"], "difficulty": "medium", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q8_lower-puna-volcano_00000068_1600", "template": "XBD-Q8", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000068", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "mixed-severe", "question": "Create a comprehensive scene-level damage summary for this post-disaster volcano scene (lower-puna-volcano). Use the building damage labels, target mask class distribution, severe-damage footprint area, and spatial dispersion analysis. Include damage counts, percentages, spatial assessment, and target mask validation.", "input_layers": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadPrePostMetadata", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "LoadTargetMaskHistogram", "ValidateMaskVsPolygonCounts", "ExportSceneChangeSummary", "Terminate"], "gt_answer": {"disaster": "lower-puna-volcano", "disaster_type": "volcano", "capture_date": "2018-05-23T20:59:21.000Z", "gsd": 2.2423599, "damage_summary": {"total_buildings": 29, "classified_buildings": 27, "counts": {"no-damage": 18, "minor-damage": 4, "major-damage": 2, "destroyed": 3, "unclassified": 2}, "severe_count": 5, "severe_pct": 18.52, "damaged_pct": 33.33}, "area_summary": {"total_footprint_m2": 123104.2482, "severe_footprint_m2": 36084.6247, "area_method": "xy_polygon_area_times_gsd_squared"}, "spatial_analysis": {"dispersion_assessment": "highly concentrated", "concentration_score": 0.8, "worst_quadrant": "Q1"}, "mask_validation": {"mask_available": false, "polygon_severe_count": 5}}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "mask_validation", "spatial_localization", "multi_source_synthesis"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}
+{"query_id": "XBD_Q4_guatemala-volcano_00000025_1601", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "guatemala-volcano_00000025", "split": "train", "qa_split": "train", "disaster": "guatemala-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "Two satellite scenes captured the same volcano event (guatemala-volcano). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "guatemala-volcano_00000025", "total_buildings": 201, "severe_count": 27, "severe_pct": 100.0, "severe_footprint_m2": 15332.6583}, "scene_b": {"scene_id": "guatemala-volcano_00000020", "total_buildings": 36, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "higher_severe_pct_scene": "guatemala-volcano_00000025", "higher_severe_area_scene": "guatemala-volcano_00000025"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "guatemala-volcano_00000020"}
+{"query_id": "XBD_Q4_hurricane-florence_00000297_1604", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000297", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "Two satellite scenes captured the same flooding event (hurricane-florence). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "hurricane-florence_00000297", "total_buildings": 15, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "hurricane-florence_00000409", "total_buildings": 31, "severe_count": 20, "severe_pct": 64.52, "severe_footprint_m2": 163973.0118}, "higher_severe_pct_scene": "hurricane-florence_00000409", "higher_severe_area_scene": "hurricane-florence_00000409"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "hurricane-florence_00000409"}
+{"query_id": "XBD_Q4_hurricane-florence_00000181_1605", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "hurricane-florence_00000181", "split": "train", "qa_split": "train", "disaster": "hurricane-florence", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Two satellite scenes captured the same flooding event (hurricane-florence). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "hurricane-florence_00000181", "total_buildings": 32, "severe_count": 11, "severe_pct": 35.48, "severe_footprint_m2": 35991.7055}, "scene_b": {"scene_id": "hurricane-florence_00000389", "total_buildings": 44, "severe_count": 27, "severe_pct": 100.0, "severe_footprint_m2": 69751.2643}, "higher_severe_pct_scene": "hurricane-florence_00000389", "higher_severe_area_scene": "hurricane-florence_00000389"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "hurricane-florence_00000389"}
+{"query_id": "XBD_Q4_hurricane-harvey_00000081_1606", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000081", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Two satellite scenes captured the same flooding event (hurricane-harvey). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "hurricane-harvey_00000081", "total_buildings": 33, "severe_count": 23, "severe_pct": 85.19, "severe_footprint_m2": 241740.9646}, "scene_b": {"scene_id": "hurricane-harvey_00000188", "total_buildings": 37, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "higher_severe_pct_scene": "hurricane-harvey_00000081", "higher_severe_area_scene": "hurricane-harvey_00000081"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "hurricane-harvey_00000188"}
