Add route and hex open-ended question sets

Dataset/combined_open_question_report.json

@@ -0,0 +1,50 @@
+{
+  "status": "PASS",
+  "datasets": {
+    "hex_pathfinding_dataset_3000": {
+      "dataset_version": "hex-pathfinding-3.0.0",
+      "images": 3000,
+      "open_questions": 3000,
+      "open_answers": 3000
+    },
+    "route_dataset_3000": {
+      "dataset_version": "route-3.0.0",
+      "images": 3000,
+      "open_questions": 3000,
+      "open_answers": 3000
+    }
+  },
+  "combined_open_questions": 6000,
+  "combined_open_answers": 6000,
+  "unique_question_ids": 6000,
+  "resolved_image_paths": 6000,
+  "public_columns": [
+    "dataset",
+    "dataset_version",
+    "question_id",
+    "image",
+    "image_path",
+    "prompt"
+  ],
+  "private_columns": [
+    "dataset",
+    "dataset_version",
+    "question_id",
+    "image",
+    "image_path",
+    "prompt",
+    "target",
+    "target_position",
+    "target_degree",
+    "lines_at_target",
+    "colors",
+    "reached",
+    "unreached",
+    "answer",
+    "home_on_boundary",
+    "neighbour_count",
+    "holes_touching",
+    "walkable_touching",
+    "hole_directions"
+  ]
+}

Dataset/combined_suite_rebuild_report.json

@@ -25,13 +25,36 @@
       "cube_structure": 15000,
       "depth_height": 15000,
       "embedded_figure": 15000,
-      "fbd": 9839
+      "fbd": 15000,
+      "fold_punch": 15000,
+      "gauge_reading": 15000,
+      "gear_train": 15000,
+      "hex_pathfinding": 15000,
+      "impossible_object": 15000,
+      "laser_mirror": 15000,
+      "line_intersect": 15000,
+      "nested_hexagons": 15000,
+      "nested_squares": 15000,
+      "nested_triangles": 15000,
+      "occluded_pattern": 15000,
+      "optical_illusion": 15000,
+      "orthographic": 15000,
+      "overlap_circles": 15000,
+      "physical_stability": 15000,
+      "polyhedron": 15000,
+      "projectile_motion": 5000,
+      "rotation_match": 15000,
+      "route_puzzle": 15000,
+      "rpm": 15000,
+      "shadow_inference": 15000,
+      "surface_topology": 15000,
+      "symmetry_pattern": 15000
     },
-    "answer_rows": 159839,
+    "answer_rows": 500000,
     "extra_prefixes": [],
     "missing_prefixes": [],
     "present": true,
-    "question_answer_row_counts_match": false,
+    "question_answer_row_counts_match": true,
     "question_prefix_counts": {
       "angle_estimation": 15000,
       "clock_reading": 15000,
@@ -69,7 +92,10 @@
       "symmetry_pattern": 15000
     },
     "question_rows": 500000,
-    "stale_or_mismatched": {}
+    "stale_or_mismatched": {
+      "hex_pathfinding": "dataset_version is ['hex-pathfinding-2.0.0'], expected hex-pathfinding-3.0.0",
+      "route": "dataset_version is ['route-2.0.0'], expected route-3.0.0"
+    }
   },
   "discovered_dataset_count": 34,
   "expected_totals": {
@@ -197,7 +223,7 @@
       "prefix": "hex_pathfinding",
       "questions": 15000,
       "slug": "hex_pathfinding",
-      "version": "hex-pathfinding-2.0.0"
+      "version": "hex-pathfinding-3.0.0"
     },
     "impossible_object_dataset_3000": {
       "answers": 15000,
@@ -317,7 +343,7 @@
       "prefix": "route_puzzle",
       "questions": 15000,
       "slug": "route",
-      "version": "route-2.0.0"
+      "version": "route-3.0.0"
     },
     "rpm_dataset_3000": {
       "answers": 15000,

Dataset/final_release_check.py

@@ -8,14 +8,23 @@ errors=[];lines=[]
 datasets=sorted(d for d in root.iterdir() if d.is_dir() and (d/'annotations.jsonl').exists())
 for d in datasets:
     rows=[json.loads(x) for x in (d/'annotations.jsonl').read_text(encoding='utf-8').splitlines()]
-    names={p.name for p in d.iterdir()};missing=required-names
+    names={p.name for p in d.iterdir() if p.name!='__pycache__'};missing=required-names
     if missing:errors.append(f'{d.name}: missing {sorted(missing)}')
-    allowed=required | {name for name in names if re.fullmatch(r'(generate|validate|flatten)_.+\.py',name)}
+    manifest=json.loads((d/'build_manifest.json').read_text(encoding='utf-8'))
+    open_files=set(manifest.get('open_question_files',[]))
+    open_support={'build_open_questions.py','validate_open_questions.py','open_validation_metrics.json','open_validation_report.txt'} if open_files else set()
+    allowed=required | open_files | open_support | {name for name in names if re.fullmatch(r'(generate|validate|flatten)_.+\.py',name)}
     extra=names-allowed
     if extra:errors.append(f'{d.name}: unexpected {sorted(extra)}')
     for prefix in ('generate_','validate_','flatten_'):
-        if sum(name.startswith(prefix) and name.endswith('.py') for name in names)!=1:
+        standard_scripts=[name for name in names if name.startswith(prefix) and name.endswith('.py') and name!='validate_open_questions.py']
+        if len(standard_scripts)!=1:
             errors.append(f'{d.name}: expected exactly one {prefix}*.py')
+    if open_files:
+        if not open_files <= names:errors.append(f'{d.name}: missing declared open-question files')
+        with (d/'open_questions.csv').open(encoding='utf-8-sig',newline='') as f:
+            reader=csv.reader(f);open_header=next(reader);open_questions=sum(1 for _ in reader)
+        if open_header!=['question_id','image','prompt'] or open_questions!=len(rows):errors.append(f'{d.name}: open_questions contract')
     if any(re.search(r'_v\d+$',p.stem) for p in d.iterdir() if p.is_file()):errors.append(f'{d.name}: version-suffixed top-level file')
     images=sum(1 for _ in (d/'images').glob('*.png'))
     if images!=len(rows):errors.append(f'{d.name}: images {images}/{len(rows)}')

Dataset/final_release_check.txt

@@ -21,7 +21,7 @@ fbd_dataset_3000: README.md, annotations.jsonl, answer_key.csv, build_manifest.j
 fold_punch_dataset_3000: README.md, annotations.jsonl, answer_key.csv, build_manifest.json, dataset_final.csv, dataset_final.jsonl, flatten_annotations.py, generate_fold_punch_dataset.py, generation_stats.json, images/, question_set.csv, validate_fold_punch_dataset.py, validation_metrics.json, validation_report.txt
 gauge_reading_dataset_3000: README.md, annotations.jsonl, answer_key.csv, build_manifest.json, dataset_final.csv, dataset_final.jsonl, flatten_annotations.py, generate_gauge_reading_dataset.py, generation_stats.json, images/, question_set.csv, validate_gauge_dataset.py, validation_metrics.json, validation_report.txt
 gear_train_dataset_3000: README.md, annotations.jsonl, answer_key.csv, build_manifest.json, dataset_final.csv, dataset_final.jsonl, flatten_annotations.py, generate_gear_train_dataset.py, generation_stats.json, images/, question_set.csv, validate_gear_train_dataset.py, validation_metrics.json, validation_report.txt
-hex_pathfinding_dataset_3000: README.md, annotations.jsonl, answer_key.csv, build_manifest.json, dataset_final.csv, dataset_final.jsonl, flatten_annotations.py, generate_hex_pathfinding_dataset.py, generation_stats.json, images/, question_set.csv, validate_hex_pathfinding_dataset.py, validation_metrics.json, validation_report.txt
+hex_pathfinding_dataset_3000: README.md, annotations.jsonl, answer_key.csv, build_manifest.json, build_open_questions.py, dataset_final.csv, dataset_final.jsonl, flatten_annotations.py, generate_hex_pathfinding_dataset.py, generation_stats.json, images/, open_annotations.jsonl, open_answer_key.csv, open_questions.csv, open_validation_metrics.json, open_validation_report.txt, question_set.csv, validate_hex_pathfinding_dataset.py, validate_open_questions.py, validation_metrics.json, validation_report.txt
 impossible_object_dataset_3000: README.md, annotations.jsonl, answer_key.csv, build_manifest.json, dataset_final.csv, dataset_final.jsonl, flatten_annotations.py, generate_impossible_object_dataset.py, generation_stats.json, images/, question_set.csv, validate_impossible_object_dataset.py, validation_metrics.json, validation_report.txt
 laser_mirror_dataset_3000: README.md, annotations.jsonl, answer_key.csv, build_manifest.json, dataset_final.csv, dataset_final.jsonl, flatten_annotations.py, generate_laser_mirror_dataset.py, generation_stats.json, images/, question_set.csv, validate_laser_mirror_dataset.py, validation_metrics.json, validation_report.txt
 line_intersection_dataset_3000: README.md, annotations.jsonl, answer_key.csv, build_manifest.json, dataset_final.csv, dataset_final.jsonl, flatten_annotations.py, generate_line_intersection_dataset.py, generation_stats.json, images/, question_set.csv, validate_line_intersect_dataset.py, validation_metrics.json, validation_report.txt
@@ -36,7 +36,7 @@ physical_stability_dataset_3000: README.md, annotations.jsonl, answer_key.csv, b
 polyhedron_dataset_3000: README.md, annotations.jsonl, answer_key.csv, build_manifest.json, dataset_final.csv, dataset_final.jsonl, flatten_annotations.py, generate_polyhedron_dataset.py, generation_stats.json, images/, question_set.csv, validate_polyhedron_dataset.py, validation_metrics.json, validation_report.txt
 projectile_motion_dataset_1000: README.md, annotations.jsonl, answer_key.csv, build_manifest.json, dataset_final.csv, dataset_final.jsonl, flatten_annotations.py, generate_projectile_motion_dataset.py, generation_stats.json, images/, question_set.csv, validate_projectile_motion_dataset.py, validation_metrics.json, validation_report.txt
 rotation_matching_dataset_3000: README.md, annotations.jsonl, answer_key.csv, build_manifest.json, dataset_final.csv, dataset_final.jsonl, flatten_annotations.py, generate_rotation_matching_dataset.py, generation_stats.json, images/, question_set.csv, validate_rotation_dataset.py, validation_metrics.json, validation_report.txt
-route_dataset_3000: README.md, annotations.jsonl, answer_key.csv, build_manifest.json, dataset_final.csv, dataset_final.jsonl, flatten_annotations.py, generate_route_dataset.py, generation_stats.json, images/, question_set.csv, validate_route_dataset.py, validation_metrics.json, validation_report.txt
+route_dataset_3000: README.md, __pycache__/, annotations.jsonl, answer_key.csv, build_manifest.json, build_open_questions.py, dataset_final.csv, dataset_final.jsonl, flatten_annotations.py, generate_route_dataset.py, generation_stats.json, images/, open_annotations.jsonl, open_answer_key.csv, open_questions.csv, open_validation_metrics.json, open_validation_report.txt, question_set.csv, validate_open_questions.py, validate_route_dataset.py, validation_metrics.json, validation_report.txt
 rpm_dataset_3000: README.md, annotations.jsonl, answer_key.csv, build_manifest.json, dataset_final.csv, dataset_final.jsonl, flatten_annotations.py, generate_rpm_dataset.py, generation_stats.json, images/, question_set.csv, validate_rpm_dataset.py, validation_metrics.json, validation_report.txt
 shadow_inference_dataset_3000: README.md, annotations.jsonl, answer_key.csv, build_manifest.json, dataset_final.csv, dataset_final.jsonl, flatten_annotations.py, generate_shadow_inference_dataset.py, generation_stats.json, images/, question_set.csv, validate_shadow_dataset.py, validation_metrics.json, validation_report.txt
 surface_topology_dataset_3000: README.md, annotations.jsonl, answer_key.csv, build_manifest.json, dataset_final.csv, dataset_final.jsonl, flatten_annotations.py, generate_surface_topology_dataset.py, generation_stats.json, images/, question_set.csv, validate_surface_topology_dataset.py, validation_metrics.json, validation_report.txt

Dataset/final_suite_audit.json

@@ -16826,7 +16826,7 @@
         "p95": 11.0
       }
     },
-    "dataset_version": "hex-pathfinding-2.0.0",
+    "dataset_version": "hex-pathfinding-3.0.0",
     "degeneracy_flags": {},
     "features_at_v_ge_0_10": {
       "difficulty_score": {
@@ -29547,7 +29547,7 @@
         "p95": 12.0
       }
     },
-    "dataset_version": "route-2.0.0",
+    "dataset_version": "route-3.0.0",
     "degeneracy_flags": {},
     "features_at_v_ge_0_10": {
       "crossing_count": {
@@ -32915,4 +32915,4 @@
       "symmetry_type"
     ]
   }
-}
+}

Dataset/final_suite_audit.md

@@ -16,7 +16,7 @@
 | fold_punch_dataset_3000 | fold-punch-2.0.0 | yes | 0 | 50.0% | 50.0% | 25.0% | 32.2% | 33.3% | 3000/3000 |
 | gauge_reading_dataset_3000 | legacy-current | no | 0 | 80.0% | 50.0% | 9.9% | 60.0% | 3.9% | 3000/3000 |
 | gear_train_dataset_3000 | legacy-current | no | 0 | 38.1% | 100.0% | 25.8% | 0.7% | 33.3% | 3000/3000 |
-| hex_pathfinding_dataset_3000 | hex-pathfinding-2.0.0 | yes | 0 | 11.1% | 50.0% | 21.6% | 22.0% | 33.3% | 3000/3000 |
+| hex_pathfinding_dataset_3000 | hex-pathfinding-3.0.0 | yes | 0 | 11.1% | 50.0% | 21.6% | 22.0% | 33.3% | 3000/3000 |
 | impossible_object_dataset_3000 | impossible-object-4.0.0 | no | 0 | 50.1% | 50.0% | 20.7% | 25.0% | 50.0% | 3000/3000 |
 | laser_mirror_dataset_3000 | legacy-current | no | 0 | 25.1% | 75.0% | 52.3% | 4.4% | 3.7% | 3000/3000 |
 | line_intersection_dataset_3000 | line-intersection-3.0.0 | yes | 0 | 40.8% | 27.3% | 20.8% | 19.3% | 49.6% | 3000/3000 |
@@ -31,7 +31,7 @@
 | polyhedron_dataset_3000 | polyhedron-4.0.0 | yes | 0 | 21.1% | 100.0% | 19.5% | 27.5% | 100.0% | 3000/3000 |
 | projectile_motion_dataset_1000 | projectile-motion-1.0.0 | no | 0 | 2.8% | 65.4% | 1.0% | 0.4% | 68.8% | 1000/1000 |
 | rotation_matching_dataset_3000 | rotation-matching-2.0.0 | no | 0 | 25.8% | 25.0% | 15.2% | 26.0% | 14.3% | 3000/3000 |
-| route_dataset_3000 | route-2.0.0 | yes | 0 | 23.4% | 51.3% | 11.3% | 3.5% | 26.3% | 3000/3000 |
+| route_dataset_3000 | route-3.0.0 | yes | 0 | 23.4% | 51.3% | 11.3% | 3.5% | 26.3% | 3000/3000 |
 | rpm_dataset_3000 | rpm-2.0.0 | yes | 0 | 41.8% | 13.1% | 29.9% | 17.1% | 51.5% | 3000/3000 |
 | shadow_inference_dataset_3000 | shadow-inference-2.0.0 | yes | 0 | 37.8% | 36.0% | 29.6% | 40.0% | 50.0% | 3000/3000 |
 | surface_topology_dataset_3000 | surface-topology-3.0.0 | yes | 0 | 27.3% | 75.0% | 51.8% | 27.3% | 52.3% | 3000/3000 |

Dataset/hex_pathfinding_dataset_3000/README.md

@@ -1,6 +1,6 @@
 # GRIP Hex Grid Pathfinding Dataset 3000
 
-Current version: `hex-pathfinding-2.0.0`. The superseded initial build is preserved under `archive/v1/`; v2 is unsuffixed at the dataset root.
+Current version: `hex-pathfinding-3.0.0`. Version 3 adds supplementary open-ended questions and leaves the existing images and five-level questions and answers unchanged.
 
 ## 1. Overview
 
@@ -61,7 +61,7 @@ Outputs include `annotations.jsonl`, `images/`, `dataset_final.csv`, `dataset_fi
 
 The validator independently rebuilds the coordinate set and traversable graph from `all_tiles`, runs its own BFS, recounts shortest paths, verifies every edge in the stored path sequence is one of the six axial neighbors, and reruns BFS after removing the Level 5 tile. It proves the outcome-specific Level 5 invariant for every item: a no-path tile is a true START/HOME cut vertex, an increase tile belongs to every original shortest path, and a same-length case retains an avoiding shortest path. PNG probes recover white, black, grey, START, and HOME as distinct classes and recover the exact grey-hole count in all 3,000 images.
 
-The v2 reference-frame declaration is `axial hex coordinates (q,r)`. All stored graph quantities and every consuming question use that same frame. Full association matrices, answer baselines, parameter distributions, guard injections, and PNG recovery totals are in `validation_metrics.json` and `validation_report.txt`.
+The reference-frame declaration is `axial hex coordinates (q,r)`. All stored graph quantities and every consuming five-level question use that same frame. Full association matrices, answer baselines, parameter distributions, guard injections, and PNG recovery totals are in `validation_metrics.json` and `validation_report.txt`.
 
 Manual launch traces confirmed 5-, 7-, and 13-move sample paths using only legal six-neighbor transitions.
 
@@ -70,3 +70,16 @@ Manual launch traces confirmed 5-, 7-, and 13-move sample paths using only legal
 Skills tested include graph construction from a visual grid, six-neighbor adjacency, hazard avoidance, optimal path planning, shortest-path multiplicity, articulation reasoning, and counterfactual replanning.
 