+{"query_id": "XBD_Q4_hurricane-harvey_00000269_1608", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "hurricane-harvey_00000269", "split": "train", "qa_split": "train", "disaster": "hurricane-harvey", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Two satellite scenes captured the same flooding event (hurricane-harvey). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "hurricane-harvey_00000269", "total_buildings": 15, "severe_count": 15, "severe_pct": 100.0, "severe_footprint_m2": 59808.9531}, "scene_b": {"scene_id": "hurricane-harvey_00000491", "total_buildings": 177, "severe_count": 115, "severe_pct": 65.71, "severe_footprint_m2": 1622768.118}, "higher_severe_pct_scene": "hurricane-harvey_00000269", "higher_severe_area_scene": "hurricane-harvey_00000491"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "hurricane-harvey_00000491"}
+{"query_id": "XBD_Q4_hurricane-matthew_00000034_1610", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000034", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Two satellite scenes captured the same wind event (hurricane-matthew). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "hurricane-matthew_00000034", "total_buildings": 19, "severe_count": 4, "severe_pct": 25.0, "severe_footprint_m2": 13668.6899}, "scene_b": {"scene_id": "hurricane-matthew_00000063", "total_buildings": 14, "severe_count": 5, "severe_pct": 35.71, "severe_footprint_m2": 17692.1073}, "higher_severe_pct_scene": "hurricane-matthew_00000063", "higher_severe_area_scene": "hurricane-matthew_00000063"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "hurricane-matthew_00000063"}
+{"query_id": "XBD_Q4_hurricane-matthew_00000002_1611", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "hurricane-matthew_00000002", "split": "train", "qa_split": "train", "disaster": "hurricane-matthew", "disaster_type": "wind", "damage_bucket": "major-heavy", "question": "Two satellite scenes captured the same wind event (hurricane-matthew). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "hurricane-matthew_00000002", "total_buildings": 38, "severe_count": 7, "severe_pct": 18.42, "severe_footprint_m2": 4250.4423}, "scene_b": {"scene_id": "hurricane-matthew_00000097", "total_buildings": 99, "severe_count": 15, "severe_pct": 15.31, "severe_footprint_m2": 18939.8828}, "higher_severe_pct_scene": "hurricane-matthew_00000002", "higher_severe_area_scene": "hurricane-matthew_00000097"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "hurricane-matthew_00000097"}
+{"query_id": "XBD_Q4_hurricane-michael_00000501_1612", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000501", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Two satellite scenes captured the same wind event (hurricane-michael). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "hurricane-michael_00000501", "total_buildings": 107, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "hurricane-michael_00000429", "total_buildings": 56, "severe_count": 9, "severe_pct": 16.07, "severe_footprint_m2": 69016.5711}, "higher_severe_pct_scene": "hurricane-michael_00000429", "higher_severe_area_scene": "hurricane-michael_00000429"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "hurricane-michael_00000429"}
+{"query_id": "XBD_Q4_hurricane-michael_00000000_1613", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "hurricane-michael_00000000", "split": "train", "qa_split": "train", "disaster": "hurricane-michael", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Two satellite scenes captured the same wind event (hurricane-michael). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "hurricane-michael_00000000", "total_buildings": 54, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "hurricane-michael_00000219", "total_buildings": 113, "severe_count": 4, "severe_pct": 3.54, "severe_footprint_m2": 15026.4051}, "higher_severe_pct_scene": "hurricane-michael_00000219", "higher_severe_area_scene": "hurricane-michael_00000219"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "hurricane-michael_00000219"}
+{"query_id": "XBD_Q4_joplin-tornado_00000054_1615", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000054", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Two satellite scenes captured the same wind event (joplin-tornado). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "joplin-tornado_00000054", "total_buildings": 23, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "joplin-tornado_00000132", "total_buildings": 174, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "higher_severe_pct_scene": "joplin-tornado_00000054", "higher_severe_area_scene": "joplin-tornado_00000054"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "joplin-tornado_00000132"}
+{"query_id": "XBD_Q4_joplin-tornado_00000042_1616", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000042", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "minor-moderate", "question": "Two satellite scenes captured the same wind event (joplin-tornado). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "joplin-tornado_00000042", "total_buildings": 150, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "joplin-tornado_00000075", "total_buildings": 86, "severe_count": 52, "severe_pct": 70.27, "severe_footprint_m2": 262303.9588}, "higher_severe_pct_scene": "joplin-tornado_00000075", "higher_severe_area_scene": "joplin-tornado_00000075"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "joplin-tornado_00000075"}
+{"query_id": "XBD_Q4_joplin-tornado_00000131_1617", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "joplin-tornado_00000131", "split": "tier3", "qa_split": "train", "disaster": "joplin-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Two satellite scenes captured the same wind event (joplin-tornado). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "joplin-tornado_00000131", "total_buildings": 36, "severe_count": 25, "severe_pct": 71.43, "severe_footprint_m2": 136829.9616}, "scene_b": {"scene_id": "joplin-tornado_00000046", "total_buildings": 245, "severe_count": 13, "severe_pct": 5.53, "severe_footprint_m2": 43945.5391}, "higher_severe_pct_scene": "joplin-tornado_00000131", "higher_severe_area_scene": "joplin-tornado_00000131"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "joplin-tornado_00000046"}
+{"query_id": "XBD_Q4_lower-puna-volcano_00000226_1618", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000226", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "no-low-damage", "question": "Two satellite scenes captured the same volcano event (lower-puna-volcano). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "lower-puna-volcano_00000226", "total_buildings": 36, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "lower-puna-volcano_00000224", "total_buildings": 43, "severe_count": 15, "severe_pct": 34.88, "severe_footprint_m2": 31371.0606}, "higher_severe_pct_scene": "lower-puna-volcano_00000224", "higher_severe_area_scene": "lower-puna-volcano_00000224"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "lower-puna-volcano_00000224"}
+{"query_id": "XBD_Q4_lower-puna-volcano_00000121_1619", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000121", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "major-heavy", "question": "Two satellite scenes captured the same volcano event (lower-puna-volcano). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "lower-puna-volcano_00000121", "total_buildings": 28, "severe_count": 7, "severe_pct": 25.0, "severe_footprint_m2": 10480.591}, "scene_b": {"scene_id": "lower-puna-volcano_00000059", "total_buildings": 29, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "higher_severe_pct_scene": "lower-puna-volcano_00000121", "higher_severe_area_scene": "lower-puna-volcano_00000121"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "lower-puna-volcano_00000059"}
+{"query_id": "XBD_Q4_lower-puna-volcano_00000135_1620", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "lower-puna-volcano_00000135", "split": "tier3", "qa_split": "train", "disaster": "lower-puna-volcano", "disaster_type": "volcano", "damage_bucket": "destroyed-heavy", "question": "Two satellite scenes captured the same volcano event (lower-puna-volcano). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "lower-puna-volcano_00000135", "total_buildings": 31, "severe_count": 30, "severe_pct": 96.77, "severe_footprint_m2": 64773.7271}, "scene_b": {"scene_id": "lower-puna-volcano_00000068", "total_buildings": 29, "severe_count": 5, "severe_pct": 18.52, "severe_footprint_m2": 36084.6247}, "higher_severe_pct_scene": "lower-puna-volcano_00000135", "higher_severe_area_scene": "lower-puna-volcano_00000135"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "lower-puna-volcano_00000068"}
+{"query_id": "XBD_Q4_mexico-earthquake_00000049_1621", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000049", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "major-heavy", "question": "Two satellite scenes captured the same earthquake event (mexico-earthquake). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "mexico-earthquake_00000049", "total_buildings": 517, "severe_count": 3, "severe_pct": 0.58, "severe_footprint_m2": 9020.1076}, "scene_b": {"scene_id": "mexico-earthquake_00000191", "total_buildings": 359, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "higher_severe_pct_scene": "mexico-earthquake_00000049", "higher_severe_area_scene": "mexico-earthquake_00000049"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "mexico-earthquake_00000191"}
+{"query_id": "XBD_Q4_mexico-earthquake_00000119_1623", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "mexico-earthquake_00000119", "split": "train", "qa_split": "train", "disaster": "mexico-earthquake", "disaster_type": "earthquake", "damage_bucket": "minor-moderate", "question": "Two satellite scenes captured the same earthquake event (mexico-earthquake). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "mexico-earthquake_00000119", "total_buildings": 54, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "mexico-earthquake_00000142", "total_buildings": 177, "severe_count": 2, "severe_pct": 1.2, "severe_footprint_m2": 6777.0004}, "higher_severe_pct_scene": "mexico-earthquake_00000142", "higher_severe_area_scene": "mexico-earthquake_00000142"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "mexico-earthquake_00000142"}