 Limitations: only hexagonal grids are included; holes are static; every scene has one START and one HOME; movement costs are uniform; there are no dynamic agents, terrain weights, diagonal-like jumps, multiple goals, or partial observability.
+
+## 8. Supplementary open-ended questions (v3)
+
+Each image has one additional open-ended question about the visible neighborhood of the green HOME tile. The prompt uses only printed labels, colors, and the fixed natural-language vocabulary `upper-left`, `upper-right`, `left`, `right`, `lower-left`, and `lower-right`; it never exposes axial coordinates. The requested facts are stored separately as `home_on_boundary`, `neighbour_count`, `holes_touching`, `walkable_touching`, and `hole_directions` so partial correctness can be measured without accepting an incorrect combined conclusion.
+
+The requested preflight showed that neighbor count is informative and should remain in the prompt: HOME is on the boundary in 1,606 images and interior in 1,394; `neighbour_count` is 3 for 475 images, 4 for 1,131, and 6 for 1,394. Walkable-neighbor counts range from 1 to 6. No open-answer field has a constant-answer baseline above 60%; complete distributions appear in `open_validation_metrics.json`.
+
+- `open_questions.csv` is public and contains exactly `question_id,image,prompt`.
+- `open_answer_key.csv` is answer-key-side and contains the target and the five separately scored ground-truth fields.
+- `open_annotations.jsonl` is answer-key-side and adds the private coordinate-based derivation. Coordinates are deliberately absent from the prompt because the render has no printed coordinate axes.
+- `build_open_questions.py` rebuilds these artifacts; `validate_open_questions.py` independently re-derives them and recovers every HOME-neighborhood fact from all 3,000 PNGs.
+
+Do not provide `open_answer_key.csv` or `open_annotations.jsonl` to a model under evaluation; the annotation derivation directly exposes the answer-generating scene structure.

Dataset/hex_pathfinding_dataset_3000/annotations.jsonl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4fad9e36e4785917af79584f6666b2ff8154aa70230954649cafe4cd57719b13
+oid sha256:8433db986c232283fc117c13ef6662fedc17164486494f1749f0cee8bde8675e
 size 14244633

Dataset/hex_pathfinding_dataset_3000/build_manifest.json

@@ -1,9 +1,35 @@
 {
-  "constraint_source": "generator constants and validation metrics",
+  "constraint_set": {
+    "existing_five_level_questions_unchanged": true,
+    "fixed_open_direction_vocabulary": [
+      "upper-left",
+      "upper-right",
+      "left",
+      "right",
+      "lower-left",
+      "lower-right"
+    ],
+    "no_coordinates_in_open_prompt": true,
+    "open_ground_truth_independently_rederived": true,
+    "open_png_recoverability_all_items": true,
+    "open_public_fields_exact": [
+      "question_id",
+      "image",
+      "prompt"
+    ],
+    "open_question_count_per_image": 1
+  },
+  "constraint_source": "generator constants and open-question validation metrics",
   "current_build_layout": "unsuffixed",
-  "dataset_version": "hex-pathfinding-2.0.0",
+  "dataset_version": "hex-pathfinding-3.0.0",
   "generator_commit": "e6975d7f82e6fc42f77fdc6434e733c10adb83d6",
   "images": 3000,
+  "open_question_files": [
+    "open_questions.csv",
+    "open_answer_key.csv",
+    "open_annotations.jsonl"
+  ],
+  "open_questions": 3000,
   "question_set_fields": [
     "question_id",
     "task",

Dataset/hex_pathfinding_dataset_3000/build_open_questions.py

@@ -0,0 +1,102 @@
+"""Build one deterministic open-ended HOME-neighborhood question per image."""
+from __future__ import annotations
+
+import csv
+import json
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parent
+DIRECTIONS = [
+    ("upper-left", (0, -1)),
+    ("upper-right", (1, -1)),
+    ("left", (-1, 0)),
+    ("right", (1, 0)),
+    ("lower-left", (-1, 1)),
+    ("lower-right", (0, 1)),
+]
+PUBLIC_COLUMNS = ["question_id", "image", "prompt"]
+ANSWER_COLUMNS = [
+    "question_id", "image", "target", "home_on_boundary", "neighbour_count",
+    "holes_touching", "walkable_touching", "hole_directions",
+]
+PROMPT = (
+    "Look closely at the green HOME hex and everything that immediately surrounds it: state whether it sits on "
+    "the outer boundary of the hexagonal field or fully inside it, count how many hexes lie directly against it, "
+    "and say how many of those are grey holes and how many are walkable. Name the position of any hole touching it "
+    "using these six direction names only: upper-left, upper-right, left, right, lower-left, lower-right. Justify "
+    "your answer by describing exactly what you see around it, and end with a confidence score from 0 to 1."
+)
+
+
+def read_records() -> list[dict]:
+    return [json.loads(line) for line in (ROOT / "annotations.jsonl").read_text(encoding="utf-8").splitlines() if line]
+
+
+def derive(record: dict) -> tuple[dict, dict, dict]:
+    tiles = {tuple(tile["coordinate"]): tile["color"] for tile in record["all_tiles"]}
+    home = tuple(record["home_coordinate"])
+    neighbours = []
+    for name, (dq, dr) in DIRECTIONS:
+        coordinate = (home[0] + dq, home[1] + dr)
+        if coordinate in tiles:
+            neighbours.append({"direction": name, "coordinate": list(coordinate), "color": tiles[coordinate]})
+    holes = [item for item in neighbours if item["color"] == "grey"]
+    hole_directions = [name for name, _ in DIRECTIONS if any(item["direction"] == name for item in holes)]
+    question_id = f"{record['id']}_open_q1"
+    public = {"question_id": question_id, "image": Path(record["image_path"]).name, "prompt": PROMPT}
+    answer = {
+        "question_id": question_id,
+        "image": Path(record["image_path"]).name,
+        "target": "HOME",
+        "home_on_boundary": len(neighbours) < 6,
+        "neighbour_count": len(neighbours),
+        "holes_touching": len(holes),
+        "walkable_touching": len(neighbours) - len(holes),
+        "hole_directions": json.dumps(hole_directions, separators=(",", ":")),
+    }
+    annotation = {
+        **public,
+        **{key: value for key, value in answer.items() if key not in {"question_id", "image"}},
+        "dataset_version": record.get("dataset_version"),
+        "scoring": {
+            "separate_exact_match_fields": [
+                "home_on_boundary", "neighbour_count", "holes_touching", "walkable_touching", "hole_directions"
+            ],
+            "confidence_range": [0, 1],
+        },
+        "derivation": {
+            "home_coordinate": list(home),
+            "direction_vocabulary": [name for name, _ in DIRECTIONS],
+            "in_grid_neighbours": neighbours,
+            "walkable_definition": "any in-grid neighbor whose stored color is not grey",
+        },
+    }
+    return public, answer, annotation
+
+
+def write_csv(path: Path, columns: list[str], rows: list[dict]) -> None:
+    with path.open("w", encoding="utf-8-sig", newline="") as handle:
+        writer = csv.DictWriter(handle, fieldnames=columns, lineterminator="\n")
+        writer.writeheader()
+        writer.writerows(rows)
+
+
+def main() -> None:
+    public_rows, answer_rows, annotations = [], [], []
+    for record in read_records():
+        public, answer, annotation = derive(record)
+        public_rows.append(public)
+        answer_rows.append(answer)
+        annotations.append(annotation)
+    if len(public_rows) != 3000:
+        raise RuntimeError(f"Expected 3,000 records, found {len(public_rows)}")
+    write_csv(ROOT / "open_questions.csv", PUBLIC_COLUMNS, public_rows)
+    write_csv(ROOT / "open_answer_key.csv", ANSWER_COLUMNS, answer_rows)
+    with (ROOT / "open_annotations.jsonl").open("w", encoding="utf-8", newline="\n") as handle:
+        for row in annotations:
+            handle.write(json.dumps(row, ensure_ascii=False, sort_keys=True, separators=(",", ":")) + "\n")
+    print(f"Wrote {len(public_rows)} hex open questions")
+
+
+if __name__ == "__main__":
+    main()

Dataset/hex_pathfinding_dataset_3000/generate_hex_pathfinding_dataset.py

@@ -5,7 +5,7 @@ from collections import Counter,deque
 from pathlib import Path
 from PIL import Image,ImageDraw,ImageFont
 BACKGROUND="#FDFAF4";WHITE="#FFFDF9";BLACK="#242A2E";GREY="#9DA3A8";HOME="#4A9B63";START="#D24A43";OUTLINE="#314958";L2="#2C78B8";L5="#8A4FA3"
-DATASET_VERSION="hex-pathfinding-2.0.0"
+DATASET_VERSION="hex-pathfinding-3.0.0"
 DIRECTIONS=((1,0),(1,-1),(0,-1),(-1,0),(-1,1),(0,1))
 def font(size,bold=False):
  try:return ImageFont.truetype(str(Path("C:/Windows/Fonts")/("arialbd.ttf" if bold else "arial.ttf")),size)

Dataset/hex_pathfinding_dataset_3000/open_validation_metrics.json

@@ -0,0 +1,115 @@
+{
+  "answer_distributions": {
+    "hole_directions": {
+      "[\"left\", \"lower-left\", \"lower-right\"]": 4,
+      "[\"left\", \"lower-left\"]": 30,
+      "[\"left\", \"lower-right\"]": 41,
+      "[\"left\", \"right\", \"lower-left\", \"lower-right\"]": 2,
+      "[\"left\", \"right\", \"lower-left\"]": 4,
+      "[\"left\", \"right\", \"lower-right\"]": 2,
+      "[\"left\", \"right\"]": 35,
+      "[\"left\"]": 183,
+      "[\"lower-left\", \"lower-right\"]": 37,
+      "[\"lower-left\"]": 213,
+      "[\"lower-right\"]": 205,
+      "[\"right\", \"lower-left\", \"lower-right\"]": 3,
+      "[\"right\", \"lower-left\"]": 33,
+      "[\"right\", \"lower-right\"]": 44,
+      "[\"right\"]": 193,
+      "[\"upper-left\", \"left\", \"lower-left\"]": 7,
+      "[\"upper-left\", \"left\", \"lower-right\"]": 3,
+      "[\"upper-left\", \"left\", \"right\", \"lower-right\"]": 1,
+      "[\"upper-left\", \"left\", \"right\"]": 7,
+      "[\"upper-left\", \"left\"]": 34,
+      "[\"upper-left\", \"lower-left\", \"lower-right\"]": 5,
+      "[\"upper-left\", \"lower-left\"]": 26,
+      "[\"upper-left\", \"lower-right\"]": 33,
+      "[\"upper-left\", \"right\", \"lower-left\"]": 7,
+      "[\"upper-left\", \"right\", \"lower-right\"]": 2,
+      "[\"upper-left\", \"right\"]": 24,
+      "[\"upper-left\", \"upper-right\", \"left\", \"lower-left\"]": 1,
+      "[\"upper-left\", \"upper-right\", \"left\"]": 3,
+      "[\"upper-left\", \"upper-right\", \"lower-left\", \"lower-right\"]": 2,
+      "[\"upper-left\", \"upper-right\", \"lower-left\"]": 7,
+      "[\"upper-left\", \"upper-right\", \"lower-right\"]": 3,
+      "[\"upper-left\", \"upper-right\", \"right\", \"lower-right\"]": 1,
+      "[\"upper-left\", \"upper-right\", \"right\"]": 6,
+      "[\"upper-left\", \"upper-right\"]": 30,
+      "[\"upper-left\"]": 173,
+      "[\"upper-right\", \"left\", \"lower-left\"]": 3,
+      "[\"upper-right\", \"left\", \"lower-right\"]": 4,
+      "[\"upper-right\", \"left\", \"right\", \"lower-left\"]": 1,
+      "[\"upper-right\", \"left\", \"right\"]": 9,
+      "[\"upper-right\", \"left\"]": 27,
+      "[\"upper-right\", \"lower-left\", \"lower-right\"]": 7,
+      "[\"upper-right\", \"lower-left\"]": 28,
+      "[\"upper-right\", \"lower-right\"]": 32,
+      "[\"upper-right\", \"right\", \"lower-left\", \"lower-right\"]": 1,
+      "[\"upper-right\", \"right\", \"lower-left\"]": 5,
+      "[\"upper-right\", \"right\", \"lower-right\"]": 5,
+      "[\"upper-right\", \"right\"]": 27,
+      "[\"upper-right\"]": 190,
+      "[]": 1257
+    },
+    "holes_touching": {
+      "0": 1257,
+      "1": 1157,
+      "2": 481,
+      "3": 96,
+      "4": 9
+    },
+    "home_on_boundary": {
+      "False": 1394,
+      "True": 1606
+    },
+    "neighbour_count": {
+      "3": 475,
+      "4": 1131,
+      "6": 1394
+    },
+    "walkable_touching": {
+      "1": 43,
+      "2": 364,
+      "3": 749,
+      "4": 837,
+      "5": 538,
+      "6": 469
+    }
+  },
+  "boundary_vs_interior": {
+    "boundary": 1606,
+    "interior": 1394
+  },
+  "constant_answer_baselines": {
+    "hole_directions": 0.419,
+    "holes_touching": 0.419,
+    "home_on_boundary": 0.5353333333333333,
+    "neighbour_count": 0.4646666666666667,
+    "walkable_touching": 0.279
+  },
+  "derivation_mismatches": [],
+  "fields_over_60_percent": {},
+  "issues": [],
+  "no_answer_leakage": true,
+  "no_coordinate_prompts": true,
+  "no_existing_question_paraphrase": true,
+  "permitted_directions": [
+    "left",
+    "lower-left",
+    "lower-right",
+    "right",
+    "upper-left",
+    "upper-right"
+  ],
+  "png_recoverability": {
+    "passed": 3000,
+    "total": 3000
+  },
+  "public_columns": [
+    "question_id",
+    "image",
+    "prompt"
+  ],
+  "question_count": 3000,
+  "template_choice": "retained neighbour_count because HOME is boundary in 1,606 items and interior in 1,394"
+}

Dataset/hex_pathfinding_dataset_3000/open_validation_report.txt

@@ -0,0 +1,11 @@
+HEX PATHFINDING OPEN-QUESTION VALIDATION REPORT
+================================================
+Questions: 3000
+Derivation mismatches: 0
+PNG-recoverable items: 3000/3000
+Boundary/interior: {'boundary': 1606, 'interior': 1394}
+Constant baselines: {'home_on_boundary': 0.5353333333333333, 'neighbour_count': 0.4646666666666667, 'holes_touching': 0.419, 'walkable_touching': 0.279, 'hole_directions': 0.419}
+Fields over 60%: {}
+Issues: 0
+None
+Summary: PASS