+{"query_id": "XBD_Q4_midwest-flooding_00000024_1624", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000024", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "no-low-damage", "question": "Two satellite scenes captured the same flooding event (midwest-flooding). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "midwest-flooding_00000024", "total_buildings": 304, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "midwest-flooding_00000189", "total_buildings": 13, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "higher_severe_pct_scene": "midwest-flooding_00000024", "higher_severe_area_scene": "midwest-flooding_00000024"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "midwest-flooding_00000189"}
+{"query_id": "XBD_Q4_midwest-flooding_00000239_1625", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "midwest-flooding_00000239", "split": "train", "qa_split": "train", "disaster": "midwest-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Two satellite scenes captured the same flooding event (midwest-flooding). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "midwest-flooding_00000239", "total_buildings": 15, "severe_count": 2, "severe_pct": 13.33, "severe_footprint_m2": 3725.6746}, "scene_b": {"scene_id": "midwest-flooding_00000011", "total_buildings": 70, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "higher_severe_pct_scene": "midwest-flooding_00000239", "higher_severe_area_scene": "midwest-flooding_00000239"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "midwest-flooding_00000011"}
+{"query_id": "XBD_Q4_moore-tornado_00000098_1627", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "moore-tornado_00000098", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "Two satellite scenes captured the same wind event (moore-tornado). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "moore-tornado_00000098", "total_buildings": 135, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "moore-tornado_00000056", "total_buildings": 271, "severe_count": 59, "severe_pct": 21.77, "severe_footprint_m2": 167613.8527}, "higher_severe_pct_scene": "moore-tornado_00000056", "higher_severe_area_scene": "moore-tornado_00000056"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "moore-tornado_00000056"}
+{"query_id": "XBD_Q4_moore-tornado_00000003_1628", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "moore-tornado_00000003", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Two satellite scenes captured the same wind event (moore-tornado). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "moore-tornado_00000003", "total_buildings": 212, "severe_count": 28, "severe_pct": 14.43, "severe_footprint_m2": 51209.6375}, "scene_b": {"scene_id": "moore-tornado_00000120", "total_buildings": 20, "severe_count": 18, "severe_pct": 90.0, "severe_footprint_m2": 85329.6156}, "higher_severe_pct_scene": "moore-tornado_00000120", "higher_severe_area_scene": "moore-tornado_00000120"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "moore-tornado_00000120"}
+{"query_id": "XBD_Q4_moore-tornado_00000037_1629", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "moore-tornado_00000037", "split": "tier3", "qa_split": "train", "disaster": "moore-tornado", "disaster_type": "wind", "damage_bucket": "no-low-damage", "question": "Two satellite scenes captured the same wind event (moore-tornado). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "moore-tornado_00000037", "total_buildings": 10, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "moore-tornado_00000013", "total_buildings": 163, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "higher_severe_pct_scene": "moore-tornado_00000037", "higher_severe_area_scene": "moore-tornado_00000037"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "moore-tornado_00000013"}
+{"query_id": "XBD_Q4_nepal-flooding_00000267_1630", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000267", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "minor-moderate", "question": "Two satellite scenes captured the same flooding event (nepal-flooding). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "nepal-flooding_00000267", "total_buildings": 180, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "nepal-flooding_00000552", "total_buildings": 26, "severe_count": 9, "severe_pct": 34.62, "severe_footprint_m2": 31881.6092}, "higher_severe_pct_scene": "nepal-flooding_00000552", "higher_severe_area_scene": "nepal-flooding_00000552"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "nepal-flooding_00000552"}