Dataset/hex_pathfinding_dataset_3000/validate_open_questions.py

@@ -0,0 +1,175 @@
+"""Independently validate HOME-neighborhood open questions and PNG evidence."""
+from __future__ import annotations
+
+import csv
+import json
+import math
+import re
+from collections import Counter
+from pathlib import Path
+
+from PIL import Image
+
+ROOT = Path(__file__).resolve().parent
+PUBLIC_COLUMNS = ["question_id", "image", "prompt"]
+ANSWER_COLUMNS = [
+    "question_id", "image", "target", "home_on_boundary", "neighbour_count",
+    "holes_touching", "walkable_touching", "hole_directions",
+]
+DIRECTIONS = [
+    ("upper-left", (0, -1)), ("upper-right", (1, -1)), ("left", (-1, 0)),
+    ("right", (1, 0)), ("lower-left", (-1, 1)), ("lower-right", (0, 1)),
+]
+PALETTE = {"white": (255, 253, 249), "black": (36, 42, 46), "grey": (157, 163, 168)}
+
+
+def read_csv(path: Path) -> list[dict[str, str]]:
+    with path.open(encoding="utf-8-sig", newline="") as handle:
+        return list(csv.DictReader(handle))
+
+
+def derive(record: dict) -> dict:
+    tiles = {tuple(tile["coordinate"]): tile["color"] for tile in record["all_tiles"]}
+    home = tuple(record["home_coordinate"])
+    neighbours = []
+    for name, (dq, dr) in DIRECTIONS:
+        coordinate = (home[0] + dq, home[1] + dr)
+        if coordinate in tiles:
+            neighbours.append((name, coordinate, tiles[coordinate]))
+    holes = [item for item in neighbours if item[2] == "grey"]
+    return {
+        "home_on_boundary": len(neighbours) < 6,
+        "neighbour_count": len(neighbours),
+        "holes_touching": len(holes),
+        "walkable_touching": len(neighbours) - len(holes),
+        "hole_directions": [name for name, _ in DIRECTIONS if any(item[0] == name for item in holes)],
+        "neighbours": neighbours,
+    }
+
+
+def pixel_geometry(canvas: int, radius: int) -> tuple[float, tuple[float, float]]:
+    size = min((canvas - 72) / (math.sqrt(3) * (2 * radius + 1)), (canvas - 112) / (3 * radius + 2))
+    return size, (canvas / 2, canvas / 2 - 22)
+
+
+def to_pixel(coordinate: tuple[int, int], size: float, center: tuple[float, float]) -> tuple[float, float]:
+    q, r = coordinate
+    return center[0] + size * math.sqrt(3) * (q + r / 2), center[1] + size * 1.5 * r
+
+
+def recover_class(image: Image.Image, coordinate: tuple[int, int], size: float, center: tuple[float, float]) -> str:
+    x0, y0 = to_pixel(coordinate, size, center)
+    samples = []
+    for angle in range(0, 360, 30):
+        radians = math.radians(angle)
+        x, y = round(x0 + .65 * size * math.cos(radians)), round(y0 + .65 * size * math.sin(radians))
+        samples.append(image.getpixel((x, y)))
+    votes = Counter(min(PALETTE, key=lambda name: sum((pixel[index] - PALETTE[name][index]) ** 2 for index in range(3))) for pixel in samples)
+    return votes.most_common(1)[0][0]
+
+
+def distribution(values) -> dict[str, int]:
+    return dict(sorted(Counter(json.dumps(value, ensure_ascii=False) if isinstance(value, list) else str(value) for value in values).items()))
+
+
+def baseline(counts: dict[str, int]) -> float:
+    return max(counts.values()) / sum(counts.values()) if counts else 0.0
+
+
+def main() -> None:
+    records = [json.loads(line) for line in (ROOT / "annotations.jsonl").read_text(encoding="utf-8").splitlines() if line]
+    public = read_csv(ROOT / "open_questions.csv")
+    answers = read_csv(ROOT / "open_answer_key.csv")
+    annotations = [json.loads(line) for line in (ROOT / "open_annotations.jsonl").read_text(encoding="utf-8").splitlines() if line]
+    issues, derivation_mismatches = [], []
+    png_pass = 0
+    fields = ["home_on_boundary", "neighbour_count", "holes_touching", "walkable_touching", "hole_directions"]
+    values = {key: [] for key in fields}
+    if len(records) != len(public) or len(public) != len(answers) or len(answers) != len(annotations):
+        issues.append("row-count mismatch among source/open files")
+    if public and list(public[0]) != PUBLIC_COLUMNS:
+        issues.append(f"public columns {list(public[0])} != {PUBLIC_COLUMNS}")
+    if answers and list(answers[0]) != ANSWER_COLUMNS:
+        issues.append(f"answer columns {list(answers[0])} != {ANSWER_COLUMNS}")
+    permitted = {name for name, _ in DIRECTIONS}
+    for position, (record, question, answer, annotation) in enumerate(zip(records, public, answers, annotations), 1):
+        expected = derive(record)
+        parsed = {
+            "home_on_boundary": answer["home_on_boundary"].lower() == "true",
+            "neighbour_count": int(answer["neighbour_count"]),
+            "holes_touching": int(answer["holes_touching"]),
+            "walkable_touching": int(answer["walkable_touching"]),
+            "hole_directions": json.loads(answer["hole_directions"]),
+        }
+        for key in fields:
+            if parsed[key] != expected[key]:
+                derivation_mismatches.append(f"{record['id']}/{key}")
+            values[key].append(expected[key])
+        qid = f"{record['id']}_open_q1"
+        if question["question_id"] != qid or answer["question_id"] != qid or annotation["question_id"] != qid:
+            issues.append(f"{record['id']}: question_id mismatch")
+        if question["image"] != Path(record["image_path"]).name or answer["image"] != question["image"]:
+            issues.append(f"{record['id']}: image mismatch")
+        prompt = question["prompt"]
+        if re.search(r"\(\s*-?\d+\s*,\s*-?\d+\s*\)", prompt) or any(word in prompt.lower() for word in ["axial", "coordinate", "q,r", "q, r"]):
+            issues.append(f"{record['id']}: prompt contains a coordinate scheme")
+        if "start" in prompt.lower() or "outlined in blue" in prompt.lower():
+            issues.append(f"{record['id']}: prompt restates Level 2 START-neighbor task")
+        if any(prompt.strip() == old["question_text"].strip() for old in record["questions"]):
+            issues.append(f"{record['id']}: exact existing-question paraphrase")
+        directions_in_prompt = {name for name in permitted if name in prompt.lower()}
+        if directions_in_prompt != permitted or not set(expected["hole_directions"]) <= permitted:
+            issues.append(f"{record['id']}: invalid or incomplete direction vocabulary")
+        if annotation.get("derivation", {}).get("home_coordinate") != record["home_coordinate"]:
+            issues.append(f"{record['id']}: annotation derivation mismatch")
+        with Image.open(ROOT / record["image_path"]) as source:
+            image = source.convert("RGB")
+            image.load()
+        size, center = pixel_geometry(record["canvas_size"][0], record["grid_radius"])
+        png_ok = image.size == tuple(record["canvas_size"])
+        for _direction, coordinate, stored_color in expected["neighbours"]:
+            recovered = recover_class(image, coordinate, size, center)
+            wanted = "grey" if stored_color == "grey" else ("black" if stored_color == "black" else "white")
+            if recovered != wanted:
+                png_ok = False
+        if png_ok:
+            png_pass += 1
+        else:
+            issues.append(f"{record['id']}: PNG HOME-neighborhood recovery")
+        if position % 500 == 0:
+            print(f"Validated {position}/{len(records)}", flush=True)
+    if derivation_mismatches:
+        issues.append(f"independent derivation mismatches: {len(derivation_mismatches)}")
+    distributions = {key: distribution(field_values) for key, field_values in values.items()}
+    baselines = {key: baseline(counts) for key, counts in distributions.items()}
+    metrics = {
+        "question_count": len(public),
+        "template_choice": "retained neighbour_count because HOME is boundary in 1,606 items and interior in 1,394",
+        "boundary_vs_interior": {"boundary": values["home_on_boundary"].count(True), "interior": values["home_on_boundary"].count(False)},
+        "answer_distributions": distributions,
+        "constant_answer_baselines": baselines,
+        "fields_over_60_percent": {key: value for key, value in baselines.items() if value > .60},
+        "derivation_mismatches": derivation_mismatches,
+        "png_recoverability": {"passed": png_pass, "total": len(records)},
+        "public_columns": PUBLIC_COLUMNS,
+        "permitted_directions": sorted(permitted),
+        "no_coordinate_prompts": not any("coordinate scheme" in issue for issue in issues),
+        "no_answer_leakage": not any("leaks" in issue for issue in issues),
+        "no_existing_question_paraphrase": not any("paraphrase" in issue or "Level 2" in issue for issue in issues),
+        "issues": issues,
+    }
+    (ROOT / "open_validation_metrics.json").write_text(json.dumps(metrics, indent=2, ensure_ascii=False, sort_keys=True) + "\n", encoding="utf-8")
+    report = [
+        "HEX PATHFINDING OPEN-QUESTION VALIDATION REPORT", "=" * 48,
+        f"Questions: {len(public)}", f"Derivation mismatches: {len(derivation_mismatches)}",
+        f"PNG-recoverable items: {png_pass}/{len(records)}", f"Boundary/interior: {metrics['boundary_vs_interior']}",
+        f"Constant baselines: {baselines}", f"Fields over 60%: {metrics['fields_over_60_percent']}",
+        f"Issues: {len(issues)}", *(issues if issues else ["None"]), f"Summary: {'PASS' if not issues else 'FAIL'}",
+    ]
+    (ROOT / "open_validation_report.txt").write_text("\n".join(report) + "\n", encoding="utf-8")
+    print("\n".join(report))
+    raise SystemExit(1 if issues else 0)
+
+
+if __name__ == "__main__":
+    main()

Dataset/open_question_prechange_hashes.json

@@ -0,0 +1,18 @@
+{
+  "hex_pathfinding_dataset_3000": {
+    "answer_key_sha256": "6dddbeddd760b86d240ed400b715d8a98973ac9e63525179b93b513c868dcdc6",
+    "dataset_final_csv_sha256": "ea8259602ed80d6a4ec4c9f1ceeaa6a7d03b323fe262b78f752290f267dd9d96",
+    "dataset_final_jsonl_sha256": "b2ceae336c0c8e86f5c2227164fc4ae314739b07765a427ac5f88adf7b39eb6f",
+    "existing_questions_semantic_sha256": "2442dbe89dd9e242e58c7a58b8d0a0e71ab3c13a10754cc0d89d51a2fe6bd041",
+    "image_tree_git_oid": "14e67e7b1f49c42c3beaaa78e20de230746a9395",
+    "question_set_sha256": "5f9fb5f503e642376e109a408bbaf12815515e30d5d489b88e82397eb58c6154"
+  },
+  "route_dataset_3000": {
+    "answer_key_sha256": "ab9fb431b5bbb44953b0f1b323b693ef0793defa14aae859aff310a6273572e4",
+    "dataset_final_csv_sha256": "909a9801222e694903332c3ebacf6633501f821d988df458ee67efae0db9f7df",
+    "dataset_final_jsonl_sha256": "b550367f58d6850656900d592c0e491f360e90201ea4e3a0cd76e10efdc59bc3",
+    "existing_questions_semantic_sha256": "4ea557a6a3feb111d4d5309a70364fd5aa7a9c2810f9cb739b717012ed053d6d",
+    "image_tree_git_oid": "cbe959b68ae060fc5eb6bc4c290988e18992a8c9",
+    "question_set_sha256": "62d9122b24f374ac2aeb96b10f9c49fa50e811fc0205427eb31afe074ba90876"
+  }
+}

Dataset/open_question_release_report.json

@@ -0,0 +1,234 @@
+{
+  "status": "PASS",
+  "scope": [
+    "route_dataset_3000",
+    "hex_pathfinding_dataset_3000"
+  ],
+  "domains": {
+    "route_dataset_3000": {
+      "dataset_version": "route-3.0.0",
+      "open_questions": 3000,
+      "open_answers": 3000,
+      "derivation_mismatches": 0,
+      "png_recoverability": {
+        "passed": 3000,
+        "total": 3000
+      },
+      "constant_answer_baselines": {
+        "answer": 0.913,
+        "colors": 0.006666666666666667,
+        "lines_at_target": 0.502,
+        "reached": 0.06833333333333333,
+        "target": 0.212,
+        "target_degree": 0.502,
+        "unreached": 0.087
+      },
+      "fields_over_60_percent": {
+        "answer": 0.913
+      },
+      "protected_file_hashes_unchanged": {
+        "question_set.csv": true,
+        "answer_key.csv": true,
+        "dataset_final.csv": true,
+        "dataset_final.jsonl": true
+      },
+      "annotation_payload_unchanged_except_version": true,
+      "images_unchanged": true,
+      "selection_distribution": {
+        "target": {
+          "A": 609,
+          "B": 582,
+          "C": 630,
+          "D": 636,
+          "E": 397,
+          "F": 146
+        },
+        "target_degree": {
+          "1": 93,
+          "2": 1090,
+          "3": 1506,
+          "4": 247,
+          "5": 63,
+          "6": 1
+        }
+      },
+      "connected_label_pairs_out_of_canonical_15": {
+        "2": 4,
+        "3": 96,
+        "4": 405,
+        "5": 809,
+        "6": 755,
+        "7": 484,
+        "8": 301,
+        "9": 119,
+        "10": 26,
+        "11": 1
+      }
+    },
+    "hex_pathfinding_dataset_3000": {
+      "dataset_version": "hex-pathfinding-3.0.0",
+      "open_questions": 3000,
+      "open_answers": 3000,
+      "derivation_mismatches": 0,
+      "png_recoverability": {
+        "passed": 3000,
+        "total": 3000
+      },
+      "constant_answer_baselines": {
+        "hole_directions": 0.419,
+        "holes_touching": 0.419,
+        "home_on_boundary": 0.5353333333333333,
+        "neighbour_count": 0.4646666666666667,
+        "walkable_touching": 0.279
+      },
+      "fields_over_60_percent": {},
+      "protected_file_hashes_unchanged": {
+        "question_set.csv": true,
+        "answer_key.csv": true,
+        "dataset_final.csv": true,
+        "dataset_final.jsonl": true
+      },
+      "annotation_payload_unchanged_except_version": true,
+      "images_unchanged": true,
+      "boundary_vs_interior": {
+        "boundary": 1606,
+        "interior": 1394
+      },
+      "template_choice": "retained neighbour_count because HOME is boundary in 1,606 items and interior in 1,394",
+      "answer_distributions": {
+        "hole_directions": {
+          "[\"left\", \"lower-left\", \"lower-right\"]": 4,
+          "[\"left\", \"lower-left\"]": 30,
+          "[\"left\", \"lower-right\"]": 41,
+          "[\"left\", \"right\", \"lower-left\", \"lower-right\"]": 2,
+          "[\"left\", \"right\", \"lower-left\"]": 4,
+          "[\"left\", \"right\", \"lower-right\"]": 2,
+          "[\"left\", \"right\"]": 35,
+          "[\"left\"]": 183,
+          "[\"lower-left\", \"lower-right\"]": 37,
+          "[\"lower-left\"]": 213,
+          "[\"lower-right\"]": 205,
+          "[\"right\", \"lower-left\", \"lower-right\"]": 3,
+          "[\"right\", \"lower-left\"]": 33,
+          "[\"right\", \"lower-right\"]": 44,
+          "[\"right\"]": 193,
+          "[\"upper-left\", \"left\", \"lower-left\"]": 7,
+          "[\"upper-left\", \"left\", \"lower-right\"]": 3,
+          "[\"upper-left\", \"left\", \"right\", \"lower-right\"]": 1,
+          "[\"upper-left\", \"left\", \"right\"]": 7,
+          "[\"upper-left\", \"left\"]": 34,
+          "[\"upper-left\", \"lower-left\", \"lower-right\"]": 5,
+          "[\"upper-left\", \"lower-left\"]": 26,
+          "[\"upper-left\", \"lower-right\"]": 33,
+          "[\"upper-left\", \"right\", \"lower-left\"]": 7,
+          "[\"upper-left\", \"right\", \"lower-right\"]": 2,
+          "[\"upper-left\", \"right\"]": 24,
+          "[\"upper-left\", \"upper-right\", \"left\", \"lower-left\"]": 1,
+          "[\"upper-left\", \"upper-right\", \"left\"]": 3,
+          "[\"upper-left\", \"upper-right\", \"lower-left\", \"lower-right\"]": 2,
+          "[\"upper-left\", \"upper-right\", \"lower-left\"]": 7,
+          "[\"upper-left\", \"upper-right\", \"lower-right\"]": 3,
+          "[\"upper-left\", \"upper-right\", \"right\", \"lower-right\"]": 1,
+          "[\"upper-left\", \"upper-right\", \"right\"]": 6,
+          "[\"upper-left\", \"upper-right\"]": 30,
+          "[\"upper-left\"]": 173,
+          "[\"upper-right\", \"left\", \"lower-left\"]": 3,
+          "[\"upper-right\", \"left\", \"lower-right\"]": 4,
+          "[\"upper-right\", \"left\", \"right\", \"lower-left\"]": 1,
+          "[\"upper-right\", \"left\", \"right\"]": 9,
+          "[\"upper-right\", \"left\"]": 27,
+          "[\"upper-right\", \"lower-left\", \"lower-right\"]": 7,
+          "[\"upper-right\", \"lower-left\"]": 28,
+          "[\"upper-right\", \"lower-right\"]": 32,
+          "[\"upper-right\", \"right\", \"lower-left\", \"lower-right\"]": 1,
+          "[\"upper-right\", \"right\", \"lower-left\"]": 5,
+          "[\"upper-right\", \"right\", \"lower-right\"]": 5,
+          "[\"upper-right\", \"right\"]": 27,
+          "[\"upper-right\"]": 190,
+          "[]": 1257
+        },
+        "holes_touching": {
+          "0": 1257,
+          "1": 1157,
+          "2": 481,
+          "3": 96,
+          "4": 9
+        },
+        "home_on_boundary": {
+          "False": 1394,
+          "True": 1606
+        },
+        "neighbour_count": {
+          "3": 475,
+          "4": 1131,
+          "6": 1394
+        },
+        "walkable_touching": {
+          "1": 43,
+          "2": 364,
+          "3": 749,
+          "4": 837,
+          "5": 538,
+          "6": 469
+        }
+      }
+    }
+  },
+  "combined_open": {
+    "status": "PASS",
+    "datasets": {
+      "hex_pathfinding_dataset_3000": {
+        "dataset_version": "hex-pathfinding-3.0.0",
+        "images": 3000,
+        "open_questions": 3000,
+        "open_answers": 3000
+      },
+      "route_dataset_3000": {
+        "dataset_version": "route-3.0.0",
+        "images": 3000,
+        "open_questions": 3000,
+        "open_answers": 3000
+      }
+    },
+    "combined_open_questions": 6000,
+    "combined_open_answers": 6000,
+    "unique_question_ids": 6000,
+    "resolved_image_paths": 6000,
+    "public_columns": [
+      "dataset",
+      "dataset_version",
+      "question_id",
+      "image",
+      "image_path",
+      "prompt"
+    ],
+    "private_columns": [
+      "dataset",
+      "dataset_version",
+      "question_id",
+      "image",
+      "image_path",
+      "prompt",
+      "target",
+      "target_position",
+      "target_degree",
+      "lines_at_target",
+      "colors",
+      "reached",
+      "unreached",
+      "answer",
+      "home_on_boundary",
+      "neighbour_count",
+      "holes_touching",
+      "walkable_touching",
+      "hole_directions"
+    ]
+  },
+  "five_level_suite": {
+    "datasets": 34,
+    "images": 100000,
+    "questions": 500000,
+    "answers": 500000
+  },
+  "protected_content": "images, renderers, existing five-level question text, and existing answers unchanged"
+}

Dataset/rebuild_combined_open_questions.py

@@ -0,0 +1,121 @@
+"""Build the separate suite-level supplementary open-question CSVs."""
+from __future__ import annotations
+
+import csv
+import json
+from collections import Counter
+from pathlib import Path
+
+
+DATASET_ROOT = Path(__file__).resolve().parent
+REPO_ROOT = DATASET_ROOT.parent
+COMBINED_ROOT = REPO_ROOT / "combined"
+PUBLIC_OUT = COMBINED_ROOT / "all_open_questions_combined.csv"
+PRIVATE_OUT = COMBINED_ROOT / "all_open_answers_combined.csv"
+REPORT_OUT = DATASET_ROOT / "combined_open_question_report.json"
+
+SOURCE_PUBLIC_COLUMNS = ["question_id", "image", "prompt"]
+PUBLIC_COLUMNS = ["dataset", "dataset_version", "question_id", "image", "image_path", "prompt"]
+PRIVATE_COLUMNS = PUBLIC_COLUMNS + [
+    "target", "target_position", "target_degree", "lines_at_target", "colors",
+    "reached", "unreached", "answer", "home_on_boundary", "neighbour_count",
+    "holes_touching", "walkable_touching", "hole_directions",
+]
+
+
+def read_csv(path: Path) -> tuple[list[str], list[dict[str, str]]]:
+    if not path.is_file():
+        raise FileNotFoundError(path)
+    with path.open(encoding="utf-8-sig", newline="") as handle:
+        reader = csv.DictReader(handle)
+        return list(reader.fieldnames or []), list(reader)
+
+
+def write_csv(path: Path, rows: list[dict[str, str]], columns: list[str]) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    with path.open("w", encoding="utf-8-sig", newline="") as handle:
+        writer = csv.DictWriter(handle, fieldnames=columns, lineterminator="\n")
+        writer.writeheader()
+        writer.writerows({column: row.get(column, "") for column in columns} for row in rows)
+
+
+def main() -> None:
+    discovered: list[Path] = []
+    incomplete: list[str] = []
+    for folder in sorted(path for path in DATASET_ROOT.iterdir() if path.is_dir()):
+        public_exists = (folder / "open_questions.csv").is_file()
+        private_exists = (folder / "open_answer_key.csv").is_file()
+        if public_exists != private_exists:
+            incomplete.append(folder.name)
+        elif public_exists:
+            discovered.append(folder)
+    if incomplete:
+        raise RuntimeError(f"Incomplete open-question file pairs: {incomplete}")
+    expected = {"route_dataset_3000", "hex_pathfinding_dataset_3000"}
+    if {folder.name for folder in discovered} != expected:
+        raise RuntimeError(f"Expected open sets for {sorted(expected)}, found {[f.name for f in discovered]}")
+
+    public_rows: list[dict[str, str]] = []
+    private_rows: list[dict[str, str]] = []
+    per_dataset: dict[str, dict[str, int | str]] = {}
+    for folder in discovered:
+        public_header, public = read_csv(folder / "open_questions.csv")
+        private_header, private = read_csv(folder / "open_answer_key.csv")
+        if public_header != SOURCE_PUBLIC_COLUMNS:
+            raise RuntimeError(f"{folder.name}: public schema is {public_header}")
+        if len(public) != 3000 or len(private) != 3000:
+            raise RuntimeError(f"{folder.name}: expected 3000 public/private rows")
+        private_by_id = {row["question_id"]: row for row in private}
+        if len(private_by_id) != len(private):
+            raise RuntimeError(f"{folder.name}: duplicate private question_id")
+        manifest = json.loads((folder / "build_manifest.json").read_text(encoding="utf-8"))
+        version = str(manifest["dataset_version"])
+        dataset = folder.name.rsplit("_dataset_", 1)[0]
+        for row in public:
+            qid = row["question_id"]
+            answer = private_by_id.get(qid)
+            if answer is None:
+                raise RuntimeError(f"{folder.name}/{qid}: no unique answer row")
+            if any(answer.get(key, "") != row[key] for key in ("question_id", "image")):
+                raise RuntimeError(f"{folder.name}/{qid}: public/private row drift")
+            image_path = (Path("Dataset") / folder.name / "images" / row["image"]).as_posix()
+            if not (REPO_ROOT / image_path).is_file():
+                raise RuntimeError(f"{folder.name}/{qid}: missing image {image_path}")
+            base = {
+                "dataset": dataset,
+                "dataset_version": version,
+                **row,
+                "image_path": image_path,
+            }
+            public_rows.append(base)
+            private_rows.append({**base, **{key: answer.get(key, "") for key in private_header[2:]}})
+        per_dataset[folder.name] = {
+            "dataset_version": version,
+            "images": len(public),
+            "open_questions": len(public),
+            "open_answers": len(private),
+        }
+
+    duplicate_ids = [qid for qid, count in Counter(row["question_id"] for row in public_rows).items() if count != 1]
+    if duplicate_ids:
+        raise RuntimeError(f"Combined open question IDs are not unique: {duplicate_ids[:5]}")
+    if len(public_rows) != 6000 or len(private_rows) != 6000:
+        raise RuntimeError("Combined open row count must be exactly 6000")
+    write_csv(PUBLIC_OUT, public_rows, PUBLIC_COLUMNS)
+    write_csv(PRIVATE_OUT, private_rows, PRIVATE_COLUMNS)
+    report = {
+        "status": "PASS",
+        "datasets": per_dataset,
+        "combined_open_questions": len(public_rows),
+        "combined_open_answers": len(private_rows),
+        "unique_question_ids": len({row["question_id"] for row in public_rows}),
+        "resolved_image_paths": len({row["image_path"] for row in public_rows}),
+        "public_columns": PUBLIC_COLUMNS,
+        "private_columns": PRIVATE_COLUMNS,
+    }
+    REPORT_OUT.write_text(json.dumps(report, indent=2) + "\n", encoding="utf-8")
+    print(json.dumps(report, indent=2))
+
+
+if __name__ == "__main__":
+    main()

Dataset/route_dataset_3000/README.md

@@ -2,7 +2,7 @@
 
 ## 1. Dataset overview
 
-The dataset contains **3,000 images** and **12,000 questions**. It was generated on **2026-08-10** through a fully synthetic, deterministic Python pipeline; images were not human-drawn and labels were not human-annotated.
+Current version: `route-3.0.0`. The dataset contains **3,000 images**, **15,000 five-level questions**, and **3,000 supplementary open-ended questions**. It was generated through a fully synthetic, deterministic Python pipeline; images were not human-drawn and labels were not human-annotated.
 
 ## 2. Full verbatim generation prompt
 
@@ -305,3 +305,16 @@ Route Counting,route_puzzle_0001.png,Count the one-colored routes that go from A
 Comparative Reasoning,route_puzzle_0001.png,How many routes does letter A have in total (i.e. touching A)?,5,"{""colors_used"":[""orange"",""amber"",""olive"",""cyan"",""brown"",""magenta"",""gray-blue"",""teal""],""crossing_count"":83,""difficulty_score"":0.4971,""line_width_px"":2.95,""num_endpoints"":4,""num_routes"":8,""seed"":1}"
 Compound Reasoning,route_puzzle_0001.png,"Is there any letter-pair with zero directly connecting routes? If yes, name one such pair; if no, answer 'none'.",CD,"{""colors_used"":[""orange"",""amber"",""olive"",""cyan"",""brown"",""magenta"",""gray-blue"",""teal""],""crossing_count"":83,""difficulty_score"":0.4971,""line_width_px"":2.95,""num_endpoints"":4,""num_routes"":8,""seed"":1}"
 ```
+
+### Supplementary open-ended questions (v3)
+
+Version `route-3.0.0` adds one open-ended route-tracing question per image without changing the existing images or five-level questions and answers. Target selection prefers an incident-line degree of 2 or 3, uses the stored seed to choose deterministically among qualifying labels, and otherwise selects among labels with the lowest non-zero degree. Because scenes contain four, five, or six endpoint labels, the common prompt inserts the actual number of other labels instead of assuming that every image has six.
+
+- `open_questions.csv` is public and contains exactly `question_id,image,prompt`.
+- `open_answer_key.csv` is answer-key-side and stores the target, target position, selected degree, incident-line count and colors, exact reached and unreached label sets, and overall conclusion.
+- `open_annotations.jsonl` is answer-key-side and adds the complete derivation, including per-label degrees and incident route paths. Its scoring declaration requires an exact reached set for the conclusion while retaining separate fields for partial-credit logging.
+- `build_open_questions.py` rebuilds the three open-question artifacts; `validate_open_questions.py` independently re-derives every answer and validates all 3,000 PNGs.
+
+The selected target-degree distribution is `{1: 93, 2: 1090, 3: 1506, 4: 247, 5: 63, 6: 1}`. The broad conclusion is structurally skewed: `some labelled sides remain unreached` occurs in 91.3% of images. Exact reached-set, unreached-set, incident-color, and incident-count fields are therefore the meaningful scoring components; their complete distributions and baselines are in `open_validation_metrics.json`. The connected-pair diagnostic ranges from 2 to 11 of the 15 canonical A–F pairs, and the worked `route_puzzle_0332` case has 7.
+
+Do not provide `open_answer_key.csv` or `open_annotations.jsonl` to a model under evaluation. The latter contains the stored path geometry and directly leaks the derivation.

Dataset/route_dataset_3000/build_manifest.json

@@ -1,9 +1,30 @@
 {
-  "constraint_source": "generator constants and validation metrics",
+  "constraint_set": {
+    "existing_five_level_questions_unchanged": true,
+    "no_coordinates_in_open_prompt": true,
+    "open_ground_truth_independently_rederived": true,
+    "open_png_recoverability_all_items": true,
+    "open_public_fields_exact": [
+      "question_id",
+      "image",
+      "prompt"
+    ],
+    "open_question_count_per_image": 1,
+    "route_colors_unique_and_rendered": true,
+    "target_has_positive_degree": true,
+    "target_selection_prefers_degree_2_or_3": true
+  },
+  "constraint_source": "generator constants and open-question validation metrics",
   "current_build_layout": "unsuffixed",
-  "dataset_version": "route-2.0.0",
+  "dataset_version": "route-3.0.0",
   "generator_commit": "e6975d7f82e6fc42f77fdc6434e733c10adb83d6",
   "images": 3000,
+  "open_question_files": [
+    "open_questions.csv",
+    "open_answer_key.csv",
+    "open_annotations.jsonl"
+  ],
+  "open_questions": 3000,
   "question_set_fields": [
     "question_id",
     "task",

Dataset/route_dataset_3000/build_open_questions.py

@@ -0,0 +1,172 @@
+"""Build one deterministic open-ended route-tracing question per image."""
+from __future__ import annotations
+
+import csv
+import json
+import random
+from collections import Counter
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parent
+PUBLIC_COLUMNS = ["question_id", "image", "prompt"]
+ANSWER_COLUMNS = [
+    "question_id", "image", "target", "target_position", "target_degree",
+    "lines_at_target", "colors", "reached", "unreached", "answer",
+]
+
+
+def read_records() -> list[dict]:
+    return [json.loads(line) for line in (ROOT / "annotations.jsonl").read_text(encoding="utf-8").splitlines() if line]
+
+
+def assign_endpoints(num_endpoints: int, canvas_size: list[int]) -> dict[str, dict]:
+    width, height = canvas_size
+    margin = 64
+    inner_width, inner_height = width - 2 * margin, height - 2 * margin
+    perimeter = 2 * (inner_width + inner_height)
+    start = inner_width / 2
+    result = {}
+    for index in range(num_endpoints):
+        distance = (start + index * perimeter / num_endpoints) % perimeter
+        letter = chr(65 + index)
+        if distance < inner_width:
+            point, side = (margin + distance, margin), "top"
+        elif distance < inner_width + inner_height:
+            point, side = (width - margin, margin + distance - inner_width), "right"
+        elif distance < 2 * inner_width + inner_height:
+            point, side = (width - margin - (distance - inner_width - inner_height), height - margin), "bottom"
+        else:
+            point, side = (margin, height - margin - (distance - 2 * inner_width - inner_height)), "left"
+        result[letter] = {"anchor": [round(point[0]), round(point[1])], "side": side}
+    return result
+
+
+def position_phrase(endpoint: dict, canvas_size: list[int]) -> str:
+    x, y = endpoint["anchor"]
+    width, height = canvas_size
+    side = endpoint["side"]
+    if side in {"top", "bottom"}:
+        horizontal = " left" if x < width * 0.38 else (" right" if x > width * 0.62 else "")
+        return f"at the {side}{horizontal}"
+    vertical = "upper " if y < height * 0.38 else ("lower " if y > height * 0.62 else "")
+    return f"on the {vertical}{side}" if vertical else f"on the {side}"
+
+
+def incident_routes(record: dict, target: str) -> list[dict]:
+    result = []
+    for index, route in enumerate(record["routes"]):
+        if target not in (route["start"], route["end"]):
+            continue
+        far_end = route["end"] if route["start"] == target else route["start"]
+        result.append({
+            "route_index": index,
+            "color": route["color"],
+            "start": route["start"],
+            "end": route["end"],
+            "far_end": far_end,
+            "num_bends": route["num_bends"],
+            "points": route["points"],
+        })
+    return result
+
+
+def select_target(record: dict) -> tuple[str, dict[str, int], list[str], str]:
+    degrees = {
+        letter: sum(letter in (route["start"], route["end"]) for route in record["routes"])
+        for letter in record["endpoint_letters"]
+    }
+    preferred = [letter for letter in record["endpoint_letters"] if degrees[letter] in (2, 3)]
+    if preferred:
+        candidates, rule = preferred, "seeded choice among degree-2-or-3 labels"
+    else:
+        positive = [value for value in degrees.values() if value >= 1]
+        if not positive:
+            raise ValueError(f"{record['id']}: no label has degree >= 1")
+        minimum = min(positive)
+        candidates = [letter for letter in record["endpoint_letters"] if degrees[letter] == minimum]
+        rule = "seeded choice among lowest non-zero-degree labels"
+    target = random.Random(int(record["seed"])).choice(candidates)
+    return target, degrees, candidates, rule
+
+
+def derive(record: dict) -> tuple[dict, dict, dict]:
+    target, degrees, candidates, selection_rule = select_target(record)
+    endpoints = assign_endpoints(record["num_endpoints"], record["canvas_size"])
+    position = position_phrase(endpoints[target], record["canvas_size"])
+    incident = incident_routes(record, target)
+    colors = [route["color"] for route in incident]
+    reached = sorted({route["far_end"] for route in incident if route["far_end"] != target})
+    unreached = sorted(set(record["endpoint_letters"]) - {target} - set(reached))
+    answer = "connected to every other labelled side" if not unreached else "some labelled sides remain unreached"
+    other_count = len(record["endpoint_letters"]) - 1
+    prompt = (
+        f"Trace the coloured lines that touch the label {target} {position} of the frame and follow each one through "
+        f"its bends to wherever it terminates: decide whether {target} is connected to every one of the other "
+        f"{other_count} labelled sides or whether some remain unreached from it, and name the label at the far end "
+        f"of each line that begins at {target}. State your conclusion, justify it by describing the paths you "
+        "followed and how you distinguished each line by colour where they overlap, and end with a confidence "
+        "score from 0 to 1 for your conclusion."
+    )
+    question_id = f"{record['id']}_open_q1"
+    public = {"question_id": question_id, "image": Path(record["image_path"]).name, "prompt": prompt}
+    answer_row = {
+        "question_id": question_id,
+        "image": Path(record["image_path"]).name,
+        "target": target,
+        "target_position": position,
+        "target_degree": len(incident),
+        "lines_at_target": len(incident),
+        "colors": json.dumps(colors, separators=(",", ":")),
+        "reached": json.dumps(reached, separators=(",", ":")),
+        "unreached": json.dumps(unreached, separators=(",", ":")),
+        "answer": answer,
+    }
+    connected_pairs = sorted({"".join(sorted((route["start"], route["end"]))) for route in record["routes"] if route["start"] != route["end"]})
+    annotation = {
+        **public,
+        **{key: value for key, value in answer_row.items() if key not in {"question_id", "image"}},
+        "dataset_version": record.get("dataset_version"),
+        "scoring": {
+            "conclusion_requires_exact_reached_set": True,
+            "partial_match_fields": ["lines_at_target", "colors", "reached", "unreached"],
+            "confidence_range": [0, 1],
+        },
+        "derivation": {
+            "degrees": degrees,
+            "qualifying_targets": candidates,
+            "selection_rule": selection_rule,
+            "endpoint_layout": endpoints,
+            "incident_routes": incident,
+            "connected_label_pairs": connected_pairs,
+            "connected_pairs_out_of_canonical_15": len(connected_pairs),
+        },
+    }
+    return public, answer_row, annotation
+
+
+def write_csv(path: Path, columns: list[str], rows: list[dict]) -> None:
+    with path.open("w", encoding="utf-8-sig", newline="") as handle:
+        writer = csv.DictWriter(handle, fieldnames=columns, lineterminator="\n")
+        writer.writeheader()
+        writer.writerows(rows)
+
+
+def main() -> None:
+    public_rows, answer_rows, annotations = [], [], []
+    for record in read_records():
+        public, answer, annotation = derive(record)
+        public_rows.append(public)
+        answer_rows.append(answer)
+        annotations.append(annotation)
+    if len(public_rows) != 3000:
+        raise RuntimeError(f"Expected 3,000 records, found {len(public_rows)}")
+    write_csv(ROOT / "open_questions.csv", PUBLIC_COLUMNS, public_rows)
+    write_csv(ROOT / "open_answer_key.csv", ANSWER_COLUMNS, answer_rows)
+    with (ROOT / "open_annotations.jsonl").open("w", encoding="utf-8", newline="\n") as handle:
+        for row in annotations:
+            handle.write(json.dumps(row, ensure_ascii=False, sort_keys=True, separators=(",", ":")) + "\n")
+    print(f"Wrote {len(public_rows)} route open questions")
+
+
+if __name__ == "__main__":
+    main()

Dataset/route_dataset_3000/generate_route_dataset.py

@@ -14,7 +14,7 @@ try:
 except ImportError:
  def tqdm(x,**_):return x
 
-BG="#FDFAF4";INK="#101719";SCALE=3;GRID=15
+BG="#FDFAF4";INK="#101719";SCALE=3;GRID=15;DATASET_VERSION="route-3.0.0"
 COLORS=[("teal","#009E9A"),("orange","#F05A17"),("blue","#0878BD"),("green","#149B43"),("red","#D7192D"),("purple","#9256C2"),("brown","#7A4938"),("amber","#C77A00"),("magenta","#D51A78"),("cyan","#00A8C2"),("olive","#799514"),("gray-blue","#557383")]
 
 def font(size):
@@ -128,7 +128,7 @@ def generate_one(index,images_dir):