+{"query_id": "XBD_Q4_nepal-flooding_00000536_1631", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000536", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "major-heavy", "question": "Two satellite scenes captured the same flooding event (nepal-flooding). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "nepal-flooding_00000536", "total_buildings": 22, "severe_count": 11, "severe_pct": 50.0, "severe_footprint_m2": 46357.425}, "scene_b": {"scene_id": "nepal-flooding_00000049", "total_buildings": 11, "severe_count": 11, "severe_pct": 100.0, "severe_footprint_m2": 36061.1812}, "higher_severe_pct_scene": "nepal-flooding_00000049", "higher_severe_area_scene": "nepal-flooding_00000536"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "nepal-flooding_00000049"}
+{"query_id": "XBD_Q4_nepal-flooding_00000380_1632", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "nepal-flooding_00000380", "split": "tier3", "qa_split": "train", "disaster": "nepal-flooding", "disaster_type": "flooding", "damage_bucket": "mixed-severe", "question": "Two satellite scenes captured the same flooding event (nepal-flooding). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "nepal-flooding_00000380", "total_buildings": 54, "severe_count": 3, "severe_pct": 6.0, "severe_footprint_m2": 6268.4598}, "scene_b": {"scene_id": "nepal-flooding_00000597", "total_buildings": 107, "severe_count": 2, "severe_pct": 1.92, "severe_footprint_m2": 2744.5802}, "higher_severe_pct_scene": "nepal-flooding_00000380", "higher_severe_area_scene": "nepal-flooding_00000380"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "nepal-flooding_00000597"}
+{"query_id": "XBD_Q4_palu-tsunami_00000000_1633", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000000", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "Two satellite scenes captured the same tsunami event (palu-tsunami). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "palu-tsunami_00000000", "total_buildings": 91, "severe_count": 10, "severe_pct": 10.99, "severe_footprint_m2": 24289.6055}, "scene_b": {"scene_id": "palu-tsunami_00000144", "total_buildings": 160, "severe_count": 3, "severe_pct": 1.88, "severe_footprint_m2": 74320.3148}, "higher_severe_pct_scene": "palu-tsunami_00000000", "higher_severe_area_scene": "palu-tsunami_00000144"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "palu-tsunami_00000144"}
+{"query_id": "XBD_Q4_palu-tsunami_00000168_1634", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000168", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "destroyed-heavy", "question": "Two satellite scenes captured the same tsunami event (palu-tsunami). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "palu-tsunami_00000168", "total_buildings": 189, "severe_count": 175, "severe_pct": 93.09, "severe_footprint_m2": 599615.0504}, "scene_b": {"scene_id": "palu-tsunami_00000120", "total_buildings": 582, "severe_count": 19, "severe_pct": 3.28, "severe_footprint_m2": 90203.9803}, "higher_severe_pct_scene": "palu-tsunami_00000168", "higher_severe_area_scene": "palu-tsunami_00000168"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "palu-tsunami_00000120"}
+{"query_id": "XBD_Q4_palu-tsunami_00000128_1635", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "palu-tsunami_00000128", "split": "train", "qa_split": "train", "disaster": "palu-tsunami", "disaster_type": "tsunami", "damage_bucket": "no-low-damage", "question": "Two satellite scenes captured the same tsunami event (palu-tsunami). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "palu-tsunami_00000128", "total_buildings": 178, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "palu-tsunami_00000093", "total_buildings": 459, "severe_count": 155, "severe_pct": 34.22, "severe_footprint_m2": 648079.6453}, "higher_severe_pct_scene": "palu-tsunami_00000093", "higher_severe_area_scene": "palu-tsunami_00000093"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "palu-tsunami_00000093"}
+{"query_id": "XBD_Q4_pinery-bushfire_00001282_1636", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001282", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Two satellite scenes captured the same fire event (pinery-bushfire). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "pinery-bushfire_00001282", "total_buildings": 12, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "pinery-bushfire_00000103", "total_buildings": 15, "severe_count": 2, "severe_pct": 15.38, "severe_footprint_m2": 961.3134}, "higher_severe_pct_scene": "pinery-bushfire_00000103", "higher_severe_area_scene": "pinery-bushfire_00000103"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "pinery-bushfire_00000103"}
+{"query_id": "XBD_Q4_pinery-bushfire_00000897_1637", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00000897", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Two satellite scenes captured the same fire event (pinery-bushfire). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "pinery-bushfire_00000897", "total_buildings": 20, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "pinery-bushfire_00001507", "total_buildings": 13, "severe_count": 1, "severe_pct": 16.67, "severe_footprint_m2": 6435.7118}, "higher_severe_pct_scene": "pinery-bushfire_00001507", "higher_severe_area_scene": "pinery-bushfire_00001507"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "pinery-bushfire_00001507"}