   sp=port_point(endpoints[a],used[a],totals[a]);used[a]+=1;ep=port_point(endpoints[b],used[b],totals[b]);used[b]+=1;bends=compatible_bends(endpoints[a]["side"],endpoints[b]["side"],rng)
   pts=generate_route(sp,ep,bounds,bends,endpoints[a]["side"],rng);routes.append({"color":name,"hex":hx,"start":a,"end":b,"num_bends":bends,"line_width":line_width,"points":[list(p) for p in pts]})
  cross=compute_crossings(routes);avg=sum(r["num_bends"] for r in routes)/nroute;difficulty=min(1,round(.22*(nroute-6)/6+.13*(nend-4)/2+.20*(avg-3)/5+.15*(3-line_width)/2+.30*min(cross,60)/60,4));iid=f"route_puzzle_{index:04d}";render(images_dir/f"{iid}.png",size,endpoints,routes,line_width)
- return {"id":iid,"image_path":f"images/{iid}.png","canvas_size":list(size),"num_endpoints":nend,"endpoint_letters":letters,"num_routes":nroute,"colors_used":[r["color"] for r in routes],"line_width_px":line_width,"seed":index,"routes":routes,"crossing_count":cross,"difficulty_score":difficulty,"questions":questions_for(iid,letters,routes,rng)}
+ return {"dataset_version":DATASET_VERSION,"id":iid,"image_path":f"images/{iid}.png","canvas_size":list(size),"num_endpoints":nend,"endpoint_letters":letters,"num_routes":nroute,"colors_used":[r["color"] for r in routes],"line_width_px":line_width,"seed":index,"routes":routes,"crossing_count":cross,"difficulty_score":difficulty,"questions":questions_for(iid,letters,routes,rng)}
 
 def validate_dataset(out):
  rows=[]

Dataset/route_dataset_3000/open_validation_metrics.json

@@ -0,0 +1,1555 @@
+{
+  "answer_distributions": {
+    "answer": {
+      "connected to every other labelled side": 261,
+      "some labelled sides remain unreached": 2739
+    },
+    "colors": {
+      "[\"amber\", \"blue\", \"brown\"]": 2,
+      "[\"amber\", \"blue\", \"cyan\"]": 1,
+      "[\"amber\", \"blue\", \"gray-blue\"]": 1,
+      "[\"amber\", \"blue\", \"green\"]": 2,
+      "[\"amber\", \"blue\", \"magenta\"]": 4,
+      "[\"amber\", \"blue\", \"olive\"]": 3,
+      "[\"amber\", \"blue\", \"purple\", \"green\", \"magenta\"]": 1,
+      "[\"amber\", \"blue\", \"purple\", \"green\"]": 1,
+      "[\"amber\", \"blue\", \"purple\"]": 1,
+      "[\"amber\", \"blue\", \"red\", \"brown\", \"purple\"]": 1,
+      "[\"amber\", \"blue\", \"red\"]": 1,
+      "[\"amber\", \"blue\", \"teal\"]": 1,
+      "[\"amber\", \"blue\"]": 5,
+      "[\"amber\", \"brown\", \"blue\"]": 1,
+      "[\"amber\", \"brown\", \"cyan\", \"green\"]": 1,
+      "[\"amber\", \"brown\", \"cyan\"]": 1,
+      "[\"amber\", \"brown\", \"gray-blue\"]": 4,
+      "[\"amber\", \"brown\", \"magenta\", \"green\"]": 1,
+      "[\"amber\", \"brown\", \"magenta\"]": 1,
+      "[\"amber\", \"brown\", \"olive\"]": 1,
+      "[\"amber\", \"brown\", \"orange\"]": 3,
+      "[\"amber\", \"brown\", \"purple\", \"olive\"]": 1,
+      "[\"amber\", \"brown\", \"purple\"]": 2,
+      "[\"amber\", \"brown\", \"red\"]": 2,
+      "[\"amber\", \"brown\", \"teal\"]": 1,
+      "[\"amber\", \"brown\"]": 5,
+      "[\"amber\", \"cyan\", \"blue\", \"brown\"]": 1,
+      "[\"amber\", \"cyan\", \"brown\", \"green\", \"orange\"]": 1,
+      "[\"amber\", \"cyan\", \"brown\"]": 2,
+      "[\"amber\", \"cyan\", \"green\"]": 2,
+      "[\"amber\", \"cyan\", \"magenta\"]": 3,
+      "[\"amber\", \"cyan\", \"olive\"]": 2,
+      "[\"amber\", \"cyan\", \"orange\"]": 1,
+      "[\"amber\", \"cyan\", \"purple\"]": 2,
+      "[\"amber\", \"cyan\", \"teal\", \"orange\"]": 1,
+      "[\"amber\", \"cyan\", \"teal\"]": 2,
+      "[\"amber\", \"cyan\"]": 11,
+      "[\"amber\", \"gray-blue\", \"blue\"]": 3,
+      "[\"amber\", \"gray-blue\", \"green\"]": 1,
+      "[\"amber\", \"gray-blue\", \"magenta\"]": 1,
+      "[\"amber\", \"gray-blue\", \"olive\"]": 2,
+      "[\"amber\", \"gray-blue\", \"purple\"]": 2,
+      "[\"amber\", \"gray-blue\", \"teal\"]": 1,
+      "[\"amber\", \"gray-blue\"]": 10,
+      "[\"amber\", \"green\", \"blue\"]": 1,
+      "[\"amber\", \"green\", \"gray-blue\"]": 1,
+      "[\"amber\", \"green\", \"magenta\"]": 1,
+      "[\"amber\", \"green\", \"orange\", \"purple\"]": 1,
+      "[\"amber\", \"green\", \"orange\"]": 2,
+      "[\"amber\", \"green\", \"purple\"]": 3,
+      "[\"amber\", \"green\", \"teal\"]": 1,
+      "[\"amber\", \"green\"]": 6,
+      "[\"amber\", \"magenta\", \"blue\"]": 2,
+      "[\"amber\", \"magenta\", \"brown\"]": 1,
+      "[\"amber\", \"magenta\", \"cyan\"]": 2,
+      "[\"amber\", \"magenta\", \"gray-blue\"]": 1,
+      "[\"amber\", \"magenta\", \"olive\", \"cyan\"]": 1,
+      "[\"amber\", \"magenta\", \"olive\"]": 3,
+      "[\"amber\", \"magenta\", \"purple\"]": 1,
+      "[\"amber\", \"magenta\", \"red\"]": 2,
+      "[\"amber\", \"magenta\", \"teal\"]": 1,
+      "[\"amber\", \"magenta\"]": 5,
+      "[\"amber\", \"olive\", \"blue\"]": 1,
+      "[\"amber\", \"olive\", \"brown\"]": 2,
+      "[\"amber\", \"olive\", \"cyan\"]": 1,
+      "[\"amber\", \"olive\", \"green\"]": 1,
+      "[\"amber\", \"olive\", \"orange\"]": 3,
+      "[\"amber\", \"olive\", \"purple\", \"orange\"]": 1,
+      "[\"amber\", \"olive\", \"purple\"]": 2,
+      "[\"amber\", \"olive\", \"red\"]": 2,
+      "[\"amber\", \"olive\"]": 9,
+      "[\"amber\", \"orange\", \"blue\"]": 2,
+      "[\"amber\", \"orange\", \"magenta\", \"olive\"]": 1,
+      "[\"amber\", \"orange\", \"magenta\"]": 1,
+      "[\"amber\", \"orange\", \"olive\"]": 1,
+      "[\"amber\", \"orange\", \"purple\"]": 1,
+      "[\"amber\", \"orange\", \"red\"]": 1,
+      "[\"amber\", \"orange\", \"teal\", \"blue\", \"red\"]": 1,
+      "[\"amber\", \"orange\", \"teal\", \"magenta\", \"blue\"]": 1,
+      "[\"amber\", \"orange\", \"teal\"]": 1,
+      "[\"amber\", \"orange\"]": 10,
+      "[\"amber\", \"purple\", \"blue\"]": 1,
+      "[\"amber\", \"purple\", \"brown\"]": 1,
+      "[\"amber\", \"purple\", \"cyan\", \"red\", \"orange\"]": 1,
+      "[\"amber\", \"purple\", \"gray-blue\"]": 1,
+      "[\"amber\", \"purple\", \"green\"]": 1,
+      "[\"amber\", \"purple\", \"olive\", \"gray-blue\"]": 1,
+      "[\"amber\", \"purple\", \"orange\"]": 1,
+      "[\"amber\", \"purple\", \"red\"]": 2,
+      "[\"amber\", \"purple\", \"teal\"]": 1,
+      "[\"amber\", \"purple\"]": 14,
+      "[\"amber\", \"red\", \"blue\"]": 1,
+      "[\"amber\", \"red\", \"brown\"]": 3,
+      "[\"amber\", \"red\", \"cyan\"]": 1,
+      "[\"amber\", \"red\", \"green\"]": 1,
+      "[\"amber\", \"red\", \"magenta\"]": 4,
+      "[\"amber\", \"red\", \"orange\"]": 4,
+      "[\"amber\", \"red\"]": 11,
+      "[\"amber\", \"teal\", \"blue\"]": 3,
+      "[\"amber\", \"teal\", \"brown\"]": 2,
+      "[\"amber\", \"teal\", \"cyan\", \"green\"]": 1,
+      "[\"amber\", \"teal\", \"cyan\"]": 2,
+      "[\"amber\", \"teal\", \"magenta\"]": 2,
+      "[\"amber\", \"teal\", \"olive\"]": 1,
+      "[\"amber\", \"teal\", \"orange\"]": 1,
+      "[\"amber\", \"teal\", \"purple\"]": 2,
+      "[\"amber\", \"teal\"]": 3,
+      "[\"amber\"]": 6,
+      "[\"blue\", \"amber\", \"gray-blue\"]": 2,
+      "[\"blue\", \"amber\", \"green\"]": 2,
+      "[\"blue\", \"amber\", \"magenta\"]": 2,
+      "[\"blue\", \"amber\", \"purple\"]": 1,
+      "[\"blue\", \"amber\", \"red\", \"brown\"]": 1,
+      "[\"blue\", \"amber\", \"red\"]": 1,
+      "[\"blue\", \"amber\", \"teal\"]": 1,
+      "[\"blue\", \"amber\"]": 10,
+      "[\"blue\", \"brown\", \"cyan\"]": 2,
+      "[\"blue\", \"brown\", \"gray-blue\"]": 2,
+      "[\"blue\", \"brown\", \"magenta\"]": 2,
+      "[\"blue\", \"brown\", \"orange\"]": 4,
+      "[\"blue\", \"brown\", \"purple\"]": 2,
+      "[\"blue\", \"brown\", \"red\", \"olive\", \"orange\"]": 1,
+      "[\"blue\", \"brown\", \"red\"]": 3,
+      "[\"blue\", \"brown\", \"teal\"]": 2,
+      "[\"blue\", \"brown\"]": 9,
+      "[\"blue\", \"cyan\", \"amber\"]": 1,
+      "[\"blue\", \"cyan\", \"brown\"]": 2,
+      "[\"blue\", \"cyan\", \"green\", \"brown\"]": 1,
+      "[\"blue\", \"cyan\", \"green\"]": 1,
+      "[\"blue\", \"cyan\", \"olive\"]": 2,
+      "[\"blue\", \"cyan\", \"purple\"]": 2,
+      "[\"blue\", \"cyan\", \"red\"]": 1,
+      "[\"blue\", \"cyan\"]": 13,
+      "[\"blue\", \"gray-blue\", \"amber\", \"orange\"]": 1,
+      "[\"blue\", \"gray-blue\", \"amber\", \"teal\"]": 1,
+      "[\"blue\", \"gray-blue\", \"amber\"]": 1,
+      "[\"blue\", \"gray-blue\", \"brown\"]": 1,
+      "[\"blue\", \"gray-blue\", \"olive\"]": 1,
+      "[\"blue\", \"gray-blue\", \"orange\", \"brown\"]": 1,
+      "[\"blue\", \"gray-blue\", \"orange\"]": 1,
+      "[\"blue\", \"gray-blue\", \"purple\"]": 1,
+      "[\"blue\", \"gray-blue\", \"red\"]": 1,
+      "[\"blue\", \"gray-blue\", \"teal\"]": 1,
+      "[\"blue\", \"gray-blue\"]": 4,
+      "[\"blue\", \"green\", \"amber\"]": 1,
+      "[\"blue\", \"green\", \"brown\"]": 2,
+      "[\"blue\", \"green\", \"cyan\"]": 1,
+      "[\"blue\", \"green\", \"gray-blue\"]": 4,
+      "[\"blue\", \"green\", \"olive\", \"brown\"]": 1,
+      "[\"blue\", \"green\", \"olive\"]": 2,
+      "[\"blue\", \"green\", \"orange\"]": 1,
+      "[\"blue\", \"green\", \"purple\", \"gray-blue\", \"olive\"]": 1,
+      "[\"blue\", \"green\", \"purple\"]": 2,
+      "[\"blue\", \"green\", \"red\"]": 1,
+      "[\"blue\", \"green\", \"teal\"]": 2,
+      "[\"blue\", \"green\"]": 3,
+      "[\"blue\", \"magenta\", \"amber\"]": 1,
+      "[\"blue\", \"magenta\", \"brown\", \"amber\"]": 1,
+      "[\"blue\", \"magenta\", \"brown\", \"red\"]": 1,
+      "[\"blue\", \"magenta\", \"brown\"]": 1,
+      "[\"blue\", \"magenta\", \"gray-blue\"]": 2,
+      "[\"blue\", \"magenta\", \"green\"]": 3,
+      "[\"blue\", \"magenta\", \"olive\"]": 2,
+      "[\"blue\", \"magenta\", \"red\"]": 3,
+      "[\"blue\", \"magenta\", \"teal\"]": 1,
+      "[\"blue\", \"magenta\"]": 12,
+      "[\"blue\", \"olive\", \"amber\"]": 3,
+      "[\"blue\", \"olive\", \"brown\", \"magenta\"]": 1,
+      "[\"blue\", \"olive\", \"brown\"]": 2,
+      "[\"blue\", \"olive\", \"cyan\", \"magenta\"]": 1,
+      "[\"blue\", \"olive\", \"cyan\"]": 1,
+      "[\"blue\", \"olive\", \"gray-blue\"]": 2,
+      "[\"blue\", \"olive\", \"green\"]": 3,
+      "[\"blue\", \"olive\", \"magenta\", \"orange\", \"red\"]": 1,
+      "[\"blue\", \"olive\", \"magenta\"]": 3,
+      "[\"blue\", \"olive\", \"orange\"]": 3,
+      "[\"blue\", \"olive\", \"purple\"]": 1,
+      "[\"blue\", \"olive\"]": 5,
+      "[\"blue\", \"orange\", \"brown\"]": 1,
+      "[\"blue\", \"orange\", \"cyan\"]": 2,
+      "[\"blue\", \"orange\", \"gray-blue\"]": 1,
+      "[\"blue\", \"orange\", \"green\", \"brown\", \"amber\"]": 1,
+      "[\"blue\", \"orange\", \"olive\"]": 2,
+      "[\"blue\", \"orange\", \"purple\"]": 2,
+      "[\"blue\", \"orange\", \"red\"]": 3,
+      "[\"blue\", \"orange\", \"teal\", \"gray-blue\"]": 1,
+      "[\"blue\", \"orange\", \"teal\"]": 1,
+      "[\"blue\", \"orange\"]": 9,
+      "[\"blue\", \"purple\", \"amber\"]": 3,
+      "[\"blue\", \"purple\", \"brown\"]": 2,
+      "[\"blue\", \"purple\", \"cyan\", \"teal\"]": 1,
+      "[\"blue\", \"purple\", \"gray-blue\"]": 2,
+      "[\"blue\", \"purple\", \"green\", \"amber\", \"magenta\"]": 1,
+      "[\"blue\", \"purple\", \"green\"]": 2,
+      "[\"blue\", \"purple\", \"magenta\", \"orange\", \"gray-blue\"]": 1,
+      "[\"blue\", \"purple\", \"magenta\"]": 1,
+      "[\"blue\", \"purple\", \"olive\"]": 2,
+      "[\"blue\", \"purple\", \"orange\"]": 1,
+      "[\"blue\", \"purple\", \"red\"]": 1,
+      "[\"blue\", \"purple\"]": 7,
+      "[\"blue\", \"red\", \"amber\"]": 2,
+      "[\"blue\", \"red\", \"brown\"]": 1,
+      "[\"blue\", \"red\", \"cyan\"]": 1,
+      "[\"blue\", \"red\", \"gray-blue\"]": 2,
+      "[\"blue\", \"red\", \"green\"]": 2,
+      "[\"blue\", \"red\", \"olive\"]": 1,
+      "[\"blue\", \"red\", \"orange\"]": 1,
+      "[\"blue\", \"red\", \"purple\"]": 1,
+      "[\"blue\", \"red\", \"teal\"]": 1,
+      "[\"blue\", \"red\"]": 7,
+      "[\"blue\", \"teal\", \"gray-blue\", \"green\"]": 1,
+      "[\"blue\", \"teal\", \"gray-blue\"]": 1,
+      "[\"blue\", \"teal\", \"green\", \"red\"]": 1,
+      "[\"blue\", \"teal\", \"green\"]": 2,
+      "[\"blue\", \"teal\", \"olive\", \"purple\"]": 1,
+      "[\"blue\", \"teal\", \"orange\", \"purple\"]": 1,
+      "[\"blue\", \"teal\", \"orange\"]": 1,
+      "[\"blue\", \"teal\", \"red\"]": 1,
+      "[\"blue\", \"teal\"]": 8,
+      "[\"blue\"]": 5,
+      "[\"brown\", \"amber\", \"blue\"]": 1,
+      "[\"brown\", \"amber\", \"cyan\", \"blue\"]": 1,
+      "[\"brown\", \"amber\", \"cyan\"]": 1,
+      "[\"brown\", \"amber\", \"gray-blue\", \"olive\"]": 1,
+      "[\"brown\", \"amber\", \"gray-blue\"]": 1,
+      "[\"brown\", \"amber\", \"green\"]": 2,
+      "[\"brown\", \"amber\", \"magenta\", \"cyan\"]": 1,
+      "[\"brown\", \"amber\", \"olive\", \"teal\"]": 1,
+      "[\"brown\", \"amber\", \"olive\"]": 1,
+      "[\"brown\", \"amber\", \"orange\"]": 3,
+      "[\"brown\", \"amber\", \"purple\", \"blue\"]": 1,
+      "[\"brown\", \"amber\", \"red\"]": 1,
+      "[\"brown\", \"amber\", \"teal\"]": 2,
+      "[\"brown\", \"amber\"]": 8,
+      "[\"brown\", \"blue\", \"amber\"]": 1,
+      "[\"brown\", \"blue\", \"cyan\", \"orange\"]": 1,
+      "[\"brown\", \"blue\", \"gray-blue\"]": 1,
+      "[\"brown\", \"blue\", \"green\"]": 2,
+      "[\"brown\", \"blue\", \"magenta\"]": 3,
+      "[\"brown\", \"blue\", \"olive\"]": 3,
+      "[\"brown\", \"blue\", \"orange\"]": 1,
+      "[\"brown\", \"blue\", \"purple\", \"teal\", \"olive\"]": 1,
+      "[\"brown\", \"blue\", \"purple\", \"teal\"]": 1,
+      "[\"brown\", \"blue\", \"purple\"]": 3,
+      "[\"brown\", \"blue\", \"red\", \"magenta\", \"orange\"]": 1,
+      "[\"brown\", \"blue\", \"red\"]": 1,
+      "[\"brown\", \"blue\"]": 10,
+      "[\"brown\", \"cyan\", \"amber\"]": 1,
+      "[\"brown\", \"cyan\", \"gray-blue\", \"blue\"]": 1,
+      "[\"brown\", \"cyan\", \"gray-blue\"]": 2,
+      "[\"brown\", \"cyan\", \"green\", \"gray-blue\"]": 1,
+      "[\"brown\", \"cyan\", \"green\"]": 2,
+      "[\"brown\", \"cyan\", \"magenta\"]": 1,
+      "[\"brown\", \"cyan\", \"olive\"]": 1,
+      "[\"brown\", \"cyan\", \"orange\", \"teal\", \"gray-blue\"]": 1,
+      "[\"brown\", \"cyan\", \"orange\"]": 2,
+      "[\"brown\", \"cyan\", \"purple\"]": 3,
+      "[\"brown\", \"cyan\", \"red\"]": 2,
+      "[\"brown\", \"cyan\", \"teal\"]": 1,
+      "[\"brown\", \"cyan\"]": 8,
+      "[\"brown\", \"gray-blue\", \"amber\"]": 2,
+      "[\"brown\", \"gray-blue\", \"blue\", \"magenta\"]": 1,
+      "[\"brown\", \"gray-blue\", \"blue\"]": 2,
+      "[\"brown\", \"gray-blue\", \"green\"]": 1,
+      "[\"brown\", \"gray-blue\", \"magenta\"]": 2,
+      "[\"brown\", \"gray-blue\", \"orange\"]": 4,
+      "[\"brown\", \"gray-blue\", \"purple\"]": 1,
+      "[\"brown\", \"gray-blue\", \"teal\"]": 1,
+      "[\"brown\", \"gray-blue\"]": 8,
+      "[\"brown\", \"green\", \"cyan\", \"blue\"]": 1,
+      "[\"brown\", \"green\", \"cyan\"]": 1,
+      "[\"brown\", \"green\", \"gray-blue\"]": 2,
+      "[\"brown\", \"green\", \"olive\", \"teal\"]": 1,
+      "[\"brown\", \"green\", \"olive\"]": 2,
+      "[\"brown\", \"green\", \"orange\", \"teal\"]": 1,
+      "[\"brown\", \"green\", \"orange\"]": 1,
+      "[\"brown\", \"green\"]": 15,
+      "[\"brown\", \"magenta\", \"amber\"]": 1,
+      "[\"brown\", \"magenta\", \"blue\"]": 1,
+      "[\"brown\", \"magenta\", \"cyan\", \"blue\"]": 1,
+      "[\"brown\", \"magenta\", \"cyan\"]": 1,
+      "[\"brown\", \"magenta\", \"gray-blue\", \"olive\"]": 1,
+      "[\"brown\", \"magenta\", \"gray-blue\"]": 1,
+      "[\"brown\", \"magenta\", \"green\"]": 2,
+      "[\"brown\", \"magenta\", \"orange\"]": 1,
+      "[\"brown\", \"magenta\", \"purple\", \"orange\"]": 1,
+      "[\"brown\", \"magenta\", \"purple\"]": 2,
+      "[\"brown\", \"magenta\", \"red\"]": 1,
+      "[\"brown\", \"magenta\", \"teal\", \"red\"]": 1,
+      "[\"brown\", \"magenta\", \"teal\"]": 2,
+      "[\"brown\", \"magenta\"]": 7,
+      "[\"brown\", \"olive\", \"amber\"]": 2,
+      "[\"brown\", \"olive\", \"blue\"]": 2,
+      "[\"brown\", \"olive\", \"cyan\", \"orange\"]": 1,
+      "[\"brown\", \"olive\", \"cyan\"]": 1,
+      "[\"brown\", \"olive\", \"gray-blue\", \"purple\", \"green\"]": 1,
+      "[\"brown\", \"olive\", \"gray-blue\"]": 1,
+      "[\"brown\", \"olive\", \"green\", \"blue\"]": 1,
+      "[\"brown\", \"olive\", \"green\", \"teal\"]": 1,
+      "[\"brown\", \"olive\", \"green\"]": 1,
+      "[\"brown\", \"olive\", \"magenta\", \"green\"]": 1,
+      "[\"brown\", \"olive\", \"purple\", \"amber\"]": 1,
+      "[\"brown\", \"olive\", \"purple\"]": 3,
+      "[\"brown\", \"olive\", \"red\"]": 2,
+      "[\"brown\", \"olive\", \"teal\"]": 2,
+      "[\"brown\", \"olive\"]": 7,
+      "[\"brown\", \"orange\", \"amber\"]": 1,
+      "[\"brown\", \"orange\", \"blue\"]": 1,
+      "[\"brown\", \"orange\", \"green\"]": 1,
+      "[\"brown\", \"orange\", \"olive\", \"green\"]": 1,
+      "[\"brown\", \"orange\", \"olive\"]": 2,
+      "[\"brown\", \"orange\", \"purple\"]": 2,
+      "[\"brown\", \"orange\", \"red\"]": 1,
+      "[\"brown\", \"orange\", \"teal\"]": 2,
+      "[\"brown\", \"orange\"]": 20,
+      "[\"brown\", \"purple\", \"blue\", \"amber\"]": 1,
+      "[\"brown\", \"purple\", \"blue\"]": 1,
+      "[\"brown\", \"purple\", \"cyan\"]": 3,
+      "[\"brown\", \"purple\", \"gray-blue\"]": 2,
+      "[\"brown\", \"purple\", \"green\"]": 1,
+      "[\"brown\", \"purple\", \"magenta\"]": 3,
+      "[\"brown\", \"purple\", \"olive\"]": 1,
+      "[\"brown\", \"purple\", \"orange\", \"amber\", \"magenta\"]": 1,
+      "[\"brown\", \"purple\", \"orange\", \"magenta\"]": 1,
+      "[\"brown\", \"purple\", \"orange\"]": 3,
+      "[\"brown\", \"purple\", \"red\", \"gray-blue\", \"teal\"]": 1,
+      "[\"brown\", \"purple\", \"teal\"]": 1,
+      "[\"brown\", \"purple\"]": 10,
+      "[\"brown\", \"red\", \"blue\"]": 2,
+      "[\"brown\", \"red\", \"cyan\"]": 1,
+      "[\"brown\", \"red\", \"gray-blue\"]": 1,
+      "[\"brown\", \"red\", \"green\", \"orange\"]": 1,
+      "[\"brown\", \"red\", \"olive\"]": 1,
+      "[\"brown\", \"red\", \"orange\", \"teal\"]": 1,
+      "[\"brown\", \"red\", \"orange\"]": 2,
+      "[\"brown\", \"red\", \"teal\"]": 1,
+      "[\"brown\", \"red\"]": 6,
+      "[\"brown\", \"teal\", \"amber\", \"green\"]": 1,
+      "[\"brown\", \"teal\", \"amber\"]": 1,
+      "[\"brown\", \"teal\", \"blue\"]": 1,
+      "[\"brown\", \"teal\", \"cyan\"]": 2,
+      "[\"brown\", \"teal\", \"green\"]": 1,
+      "[\"brown\", \"teal\", \"magenta\"]": 2,
+      "[\"brown\", \"teal\", \"olive\"]": 1,
+      "[\"brown\", \"teal\", \"orange\", \"green\", \"magenta\"]": 1,
+      "[\"brown\", \"teal\", \"orange\"]": 2,
+      "[\"brown\", \"teal\", \"purple\", \"olive\", \"green\"]": 1,
+      "[\"brown\", \"teal\", \"red\"]": 2,
+      "[\"brown\", \"teal\"]": 13,
+      "[\"brown\"]": 6,
+      "[\"cyan\", \"amber\", \"blue\"]": 2,
+      "[\"cyan\", \"amber\", \"gray-blue\"]": 1,
+      "[\"cyan\", \"amber\", \"magenta\"]": 1,
+      "[\"cyan\", \"amber\", \"orange\"]": 2,
+      "[\"cyan\", \"amber\", \"purple\", \"brown\"]": 1,
+      "[\"cyan\", \"amber\", \"purple\"]": 2,
+      "[\"cyan\", \"amber\", \"red\"]": 1,
+      "[\"cyan\", \"amber\", \"teal\", \"blue\", \"gray-blue\"]": 1,
+      "[\"cyan\", \"amber\", \"teal\"]": 3,
+      "[\"cyan\", \"amber\"]": 4,
+      "[\"cyan\", \"blue\", \"amber\"]": 1,
+      "[\"cyan\", \"blue\", \"brown\"]": 2,
+      "[\"cyan\", \"blue\", \"green\", \"olive\"]": 1,
+      "[\"cyan\", \"blue\", \"green\"]": 2,
+      "[\"cyan\", \"blue\", \"olive\"]": 2,
+      "[\"cyan\", \"blue\", \"orange\"]": 2,
+      "[\"cyan\", \"blue\", \"red\"]": 1,
+      "[\"cyan\", \"blue\", \"teal\"]": 2,
+      "[\"cyan\", \"blue\"]": 13,
+      "[\"cyan\", \"brown\", \"blue\", \"red\", \"purple\"]": 1,
+      "[\"cyan\", \"brown\", \"gray-blue\"]": 1,
+      "[\"cyan\", \"brown\", \"green\", \"purple\"]": 1,
+      "[\"cyan\", \"brown\", \"green\"]": 3,
+      "[\"cyan\", \"brown\", \"magenta\"]": 3,
+      "[\"cyan\", \"brown\", \"olive\"]": 3,
+      "[\"cyan\", \"brown\", \"orange\"]": 2,
+      "[\"cyan\", \"brown\", \"purple\"]": 1,
+      "[\"cyan\", \"brown\", \"red\"]": 1,
+      "[\"cyan\", \"brown\", \"teal\"]": 1,
+      "[\"cyan\", \"brown\"]": 7,
+      "[\"cyan\", \"gray-blue\", \"green\"]": 2,
+      "[\"cyan\", \"gray-blue\", \"magenta\"]": 1,
+      "[\"cyan\", \"gray-blue\", \"olive\", \"red\"]": 1,
+      "[\"cyan\", \"gray-blue\", \"orange\"]": 2,
+      "[\"cyan\", \"gray-blue\", \"red\"]": 1,
+      "[\"cyan\", \"gray-blue\", \"teal\", \"blue\"]": 1,
+      "[\"cyan\", \"gray-blue\"]": 7,
+      "[\"cyan\", \"green\", \"amber\"]": 2,
+      "[\"cyan\", \"green\", \"brown\"]": 2,
+      "[\"cyan\", \"green\", \"gray-blue\", \"blue\", \"orange\"]": 1,
+      "[\"cyan\", \"green\", \"gray-blue\", \"blue\"]": 1,
+      "[\"cyan\", \"green\", \"gray-blue\", \"olive\"]": 1,
+      "[\"cyan\", \"green\", \"gray-blue\"]": 2,
+      "[\"cyan\", \"green\", \"magenta\"]": 1,
+      "[\"cyan\", \"green\", \"olive\"]": 3,
+      "[\"cyan\", \"green\", \"red\", \"purple\"]": 1,
+      "[\"cyan\", \"green\", \"red\"]": 2,
+      "[\"cyan\", \"green\", \"teal\"]": 1,
+      "[\"cyan\", \"green\"]": 8,
+      "[\"cyan\", \"magenta\", \"blue\"]": 1,
+      "[\"cyan\", \"magenta\", \"gray-blue\"]": 1,
+      "[\"cyan\", \"magenta\", \"green\"]": 2,
+      "[\"cyan\", \"magenta\", \"olive\"]": 1,
+      "[\"cyan\", \"magenta\", \"orange\"]": 1,
+      "[\"cyan\", \"magenta\", \"purple\"]": 1,
+      "[\"cyan\", \"magenta\", \"red\", \"green\"]": 1,
+      "[\"cyan\", \"magenta\"]": 9,
+      "[\"cyan\", \"olive\", \"amber\", \"orange\"]": 1,
+      "[\"cyan\", \"olive\", \"blue\"]": 2,
+      "[\"cyan\", \"olive\", \"brown\"]": 2,
+      "[\"cyan\", \"olive\", \"gray-blue\"]": 1,
+      "[\"cyan\", \"olive\", \"green\"]": 1,
+      "[\"cyan\", \"olive\", \"magenta\"]": 1,
+      "[\"cyan\", \"olive\", \"orange\"]": 2,
+      "[\"cyan\", \"olive\", \"red\"]": 2,
+      "[\"cyan\", \"olive\", \"teal\", \"brown\"]": 1,
+      "[\"cyan\", \"olive\", \"teal\"]": 3,
+      "[\"cyan\", \"olive\"]": 9,
+      "[\"cyan\", \"orange\", \"brown\"]": 1,
+      "[\"cyan\", \"orange\", \"gray-blue\"]": 1,
+      "[\"cyan\", \"orange\", \"olive\"]": 1,
+      "[\"cyan\", \"orange\", \"purple\", \"gray-blue\", \"red\"]": 1,
+      "[\"cyan\", \"orange\", \"purple\"]": 2,
+      "[\"cyan\", \"orange\", \"red\"]": 1,
+      "[\"cyan\", \"orange\", \"teal\", \"amber\", \"blue\"]": 1,
+      "[\"cyan\", \"orange\", \"teal\", \"brown\"]": 1,
+      "[\"cyan\", \"orange\", \"teal\"]": 1,
+      "[\"cyan\", \"orange\"]": 12,
+      "[\"cyan\", \"purple\", \"amber\", \"brown\", \"blue\"]": 1,
+      "[\"cyan\", \"purple\", \"gray-blue\", \"brown\", \"red\"]": 1,
+      "[\"cyan\", \"purple\", \"gray-blue\"]": 1,
+      "[\"cyan\", \"purple\", \"magenta\"]": 1,
+      "[\"cyan\", \"purple\", \"olive\"]": 4,
+      "[\"cyan\", \"purple\", \"red\"]": 2,
+      "[\"cyan\", \"purple\", \"teal\"]": 1,
+      "[\"cyan\", \"purple\"]": 7,
+      "[\"cyan\", \"red\", \"amber\"]": 3,
+      "[\"cyan\", \"red\", \"blue\"]": 1,
+      "[\"cyan\", \"red\", \"brown\"]": 1,
+      "[\"cyan\", \"red\", \"gray-blue\"]": 3,
+      "[\"cyan\", \"red\", \"olive\"]": 2,
+      "[\"cyan\", \"red\", \"orange\"]": 4,
+      "[\"cyan\", \"red\", \"purple\", \"green\"]": 1,
+      "[\"cyan\", \"red\", \"purple\"]": 1,
+      "[\"cyan\", \"red\", \"teal\"]": 1,
+      "[\"cyan\", \"red\"]": 7,
+      "[\"cyan\", \"teal\", \"amber\", \"brown\"]": 1,
+      "[\"cyan\", \"teal\", \"blue\", \"olive\", \"brown\"]": 1,
+      "[\"cyan\", \"teal\", \"blue\", \"olive\"]": 1,
+      "[\"cyan\", \"teal\", \"blue\"]": 1,
+      "[\"cyan\", \"teal\", \"brown\", \"gray-blue\"]": 1,
+      "[\"cyan\", \"teal\", \"brown\"]": 3,
+      "[\"cyan\", \"teal\", \"gray-blue\"]": 1,
+      "[\"cyan\", \"teal\", \"green\"]": 2,
+      "[\"cyan\", \"teal\", \"magenta\", \"gray-blue\"]": 1,
+      "[\"cyan\", \"teal\", \"magenta\"]": 1,
+      "[\"cyan\", \"teal\", \"orange\"]": 1,
+      "[\"cyan\", \"teal\", \"red\"]": 1,
+      "[\"cyan\", \"teal\"]": 10,
+      "[\"cyan\"]": 13,
+      "[\"gray-blue\", \"amber\", \"blue\"]": 1,
+      "[\"gray-blue\", \"amber\", \"brown\", \"red\"]": 1,
+      "[\"gray-blue\", \"amber\", \"brown\"]": 1,
+      "[\"gray-blue\", \"amber\", \"cyan\"]": 2,
+      "[\"gray-blue\", \"amber\", \"orange\"]": 2,
+      "[\"gray-blue\", \"amber\", \"purple\"]": 1,
+      "[\"gray-blue\", \"amber\", \"red\", \"cyan\"]": 1,
+      "[\"gray-blue\", \"amber\", \"red\"]": 2,
+      "[\"gray-blue\", \"amber\"]": 15,
+      "[\"gray-blue\", \"blue\", \"amber\"]": 1,
+      "[\"gray-blue\", \"blue\", \"cyan\"]": 1,
+      "[\"gray-blue\", \"blue\", \"green\", \"brown\"]": 1,
+      "[\"gray-blue\", \"blue\", \"green\"]": 3,
+      "[\"gray-blue\", \"blue\", \"magenta\"]": 1,
+      "[\"gray-blue\", \"blue\", \"olive\"]": 2,
+      "[\"gray-blue\", \"blue\", \"orange\"]": 2,
+      "[\"gray-blue\", \"blue\", \"red\", \"green\"]": 1,
+      "[\"gray-blue\", \"blue\", \"red\"]": 6,
+      "[\"gray-blue\", \"blue\", \"teal\"]": 1,
+      "[\"gray-blue\", \"blue\"]": 10,
+      "[\"gray-blue\", \"brown\", \"amber\"]": 2,
+      "[\"gray-blue\", \"brown\", \"blue\"]": 2,
+      "[\"gray-blue\", \"brown\", \"cyan\"]": 2,
+      "[\"gray-blue\", \"brown\", \"green\"]": 1,
+      "[\"gray-blue\", \"brown\", \"magenta\"]": 3,
+      "[\"gray-blue\", \"brown\", \"orange\"]": 1,
+      "[\"gray-blue\", \"brown\", \"purple\"]": 2,
+      "[\"gray-blue\", \"brown\", \"red\"]": 1,
+      "[\"gray-blue\", \"brown\", \"teal\"]": 3,
+      "[\"gray-blue\", \"brown\"]": 5,
+      "[\"gray-blue\", \"cyan\", \"amber\", \"red\", \"teal\"]": 1,
+      "[\"gray-blue\", \"cyan\", \"blue\"]": 1,
+      "[\"gray-blue\", \"cyan\", \"brown\", \"olive\", \"amber\"]": 1,
+      "[\"gray-blue\", \"cyan\", \"green\"]": 1,
+      "[\"gray-blue\", \"cyan\", \"magenta\"]": 1,
+      "[\"gray-blue\", \"cyan\", \"olive\"]": 1,
+      "[\"gray-blue\", \"cyan\", \"purple\"]": 2,
+      "[\"gray-blue\", \"cyan\", \"red\", \"teal\"]": 2,
+      "[\"gray-blue\", \"cyan\", \"red\"]": 2,
+      "[\"gray-blue\", \"cyan\", \"teal\"]": 1,
+      "[\"gray-blue\", \"cyan\"]": 14,
+      "[\"gray-blue\", \"green\", \"amber\", \"blue\"]": 1,
+      "[\"gray-blue\", \"green\", \"amber\"]": 3,
+      "[\"gray-blue\", \"green\", \"cyan\"]": 1,
+      "[\"gray-blue\", \"green\", \"magenta\", \"amber\"]": 1,