+{"query_id": "XBD_Q4_pinery-bushfire_00001719_1638", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "pinery-bushfire_00001719", "split": "tier3", "qa_split": "train", "disaster": "pinery-bushfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Two satellite scenes captured the same fire event (pinery-bushfire). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "pinery-bushfire_00001719", "total_buildings": 28, "severe_count": 6, "severe_pct": 23.08, "severe_footprint_m2": 2090.2815}, "scene_b": {"scene_id": "pinery-bushfire_00000690", "total_buildings": 10, "severe_count": 3, "severe_pct": 33.33, "severe_footprint_m2": 4114.8143}, "higher_severe_pct_scene": "pinery-bushfire_00000690", "higher_severe_area_scene": "pinery-bushfire_00000690"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "pinery-bushfire_00000690"}
+{"query_id": "XBD_Q4_portugal-wildfire_00001682_1639", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001682", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Two satellite scenes captured the same fire event (portugal-wildfire). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "portugal-wildfire_00001682", "total_buildings": 90, "severe_count": 24, "severe_pct": 26.97, "severe_footprint_m2": 59843.6872}, "scene_b": {"scene_id": "portugal-wildfire_00000791", "total_buildings": 10, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "higher_severe_pct_scene": "portugal-wildfire_00001682", "higher_severe_area_scene": "portugal-wildfire_00001682"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "portugal-wildfire_00000791"}
+{"query_id": "XBD_Q4_portugal-wildfire_00001399_1640", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00001399", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Two satellite scenes captured the same fire event (portugal-wildfire). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "portugal-wildfire_00001399", "total_buildings": 102, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "portugal-wildfire_00000627", "total_buildings": 23, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "higher_severe_pct_scene": "portugal-wildfire_00001399", "higher_severe_area_scene": "portugal-wildfire_00001399"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "portugal-wildfire_00000627"}
+{"query_id": "XBD_Q4_portugal-wildfire_00000698_1641", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "portugal-wildfire_00000698", "split": "tier3", "qa_split": "train", "disaster": "portugal-wildfire", "disaster_type": "fire", "damage_bucket": "major-heavy", "question": "Two satellite scenes captured the same fire event (portugal-wildfire). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "portugal-wildfire_00000698", "total_buildings": 12, "severe_count": 1, "severe_pct": 8.33, "severe_footprint_m2": 2127.463}, "scene_b": {"scene_id": "portugal-wildfire_00001572", "total_buildings": 55, "severe_count": 27, "severe_pct": 50.0, "severe_footprint_m2": 61858.7255}, "higher_severe_pct_scene": "portugal-wildfire_00001572", "higher_severe_area_scene": "portugal-wildfire_00001572"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "portugal-wildfire_00001572"}
+{"query_id": "XBD_Q4_santa-rosa-wildfire_00000166_1642", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "santa-rosa-wildfire_00000166", "split": "train", "qa_split": "train", "disaster": "santa-rosa-wildfire", "disaster_type": "fire", "damage_bucket": "no-low-damage", "question": "Two satellite scenes captured the same fire event (santa-rosa-wildfire). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "santa-rosa-wildfire_00000166", "total_buildings": 208, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "santa-rosa-wildfire_00000170", "total_buildings": 28, "severe_count": 18, "severe_pct": 64.29, "severe_footprint_m2": 63339.5145}, "higher_severe_pct_scene": "santa-rosa-wildfire_00000170", "higher_severe_area_scene": "santa-rosa-wildfire_00000170"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "santa-rosa-wildfire_00000170"}
+{"query_id": "XBD_Q4_socal-fire_00000494_1645", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "socal-fire_00000494", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Two satellite scenes captured the same fire event (socal-fire). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "socal-fire_00000494", "total_buildings": 21, "severe_count": 12, "severe_pct": 57.14, "severe_footprint_m2": 65513.695}, "scene_b": {"scene_id": "socal-fire_00000436", "total_buildings": 18, "severe_count": 13, "severe_pct": 72.22, "severe_footprint_m2": 37870.0961}, "higher_severe_pct_scene": "socal-fire_00000436", "higher_severe_area_scene": "socal-fire_00000494"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "socal-fire_00000436"}
+{"query_id": "XBD_Q4_socal-fire_00000358_1647", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "socal-fire_00000358", "split": "train", "qa_split": "train", "disaster": "socal-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Two satellite scenes captured the same fire event (socal-fire). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "socal-fire_00000358", "total_buildings": 17, "severe_count": 16, "severe_pct": 94.12, "severe_footprint_m2": 28226.1103}, "scene_b": {"scene_id": "socal-fire_00000658", "total_buildings": 65, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "higher_severe_pct_scene": "socal-fire_00000358", "higher_severe_area_scene": "socal-fire_00000358"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "socal-fire_00000658"}
+{"query_id": "XBD_Q4_sunda-tsunami_00000023_1648", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000023", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "major-heavy", "question": "Two satellite scenes captured the same tsunami event (sunda-tsunami). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "sunda-tsunami_00000023", "total_buildings": 1195, "severe_count": 3, "severe_pct": 0.25, "severe_footprint_m2": 6788.7146}, "scene_b": {"scene_id": "sunda-tsunami_00000119", "total_buildings": 15, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "higher_severe_pct_scene": "sunda-tsunami_00000023", "higher_severe_area_scene": "sunda-tsunami_00000023"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "sunda-tsunami_00000119"}
+{"query_id": "XBD_Q4_sunda-tsunami_00000008_1649", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "sunda-tsunami_00000008", "split": "tier3", "qa_split": "train", "disaster": "sunda-tsunami", "disaster_type": "tsunami", "damage_bucket": "no-low-damage", "question": "Two satellite scenes captured the same tsunami event (sunda-tsunami). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "sunda-tsunami_00000008", "total_buildings": 18, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "scene_b": {"scene_id": "sunda-tsunami_00000005", "total_buildings": 51, "severe_count": 8, "severe_pct": 15.69, "severe_footprint_m2": 3945.1524}, "higher_severe_pct_scene": "sunda-tsunami_00000005", "higher_severe_area_scene": "sunda-tsunami_00000005"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "sunda-tsunami_00000005"}
+{"query_id": "XBD_Q4_tuscaloosa-tornado_00000324_1650", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000324", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "destroyed-heavy", "question": "Two satellite scenes captured the same wind event (tuscaloosa-tornado). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "tuscaloosa-tornado_00000324", "total_buildings": 73, "severe_count": 4, "severe_pct": 5.97, "severe_footprint_m2": 3005.1628}, "scene_b": {"scene_id": "tuscaloosa-tornado_00000030", "total_buildings": 124, "severe_count": 21, "severe_pct": 17.36, "severe_footprint_m2": 115258.9714}, "higher_severe_pct_scene": "tuscaloosa-tornado_00000030", "higher_severe_area_scene": "tuscaloosa-tornado_00000030"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "tuscaloosa-tornado_00000030"}
+{"query_id": "XBD_Q4_tuscaloosa-tornado_00000112_1651", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000112", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Two satellite scenes captured the same wind event (tuscaloosa-tornado). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "tuscaloosa-tornado_00000112", "total_buildings": 112, "severe_count": 44, "severe_pct": 41.12, "severe_footprint_m2": 198723.5842}, "scene_b": {"scene_id": "tuscaloosa-tornado_00000059", "total_buildings": 70, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "higher_severe_pct_scene": "tuscaloosa-tornado_00000112", "higher_severe_area_scene": "tuscaloosa-tornado_00000112"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "tuscaloosa-tornado_00000059"}
+{"query_id": "XBD_Q4_tuscaloosa-tornado_00000095_1652", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "tuscaloosa-tornado_00000095", "split": "tier3", "qa_split": "train", "disaster": "tuscaloosa-tornado", "disaster_type": "wind", "damage_bucket": "mixed-severe", "question": "Two satellite scenes captured the same wind event (tuscaloosa-tornado). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "tuscaloosa-tornado_00000095", "total_buildings": 120, "severe_count": 29, "severe_pct": 25.44, "severe_footprint_m2": 116151.4108}, "scene_b": {"scene_id": "tuscaloosa-tornado_00000250", "total_buildings": 113, "severe_count": 0, "severe_pct": 0.0, "severe_footprint_m2": 0.0}, "higher_severe_pct_scene": "tuscaloosa-tornado_00000095", "higher_severe_area_scene": "tuscaloosa-tornado_00000095"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "tuscaloosa-tornado_00000250"}
+{"query_id": "XBD_Q4_woolsey-fire_00000088_1653", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000088", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Two satellite scenes captured the same fire event (woolsey-fire). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "woolsey-fire_00000088", "total_buildings": 16, "severe_count": 5, "severe_pct": 31.25, "severe_footprint_m2": 11109.9175}, "scene_b": {"scene_id": "woolsey-fire_00000497", "total_buildings": 69, "severe_count": 12, "severe_pct": 18.46, "severe_footprint_m2": 31863.5148}, "higher_severe_pct_scene": "woolsey-fire_00000088", "higher_severe_area_scene": "woolsey-fire_00000497"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "woolsey-fire_00000497"}