+      "[\"gray-blue\", \"green\", \"magenta\", \"brown\"]": 1,
+      "[\"gray-blue\", \"green\", \"magenta\"]": 1,
+      "[\"gray-blue\", \"green\", \"olive\"]": 1,
+      "[\"gray-blue\", \"green\", \"orange\"]": 2,
+      "[\"gray-blue\", \"green\", \"purple\", \"amber\"]": 1,
+      "[\"gray-blue\", \"green\", \"red\"]": 2,
+      "[\"gray-blue\", \"green\", \"teal\"]": 2,
+      "[\"gray-blue\", \"green\"]": 11,
+      "[\"gray-blue\", \"magenta\", \"amber\", \"brown\", \"teal\"]": 1,
+      "[\"gray-blue\", \"magenta\", \"amber\"]": 1,
+      "[\"gray-blue\", \"magenta\", \"blue\"]": 2,
+      "[\"gray-blue\", \"magenta\", \"brown\"]": 1,
+      "[\"gray-blue\", \"magenta\", \"cyan\"]": 1,
+      "[\"gray-blue\", \"magenta\", \"green\"]": 1,
+      "[\"gray-blue\", \"magenta\", \"olive\"]": 1,
+      "[\"gray-blue\", \"magenta\", \"orange\", \"purple\"]": 1,
+      "[\"gray-blue\", \"magenta\", \"purple\"]": 1,
+      "[\"gray-blue\", \"magenta\", \"teal\"]": 1,
+      "[\"gray-blue\", \"magenta\"]": 6,
+      "[\"gray-blue\", \"olive\", \"blue\"]": 2,
+      "[\"gray-blue\", \"olive\", \"cyan\", \"purple\"]": 1,
+      "[\"gray-blue\", \"olive\", \"cyan\"]": 2,
+      "[\"gray-blue\", \"olive\", \"green\", \"amber\"]": 1,
+      "[\"gray-blue\", \"olive\", \"green\", \"blue\"]": 1,
+      "[\"gray-blue\", \"olive\", \"magenta\"]": 1,
+      "[\"gray-blue\", \"olive\", \"orange\"]": 1,
+      "[\"gray-blue\", \"olive\", \"purple\"]": 1,
+      "[\"gray-blue\", \"olive\", \"red\"]": 4,
+      "[\"gray-blue\", \"olive\", \"teal\", \"green\"]": 1,
+      "[\"gray-blue\", \"olive\"]": 8,
+      "[\"gray-blue\", \"orange\", \"amber\"]": 2,
+      "[\"gray-blue\", \"orange\", \"blue\"]": 2,
+      "[\"gray-blue\", \"orange\", \"brown\"]": 1,
+      "[\"gray-blue\", \"orange\", \"cyan\"]": 1,
+      "[\"gray-blue\", \"orange\", \"green\"]": 2,
+      "[\"gray-blue\", \"orange\", \"magenta\"]": 1,
+      "[\"gray-blue\", \"orange\", \"teal\"]": 1,
+      "[\"gray-blue\", \"orange\"]": 7,
+      "[\"gray-blue\", \"purple\", \"amber\", \"brown\"]": 1,
+      "[\"gray-blue\", \"purple\", \"blue\"]": 2,
+      "[\"gray-blue\", \"purple\", \"brown\"]": 3,
+      "[\"gray-blue\", \"purple\", \"cyan\", \"green\", \"olive\"]": 1,
+      "[\"gray-blue\", \"purple\", \"cyan\", \"green\"]": 1,
+      "[\"gray-blue\", \"purple\", \"cyan\"]": 1,
+      "[\"gray-blue\", \"purple\", \"magenta\"]": 1,
+      "[\"gray-blue\", \"purple\", \"olive\"]": 1,
+      "[\"gray-blue\", \"purple\", \"orange\"]": 1,
+      "[\"gray-blue\", \"purple\", \"red\"]": 2,
+      "[\"gray-blue\", \"purple\", \"teal\", \"magenta\"]": 1,
+      "[\"gray-blue\", \"purple\", \"teal\"]": 1,
+      "[\"gray-blue\", \"purple\"]": 5,
+      "[\"gray-blue\", \"red\", \"amber\"]": 4,
+      "[\"gray-blue\", \"red\", \"blue\", \"amber\"]": 1,
+      "[\"gray-blue\", \"red\", \"blue\", \"brown\"]": 1,
+      "[\"gray-blue\", \"red\", \"brown\"]": 1,
+      "[\"gray-blue\", \"red\", \"cyan\"]": 1,
+      "[\"gray-blue\", \"red\", \"magenta\"]": 1,
+      "[\"gray-blue\", \"red\", \"olive\"]": 2,
+      "[\"gray-blue\", \"red\", \"orange\", \"cyan\"]": 1,
+      "[\"gray-blue\", \"red\", \"orange\", \"green\"]": 1,
+      "[\"gray-blue\", \"red\", \"orange\"]": 1,
+      "[\"gray-blue\", \"red\", \"teal\"]": 3,
+      "[\"gray-blue\", \"red\"]": 6,
+      "[\"gray-blue\", \"teal\", \"amber\", \"blue\"]": 1,
+      "[\"gray-blue\", \"teal\", \"amber\"]": 1,
+      "[\"gray-blue\", \"teal\", \"blue\"]": 2,
+      "[\"gray-blue\", \"teal\", \"brown\"]": 1,
+      "[\"gray-blue\", \"teal\", \"green\", \"blue\"]": 1,
+      "[\"gray-blue\", \"teal\", \"green\"]": 2,
+      "[\"gray-blue\", \"teal\", \"magenta\"]": 1,
+      "[\"gray-blue\", \"teal\", \"olive\"]": 2,
+      "[\"gray-blue\", \"teal\", \"red\"]": 2,
+      "[\"gray-blue\", \"teal\"]": 9,
+      "[\"gray-blue\"]": 8,
+      "[\"green\", \"amber\", \"blue\"]": 1,
+      "[\"green\", \"amber\", \"brown\"]": 2,
+      "[\"green\", \"amber\", \"cyan\"]": 1,
+      "[\"green\", \"amber\", \"magenta\", \"brown\"]": 1,
+      "[\"green\", \"amber\", \"magenta\", \"purple\"]": 1,
+      "[\"green\", \"amber\", \"magenta\"]": 2,
+      "[\"green\", \"amber\", \"olive\"]": 1,
+      "[\"green\", \"amber\", \"orange\", \"cyan\"]": 1,
+      "[\"green\", \"amber\", \"purple\"]": 1,
+      "[\"green\", \"amber\", \"red\"]": 1,
+      "[\"green\", \"amber\"]": 2,
+      "[\"green\", \"blue\", \"amber\", \"teal\"]": 1,
+      "[\"green\", \"blue\", \"amber\"]": 1,
+      "[\"green\", \"blue\", \"brown\", \"amber\"]": 1,
+      "[\"green\", \"blue\", \"brown\"]": 1,
+      "[\"green\", \"blue\", \"gray-blue\", \"orange\"]": 1,
+      "[\"green\", \"blue\", \"gray-blue\"]": 1,
+      "[\"green\", \"blue\", \"magenta\", \"olive\"]": 1,
+      "[\"green\", \"blue\", \"magenta\"]": 1,
+      "[\"green\", \"blue\", \"olive\"]": 1,
+      "[\"green\", \"blue\", \"purple\", \"brown\"]": 1,
+      "[\"green\", \"blue\", \"purple\"]": 2,
+      "[\"green\", \"blue\", \"red\"]": 2,
+      "[\"green\", \"blue\"]": 12,
+      "[\"green\", \"brown\", \"amber\"]": 2,
+      "[\"green\", \"brown\", \"blue\"]": 4,
+      "[\"green\", \"brown\", \"cyan\"]": 1,
+      "[\"green\", \"brown\", \"gray-blue\"]": 1,
+      "[\"green\", \"brown\", \"magenta\"]": 1,
+      "[\"green\", \"brown\", \"purple\"]": 2,
+      "[\"green\", \"brown\", \"red\"]": 2,
+      "[\"green\", \"brown\", \"teal\"]": 1,
+      "[\"green\", \"brown\"]": 10,
+      "[\"green\", \"cyan\", \"amber\"]": 1,
+      "[\"green\", \"cyan\", \"blue\"]": 1,
+      "[\"green\", \"cyan\", \"brown\"]": 1,
+      "[\"green\", \"cyan\", \"gray-blue\", \"amber\"]": 1,
+      "[\"green\", \"cyan\", \"magenta\"]": 4,
+      "[\"green\", \"cyan\", \"olive\"]": 1,
+      "[\"green\", \"cyan\", \"red\"]": 2,
+      "[\"green\", \"cyan\", \"teal\"]": 2,
+      "[\"green\", \"cyan\"]": 5,
+      "[\"green\", \"gray-blue\", \"blue\", \"cyan\", \"olive\"]": 1,
+      "[\"green\", \"gray-blue\", \"blue\"]": 2,
+      "[\"green\", \"gray-blue\", \"brown\"]": 1,
+      "[\"green\", \"gray-blue\", \"cyan\", \"magenta\", \"purple\"]": 1,
+      "[\"green\", \"gray-blue\", \"cyan\"]": 2,
+      "[\"green\", \"gray-blue\", \"olive\"]": 2,
+      "[\"green\", \"gray-blue\", \"orange\"]": 2,
+      "[\"green\", \"gray-blue\", \"purple\"]": 1,
+      "[\"green\", \"gray-blue\", \"red\"]": 1,
+      "[\"green\", \"gray-blue\"]": 7,
+      "[\"green\", \"magenta\", \"amber\"]": 1,
+      "[\"green\", \"magenta\", \"blue\"]": 2,
+      "[\"green\", \"magenta\", \"brown\"]": 2,
+      "[\"green\", \"magenta\", \"cyan\", \"brown\"]": 1,
+      "[\"green\", \"magenta\", \"gray-blue\"]": 1,
+      "[\"green\", \"magenta\", \"olive\", \"brown\"]": 1,
+      "[\"green\", \"magenta\", \"olive\"]": 1,
+      "[\"green\", \"magenta\", \"orange\"]": 1,
+      "[\"green\", \"magenta\", \"purple\"]": 1,
+      "[\"green\", \"magenta\", \"red\"]": 1,
+      "[\"green\", \"magenta\"]": 12,
+      "[\"green\", \"olive\", \"blue\"]": 3,
+      "[\"green\", \"olive\", \"brown\"]": 3,
+      "[\"green\", \"olive\", \"cyan\", \"blue\"]": 1,
+      "[\"green\", \"olive\", \"gray-blue\"]": 2,
+      "[\"green\", \"olive\", \"magenta\"]": 2,
+      "[\"green\", \"olive\", \"purple\"]": 1,
+      "[\"green\", \"olive\", \"red\", \"orange\"]": 1,
+      "[\"green\", \"olive\", \"teal\", \"orange\"]": 1,
+      "[\"green\", \"olive\", \"teal\"]": 1,
+      "[\"green\", \"olive\"]": 7,
+      "[\"green\", \"orange\", \"amber\"]": 1,
+      "[\"green\", \"orange\", \"blue\"]": 2,
+      "[\"green\", \"orange\", \"brown\"]": 1,
+      "[\"green\", \"orange\", \"gray-blue\"]": 3,
+      "[\"green\", \"orange\", \"olive\"]": 1,
+      "[\"green\", \"orange\", \"purple\"]": 1,
+      "[\"green\", \"orange\", \"red\", \"blue\", \"amber\"]": 1,
+      "[\"green\", \"orange\", \"red\", \"gray-blue\"]": 1,
+      "[\"green\", \"orange\", \"red\"]": 1,
+      "[\"green\", \"orange\", \"teal\"]": 1,
+      "[\"green\", \"orange\"]": 10,
+      "[\"green\", \"purple\", \"amber\"]": 2,
+      "[\"green\", \"purple\", \"blue\"]": 1,
+      "[\"green\", \"purple\", \"brown\"]": 1,
+      "[\"green\", \"purple\", \"cyan\"]": 1,
+      "[\"green\", \"purple\", \"gray-blue\"]": 3,
+      "[\"green\", \"purple\", \"magenta\", \"amber\"]": 1,
+      "[\"green\", \"purple\", \"magenta\"]": 3,
+      "[\"green\", \"purple\", \"orange\"]": 2,
+      "[\"green\", \"purple\", \"red\", \"gray-blue\", \"amber\"]": 1,
+      "[\"green\", \"purple\"]": 12,
+      "[\"green\", \"red\", \"amber\", \"teal\"]": 1,
+      "[\"green\", \"red\", \"amber\"]": 1,
+      "[\"green\", \"red\", \"blue\", \"olive\"]": 1,
+      "[\"green\", \"red\", \"cyan\"]": 3,
+      "[\"green\", \"red\", \"gray-blue\", \"brown\"]": 1,
+      "[\"green\", \"red\", \"gray-blue\"]": 1,
+      "[\"green\", \"red\", \"magenta\"]": 1,
+      "[\"green\", \"red\", \"teal\"]": 1,
+      "[\"green\", \"red\"]": 10,
+      "[\"green\", \"teal\", \"blue\"]": 1,
+      "[\"green\", \"teal\", \"brown\"]": 3,
+      "[\"green\", \"teal\", \"cyan\"]": 1,
+      "[\"green\", \"teal\", \"gray-blue\"]": 2,
+      "[\"green\", \"teal\", \"olive\"]": 1,
+      "[\"green\", \"teal\", \"orange\"]": 1,
+      "[\"green\", \"teal\", \"purple\", \"brown\", \"gray-blue\"]": 1,
+      "[\"green\", \"teal\", \"purple\"]": 2,
+      "[\"green\", \"teal\", \"red\"]": 1,
+      "[\"green\", \"teal\"]": 4,
+      "[\"green\"]": 11,
+      "[\"magenta\", \"amber\", \"blue\"]": 3,
+      "[\"magenta\", \"amber\", \"brown\"]": 2,
+      "[\"magenta\", \"amber\", \"olive\"]": 3,
+      "[\"magenta\", \"amber\", \"purple\"]": 3,
+      "[\"magenta\", \"amber\", \"teal\"]": 2,
+      "[\"magenta\", \"amber\"]": 14,
+      "[\"magenta\", \"blue\", \"amber\", \"brown\"]": 1,
+      "[\"magenta\", \"blue\", \"brown\", \"purple\"]": 1,
+      "[\"magenta\", \"blue\", \"brown\"]": 2,
+      "[\"magenta\", \"blue\", \"gray-blue\"]": 2,
+      "[\"magenta\", \"blue\", \"olive\"]": 2,
+      "[\"magenta\", \"blue\", \"orange\"]": 4,
+      "[\"magenta\", \"blue\", \"purple\"]": 3,
+      "[\"magenta\", \"blue\", \"red\"]": 3,
+      "[\"magenta\", \"blue\", \"teal\"]": 1,
+      "[\"magenta\", \"blue\"]": 8,
+      "[\"magenta\", \"brown\", \"blue\"]": 2,
+      "[\"magenta\", \"brown\", \"cyan\"]": 1,
+      "[\"magenta\", \"brown\", \"gray-blue\", \"amber\", \"teal\"]": 1,
+      "[\"magenta\", \"brown\", \"gray-blue\", \"cyan\"]": 1,
+      "[\"magenta\", \"brown\", \"gray-blue\"]": 1,
+      "[\"magenta\", \"brown\", \"green\"]": 1,
+      "[\"magenta\", \"brown\", \"olive\"]": 1,
+      "[\"magenta\", \"brown\", \"orange\", \"cyan\"]": 1,
+      "[\"magenta\", \"brown\", \"orange\"]": 1,
+      "[\"magenta\", \"brown\", \"purple\"]": 1,
+      "[\"magenta\", \"brown\", \"red\", \"green\"]": 1,
+      "[\"magenta\", \"brown\", \"red\"]": 1,
+      "[\"magenta\", \"brown\", \"teal\", \"olive\"]": 1,
+      "[\"magenta\", \"brown\"]": 12,
+      "[\"magenta\", \"cyan\", \"amber\", \"blue\"]": 1,
+      "[\"magenta\", \"cyan\", \"amber\", \"green\"]": 1,
+      "[\"magenta\", \"cyan\", \"amber\"]": 1,
+      "[\"magenta\", \"cyan\", \"blue\", \"olive\"]": 1,
+      "[\"magenta\", \"cyan\", \"blue\", \"orange\"]": 1,
+      "[\"magenta\", \"cyan\", \"blue\"]": 2,
+      "[\"magenta\", \"cyan\", \"brown\"]": 4,
+      "[\"magenta\", \"cyan\", \"gray-blue\", \"olive\"]": 1,
+      "[\"magenta\", \"cyan\", \"gray-blue\"]": 1,
+      "[\"magenta\", \"cyan\", \"green\", \"orange\"]": 1,
+      "[\"magenta\", \"cyan\", \"green\"]": 2,
+      "[\"magenta\", \"cyan\", \"olive\"]": 4,
+      "[\"magenta\", \"cyan\", \"orange\"]": 2,
+      "[\"magenta\", \"cyan\", \"purple\"]": 2,
+      "[\"magenta\", \"cyan\", \"red\", \"amber\"]": 1,
+      "[\"magenta\", \"cyan\", \"red\"]": 3,
+      "[\"magenta\", \"cyan\", \"teal\", \"blue\"]": 1,
+      "[\"magenta\", \"cyan\"]": 9,
+      "[\"magenta\", \"gray-blue\", \"amber\"]": 2,
+      "[\"magenta\", \"gray-blue\", \"brown\", \"blue\"]": 1,
+      "[\"magenta\", \"gray-blue\", \"brown\"]": 2,
+      "[\"magenta\", \"gray-blue\", \"cyan\", \"brown\"]": 1,
+      "[\"magenta\", \"gray-blue\", \"cyan\"]": 1,
+      "[\"magenta\", \"gray-blue\", \"green\"]": 1,
+      "[\"magenta\", \"gray-blue\", \"olive\"]": 1,
+      "[\"magenta\", \"gray-blue\", \"orange\", \"amber\"]": 1,
+      "[\"magenta\", \"gray-blue\", \"purple\"]": 3,
+      "[\"magenta\", \"gray-blue\", \"red\"]": 1,
+      "[\"magenta\", \"gray-blue\", \"teal\"]": 4,
+      "[\"magenta\", \"gray-blue\"]": 5,
+      "[\"magenta\", \"green\", \"amber\"]": 2,
+      "[\"magenta\", \"green\", \"blue\"]": 1,
+      "[\"magenta\", \"green\", \"gray-blue\"]": 2,
+      "[\"magenta\", \"green\", \"olive\"]": 1,
+      "[\"magenta\", \"green\", \"purple\", \"olive\"]": 1,
+      "[\"magenta\", \"green\", \"teal\", \"orange\"]": 1,
+      "[\"magenta\", \"green\", \"teal\"]": 2,
+      "[\"magenta\", \"green\"]": 7,
+      "[\"magenta\", \"olive\", \"blue\"]": 1,
+      "[\"magenta\", \"olive\", \"brown\"]": 3,
+      "[\"magenta\", \"olive\", \"cyan\", \"purple\"]": 1,
+      "[\"magenta\", \"olive\", \"cyan\"]": 1,
+      "[\"magenta\", \"olive\", \"green\"]": 2,