+{"query_id": "XBD_Q4_woolsey-fire_00000568_1654", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000568", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Two satellite scenes captured the same fire event (woolsey-fire). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "woolsey-fire_00000568", "total_buildings": 51, "severe_count": 1, "severe_pct": 2.0, "severe_footprint_m2": 6786.3652}, "scene_b": {"scene_id": "woolsey-fire_00000419", "total_buildings": 11, "severe_count": 1, "severe_pct": 10.0, "severe_footprint_m2": 768.5366}, "higher_severe_pct_scene": "woolsey-fire_00000419", "higher_severe_area_scene": "woolsey-fire_00000568"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "woolsey-fire_00000419"}
+{"query_id": "XBD_Q4_woolsey-fire_00000557_1655", "template": "XBD-Q4", "source_dataset": "xBD", "scene_id": "woolsey-fire_00000557", "split": "tier3", "qa_split": "train", "disaster": "woolsey-fire", "disaster_type": "fire", "damage_bucket": "destroyed-heavy", "question": "Two satellite scenes captured the same fire event (woolsey-fire). Compare their severe-damage ratio and severe-damage footprint area. Which scene shows greater structural destruction?", "input_layers": {"scene_a": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}, "scene_b": {"pre_image": "pre.png", "post_image": "post.png", "post_target_mask": "post_target.png", "damage_polygons_xy": "building_damage_xy.geojson", "damage_polygons_geo": "building_damage_lnglat.geojson", "quadrants": "quadrants.geojson", "grid_4x4": "grid_4x4.geojson"}}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "LoadSecondScene", "CountDamageDistribution", "ComputeSevereDamageAreaFromXYGSD", "CompareScenesTableOutput", "Terminate"], "gt_answer": {"scene_a": {"scene_id": "woolsey-fire_00000557", "total_buildings": 22, "severe_count": 5, "severe_pct": 22.73, "severe_footprint_m2": 20555.6206}, "scene_b": {"scene_id": "woolsey-fire_00000262", "total_buildings": 31, "severe_count": 11, "severe_pct": 35.48, "severe_footprint_m2": 13626.9703}, "higher_severe_pct_scene": "woolsey-fire_00000262", "higher_severe_area_scene": "woolsey-fire_00000557"}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["counting", "area_computation", "cross_scene_comparison"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00", "scene_id_b": "woolsey-fire_00000262"}
+{"query_id": "XBD_Q6_subset_1656", "template": "XBD-Q6", "source_dataset": "xBD", "scene_id": "ALL_SELECTED_SCENES", "split": "mixed", "qa_split": "train", "disaster": "multiple", "disaster_type": "multiple", "damage_bucket": "mixed", "question": "Across the 300-scene sampled subset covering multiple disaster events, which disaster event has the highest average severe-damage percentage? Provide the full event ranking with scene counts and average severe percentages.", "input_layers": {"subset_csv": "metadata/xbd_selected_subset.csv"}, "gt_tool_chain": ["LoadXBDScene", "LoadDamagePolygons", "ValidateGSD", "GroupByDisasterEvent", "ComputeAverageSeverePctPerEvent", "RankEventsBySeverePct", "Terminate"], "gt_answer": {"highest_avg_severe_pct_event": "santa-rosa-wildfire", "highest_avg_severe_pct": 50.7, "total_scenes_in_subset": 300, "event_ranking": [{"disaster": "santa-rosa-wildfire", "avg_severe_pct": 50.7, "scene_count": 16}, {"disaster": "hurricane-harvey", "avg_severe_pct": 48.99, "scene_count": 24}, {"disaster": "hurricane-matthew", "avg_severe_pct": 43.77, "scene_count": 25}, {"disaster": "guatemala-volcano", "avg_severe_pct": 40.0, "scene_count": 4}, {"disaster": "hurricane-florence", "avg_severe_pct": 34.36, "scene_count": 18}, {"disaster": "socal-fire", "avg_severe_pct": 30.8, "scene_count": 20}, {"disaster": "joplin-tornado", "avg_severe_pct": 30.57, "scene_count": 10}, {"disaster": "lower-puna-volcano", "avg_severe_pct": 25.02, "scene_count": 7}, {"disaster": "nepal-flooding", "avg_severe_pct": 23.37, "scene_count": 31}, {"disaster": "woolsey-fire", "avg_severe_pct": 21.43, "scene_count": 14}, {"disaster": "palu-tsunami", "avg_severe_pct": 20.76, "scene_count": 10}, {"disaster": "tuscaloosa-tornado", "avg_severe_pct": 15.94, "scene_count": 14}, {"disaster": "moore-tornado", "avg_severe_pct": 15.22, "scene_count": 10}, {"disaster": "pinery-bushfire", "avg_severe_pct": 15.22, "scene_count": 8}, {"disaster": "midwest-flooding", "avg_severe_pct": 13.58, "scene_count": 12}, {"disaster": "hurricane-michael", "avg_severe_pct": 12.3, "scene_count": 39}, {"disaster": "portugal-wildfire", "avg_severe_pct": 9.6, "scene_count": 25}, {"disaster": "sunda-tsunami", "avg_severe_pct": 3.98, "scene_count": 4}, {"disaster": "mexico-earthquake", "avg_severe_pct": 0.2, "scene_count": 9}]}, "gt_type": "direct_xbd_labels_plus_deterministic_geometry", "required_reasoning": ["aggregation", "averaging", "ranking"], "difficulty": "hard", "pipeline_version": "v1", "created_at": "2026-05-13T20:40:43.830608+00:00"}