+      "[\"magenta\", \"olive\", \"orange\", \"purple\"]": 1,
+      "[\"magenta\", \"olive\", \"red\"]": 2,
+      "[\"magenta\", \"olive\", \"teal\"]": 1,
+      "[\"magenta\", \"olive\"]": 11,
+      "[\"magenta\", \"orange\", \"amber\"]": 1,
+      "[\"magenta\", \"orange\", \"blue\"]": 2,
+      "[\"magenta\", \"orange\", \"brown\"]": 1,
+      "[\"magenta\", \"orange\", \"cyan\"]": 1,
+      "[\"magenta\", \"orange\", \"gray-blue\", \"blue\"]": 1,
+      "[\"magenta\", \"orange\", \"green\"]": 2,
+      "[\"magenta\", \"orange\", \"olive\", \"blue\"]": 1,
+      "[\"magenta\", \"orange\", \"olive\"]": 1,
+      "[\"magenta\", \"orange\", \"purple\"]": 1,
+      "[\"magenta\", \"orange\", \"red\", \"olive\"]": 1,
+      "[\"magenta\", \"orange\"]": 4,
+      "[\"magenta\", \"purple\", \"blue\"]": 2,
+      "[\"magenta\", \"purple\", \"brown\"]": 1,
+      "[\"magenta\", \"purple\", \"cyan\"]": 2,
+      "[\"magenta\", \"purple\", \"gray-blue\"]": 1,
+      "[\"magenta\", \"purple\", \"green\", \"orange\"]": 1,
+      "[\"magenta\", \"purple\", \"olive\"]": 2,
+      "[\"magenta\", \"purple\", \"orange\"]": 1,
+      "[\"magenta\", \"purple\", \"teal\", \"cyan\"]": 1,
+      "[\"magenta\", \"purple\", \"teal\"]": 2,
+      "[\"magenta\", \"purple\"]": 6,
+      "[\"magenta\", \"red\", \"amber\"]": 1,
+      "[\"magenta\", \"red\", \"gray-blue\"]": 1,
+      "[\"magenta\", \"red\", \"green\", \"amber\"]": 1,
+      "[\"magenta\", \"red\", \"green\"]": 3,
+      "[\"magenta\", \"red\", \"olive\", \"cyan\"]": 1,
+      "[\"magenta\", \"red\", \"olive\"]": 1,
+      "[\"magenta\", \"red\", \"orange\"]": 1,
+      "[\"magenta\", \"red\"]": 7,
+      "[\"magenta\", \"teal\", \"brown\", \"amber\"]": 1,
+      "[\"magenta\", \"teal\", \"brown\"]": 1,
+      "[\"magenta\", \"teal\", \"cyan\"]": 4,
+      "[\"magenta\", \"teal\", \"gray-blue\"]": 2,
+      "[\"magenta\", \"teal\", \"olive\", \"amber\", \"brown\"]": 1,
+      "[\"magenta\", \"teal\", \"orange\"]": 2,
+      "[\"magenta\", \"teal\", \"red\"]": 1,
+      "[\"magenta\", \"teal\"]": 9,
+      "[\"magenta\"]": 10,
+      "[\"olive\", \"amber\", \"brown\"]": 1,
+      "[\"olive\", \"amber\", \"cyan\"]": 2,
+      "[\"olive\", \"amber\", \"gray-blue\"]": 1,
+      "[\"olive\", \"amber\", \"green\"]": 2,
+      "[\"olive\", \"amber\", \"orange\", \"cyan\"]": 1,
+      "[\"olive\", \"amber\", \"orange\"]": 1,
+      "[\"olive\", \"amber\", \"purple\"]": 1,
+      "[\"olive\", \"amber\", \"red\", \"gray-blue\"]": 1,
+      "[\"olive\", \"amber\", \"red\"]": 2,
+      "[\"olive\", \"amber\", \"teal\"]": 2,
+      "[\"olive\", \"amber\"]": 8,
+      "[\"olive\", \"blue\", \"amber\", \"magenta\", \"orange\"]": 1,
+      "[\"olive\", \"blue\", \"amber\"]": 2,
+      "[\"olive\", \"blue\", \"brown\"]": 1,
+      "[\"olive\", \"blue\", \"cyan\", \"magenta\"]": 1,
+      "[\"olive\", \"blue\", \"cyan\"]": 1,
+      "[\"olive\", \"blue\", \"gray-blue\"]": 1,
+      "[\"olive\", \"blue\", \"green\", \"teal\"]": 1,
+      "[\"olive\", \"blue\", \"green\"]": 5,
+      "[\"olive\", \"blue\", \"orange\", \"magenta\", \"cyan\"]": 1,
+      "[\"olive\", \"blue\", \"red\"]": 3,
+      "[\"olive\", \"blue\", \"teal\"]": 2,
+      "[\"olive\", \"blue\"]": 7,
+      "[\"olive\", \"brown\", \"amber\"]": 1,
+      "[\"olive\", \"brown\", \"blue\", \"purple\"]": 1,
+      "[\"olive\", \"brown\", \"blue\"]": 2,
+      "[\"olive\", \"brown\", \"cyan\"]": 2,
+      "[\"olive\", \"brown\", \"gray-blue\"]": 1,
+      "[\"olive\", \"brown\", \"green\"]": 2,
+      "[\"olive\", \"brown\", \"magenta\", \"purple\", \"blue\"]": 1,
+      "[\"olive\", \"brown\", \"magenta\"]": 1,
+      "[\"olive\", \"brown\", \"red\"]": 1,
+      "[\"olive\", \"brown\", \"teal\", \"amber\"]": 1,
+      "[\"olive\", \"brown\"]": 9,
+      "[\"olive\", \"cyan\", \"amber\", \"magenta\"]": 1,
+      "[\"olive\", \"cyan\", \"amber\"]": 1,
+      "[\"olive\", \"cyan\", \"brown\"]": 1,
+      "[\"olive\", \"cyan\", \"gray-blue\", \"amber\"]": 1,
+      "[\"olive\", \"cyan\", \"gray-blue\"]": 1,
+      "[\"olive\", \"cyan\", \"magenta\", \"teal\", \"purple\"]": 1,
+      "[\"olive\", \"cyan\", \"magenta\"]": 1,
+      "[\"olive\", \"cyan\", \"orange\"]": 1,
+      "[\"olive\", \"cyan\", \"purple\"]": 2,
+      "[\"olive\", \"cyan\", \"red\"]": 1,
+      "[\"olive\", \"cyan\", \"teal\"]": 2,
+      "[\"olive\", \"cyan\"]": 7,
+      "[\"olive\", \"gray-blue\", \"blue\", \"amber\", \"cyan\"]": 1,
+      "[\"olive\", \"gray-blue\", \"cyan\"]": 1,
+      "[\"olive\", \"gray-blue\", \"green\"]": 1,
+      "[\"olive\", \"gray-blue\", \"magenta\"]": 4,
+      "[\"olive\", \"gray-blue\", \"purple\"]": 1,
+      "[\"olive\", \"gray-blue\", \"red\"]": 1,
+      "[\"olive\", \"gray-blue\"]": 8,
+      "[\"olive\", \"green\", \"amber\"]": 1,
+      "[\"olive\", \"green\", \"blue\"]": 1,
+      "[\"olive\", \"green\", \"brown\"]": 1,
+      "[\"olive\", \"green\", \"gray-blue\"]": 1,
+      "[\"olive\", \"green\", \"magenta\"]": 2,
+      "[\"olive\", \"green\", \"orange\"]": 1,
+      "[\"olive\", \"green\", \"red\", \"blue\"]": 1,
+      "[\"olive\", \"green\", \"red\"]": 1,
+      "[\"olive\", \"green\", \"teal\", \"magenta\"]": 1,
+      "[\"olive\", \"green\"]": 7,
+      "[\"olive\", \"magenta\", \"blue\"]": 4,
+      "[\"olive\", \"magenta\", \"brown\", \"cyan\"]": 1,
+      "[\"olive\", \"magenta\", \"brown\"]": 1,
+      "[\"olive\", \"magenta\", \"cyan\"]": 1,
+      "[\"olive\", \"magenta\", \"orange\"]": 3,
+      "[\"olive\", \"magenta\", \"purple\", \"blue\"]": 1,
+      "[\"olive\", \"magenta\", \"purple\"]": 3,
+      "[\"olive\", \"magenta\", \"teal\", \"brown\"]": 1,
+      "[\"olive\", \"magenta\", \"teal\"]": 1,
+      "[\"olive\", \"magenta\"]": 5,
+      "[\"olive\", \"orange\", \"amber\"]": 1,
+      "[\"olive\", \"orange\", \"blue\"]": 3,
+      "[\"olive\", \"orange\", \"brown\", \"blue\"]": 1,
+      "[\"olive\", \"orange\", \"brown\"]": 2,
+      "[\"olive\", \"orange\", \"cyan\"]": 3,
+      "[\"olive\", \"orange\", \"gray-blue\"]": 2,
+      "[\"olive\", \"orange\", \"magenta\"]": 3,
+      "[\"olive\", \"orange\", \"purple\", \"brown\"]": 1,
+      "[\"olive\", \"orange\", \"purple\", \"red\", \"gray-blue\"]": 1,
+      "[\"olive\", \"orange\", \"red\", \"magenta\", \"green\"]": 1,
+      "[\"olive\", \"orange\", \"teal\"]": 2,
+      "[\"olive\", \"orange\"]": 4,
+      "[\"olive\", \"purple\", \"amber\"]": 2,
+      "[\"olive\", \"purple\", \"blue\"]": 1,
+      "[\"olive\", \"purple\", \"brown\"]": 1,
+      "[\"olive\", \"purple\", \"gray-blue\", \"brown\"]": 1,
+      "[\"olive\", \"purple\", \"gray-blue\"]": 2,
+      "[\"olive\", \"purple\", \"green\"]": 1,
+      "[\"olive\", \"purple\", \"magenta\"]": 2,
+      "[\"olive\", \"purple\", \"orange\", \"red\"]": 1,
+      "[\"olive\", \"purple\", \"orange\"]": 3,
+      "[\"olive\", \"purple\", \"red\"]": 1,
+      "[\"olive\", \"purple\", \"teal\"]": 1,
+      "[\"olive\", \"purple\"]": 7,
+      "[\"olive\", \"red\", \"amber\", \"purple\"]": 1,
+      "[\"olive\", \"red\", \"amber\"]": 1,
+      "[\"olive\", \"red\", \"blue\"]": 1,
+      "[\"olive\", \"red\", \"brown\"]": 1,
+      "[\"olive\", \"red\", \"gray-blue\"]": 1,
+      "[\"olive\", \"red\", \"green\", \"magenta\"]": 1,
+      "[\"olive\", \"red\", \"green\"]": 1,
+      "[\"olive\", \"red\", \"magenta\"]": 1,
+      "[\"olive\", \"red\", \"orange\"]": 2,
+      "[\"olive\", \"red\", \"purple\"]": 2,
+      "[\"olive\", \"red\"]": 5,
+      "[\"olive\", \"teal\", \"amber\", \"blue\"]": 1,
+      "[\"olive\", \"teal\", \"amber\"]": 1,
+      "[\"olive\", \"teal\", \"blue\", \"magenta\", \"cyan\"]": 1,
+      "[\"olive\", \"teal\", \"blue\"]": 4,
+      "[\"olive\", \"teal\", \"brown\"]": 1,
+      "[\"olive\", \"teal\", \"cyan\"]": 3,
+      "[\"olive\", \"teal\", \"gray-blue\"]": 1,
+      "[\"olive\", \"teal\", \"magenta\", \"cyan\"]": 1,
+      "[\"olive\", \"teal\", \"orange\"]": 2,
+      "[\"olive\", \"teal\", \"red\", \"magenta\"]": 1,
+      "[\"olive\", \"teal\", \"red\"]": 2,
+      "[\"olive\", \"teal\"]": 11,
+      "[\"olive\"]": 9,
+      "[\"orange\", \"amber\", \"blue\", \"gray-blue\"]": 1,
+      "[\"orange\", \"amber\", \"blue\"]": 2,
+      "[\"orange\", \"amber\", \"brown\", \"teal\"]": 1,
+      "[\"orange\", \"amber\", \"brown\"]": 1,
+      "[\"orange\", \"amber\", \"cyan\"]": 1,
+      "[\"orange\", \"amber\", \"gray-blue\"]": 1,
+      "[\"orange\", \"amber\", \"green\", \"gray-blue\"]": 1,
+      "[\"orange\", \"amber\", \"green\"]": 1,
+      "[\"orange\", \"amber\", \"olive\"]": 2,
+      "[\"orange\", \"amber\", \"red\"]": 2,
+      "[\"orange\", \"amber\"]": 8,
+      "[\"orange\", \"blue\", \"brown\"]": 1,
+      "[\"orange\", \"blue\", \"cyan\"]": 3,
+      "[\"orange\", \"blue\", \"gray-blue\"]": 2,
+      "[\"orange\", \"blue\", \"magenta\", \"red\"]": 1,
+      "[\"orange\", \"blue\", \"magenta\"]": 1,
+      "[\"orange\", \"blue\", \"red\"]": 4,
+      "[\"orange\", \"blue\", \"teal\"]": 3,
+      "[\"orange\", \"blue\"]": 6,
+      "[\"orange\", \"brown\", \"blue\"]": 1,
+      "[\"orange\", \"brown\", \"green\"]": 1,
+      "[\"orange\", \"brown\", \"magenta\"]": 1,
+      "[\"orange\", \"brown\", \"olive\"]": 1,
+      "[\"orange\", \"brown\", \"purple\", \"teal\"]": 1,
+      "[\"orange\", \"brown\", \"purple\"]": 1,
+      "[\"orange\", \"brown\", \"red\", \"purple\", \"magenta\"]": 1,
+      "[\"orange\", \"brown\", \"red\"]": 2,
+      "[\"orange\", \"brown\", \"teal\"]": 1,
+      "[\"orange\", \"brown\"]": 7,
+      "[\"orange\", \"cyan\", \"amber\"]": 2,
+      "[\"orange\", \"cyan\", \"blue\"]": 2,
+      "[\"orange\", \"cyan\", \"brown\"]": 4,
+      "[\"orange\", \"cyan\", \"gray-blue\"]": 1,
+      "[\"orange\", \"cyan\", \"green\", \"brown\"]": 1,
+      "[\"orange\", \"cyan\", \"magenta\"]": 3,
+      "[\"orange\", \"cyan\", \"olive\"]": 1,
+      "[\"orange\", \"cyan\", \"purple\", \"blue\"]": 1,
+      "[\"orange\", \"cyan\", \"teal\"]": 2,
+      "[\"orange\", \"cyan\"]": 7,
+      "[\"orange\", \"gray-blue\", \"amber\"]": 4,
+      "[\"orange\", \"gray-blue\", \"cyan\"]": 1,
+      "[\"orange\", \"gray-blue\", \"magenta\"]": 2,
+      "[\"orange\", \"gray-blue\", \"olive\"]": 2,
+      "[\"orange\", \"gray-blue\", \"red\"]": 3,
+      "[\"orange\", \"gray-blue\"]": 10,
+      "[\"orange\", \"green\", \"blue\"]": 1,
+      "[\"orange\", \"green\", \"gray-blue\"]": 1,
+      "[\"orange\", \"green\", \"magenta\"]": 2,
+      "[\"orange\", \"green\", \"purple\", \"brown\"]": 1,
+      "[\"orange\", \"green\", \"teal\", \"brown\"]": 2,
+      "[\"orange\", \"green\", \"teal\"]": 1,
+      "[\"orange\", \"green\"]": 3,
+      "[\"orange\", \"magenta\", \"blue\"]": 1,
+      "[\"orange\", \"magenta\", \"brown\"]": 2,
+      "[\"orange\", \"magenta\", \"cyan\"]": 2,
+      "[\"orange\", \"magenta\", \"olive\"]": 1,
+      "[\"orange\", \"magenta\", \"red\"]": 2,
+      "[\"orange\", \"magenta\", \"teal\"]": 1,
+      "[\"orange\", \"magenta\"]": 6,
+      "[\"orange\", \"olive\", \"amber\"]": 2,
+      "[\"orange\", \"olive\", \"blue\"]": 1,
+      "[\"orange\", \"olive\", \"brown\"]": 3,
+      "[\"orange\", \"olive\", \"cyan\", \"gray-blue\"]": 1,
+      "[\"orange\", \"olive\", \"cyan\"]": 1,
+      "[\"orange\", \"olive\", \"gray-blue\"]": 2,
+      "[\"orange\", \"olive\", \"green\"]": 1,
+      "[\"orange\", \"olive\", \"magenta\"]": 3,
+      "[\"orange\", \"olive\", \"purple\"]": 2,
+      "[\"orange\", \"olive\", \"red\"]": 3,
+      "[\"orange\", \"olive\"]": 6,
+      "[\"orange\", \"purple\", \"brown\"]": 1,
+      "[\"orange\", \"purple\", \"cyan\"]": 2,
+      "[\"orange\", \"purple\", \"gray-blue\"]": 1,
+      "[\"orange\", \"purple\", \"olive\"]": 4,
+      "[\"orange\", \"purple\", \"red\"]": 1,
+      "[\"orange\", \"purple\", \"teal\"]": 1,
+      "[\"orange\", \"purple\"]": 17,
+      "[\"orange\", \"red\", \"blue\"]": 1,
+      "[\"orange\", \"red\", \"brown\"]": 2,
+      "[\"orange\", \"red\", \"cyan\", \"teal\"]": 1,
+      "[\"orange\", \"red\", \"cyan\"]": 3,
+      "[\"orange\", \"red\", \"gray-blue\", \"purple\"]": 1,
+      "[\"orange\", \"red\", \"gray-blue\"]": 1,
+      "[\"orange\", \"red\", \"magenta\"]": 2,
+      "[\"orange\", \"red\", \"purple\"]": 3,
+      "[\"orange\", \"red\", \"teal\"]": 1,
+      "[\"orange\", \"red\"]": 8,
+      "[\"orange\", \"teal\", \"blue\"]": 3,
+      "[\"orange\", \"teal\", \"brown\"]": 1,
+      "[\"orange\", \"teal\", \"cyan\"]": 1,
+      "[\"orange\", \"teal\", \"gray-blue\"]": 3,
+      "[\"orange\", \"teal\", \"green\"]": 2,
+      "[\"orange\", \"teal\", \"magenta\", \"brown\"]": 2,
+      "[\"orange\", \"teal\", \"magenta\"]": 1,
+      "[\"orange\", \"teal\", \"olive\"]": 3,
+      "[\"orange\", \"teal\", \"red\"]": 2,
+      "[\"orange\", \"teal\"]": 10,
+      "[\"orange\"]": 7,
+      "[\"purple\", \"amber\", \"cyan\", \"magenta\"]": 1,
+      "[\"purple\", \"amber\", \"cyan\"]": 1,
+      "[\"purple\", \"amber\", \"gray-blue\"]": 1,
+      "[\"purple\", \"amber\", \"green\", \"teal\"]": 1,
+      "[\"purple\", \"amber\", \"olive\"]": 1,
+      "[\"purple\", \"amber\", \"orange\"]": 1,
+      "[\"purple\", \"amber\", \"teal\"]": 1,
+      "[\"purple\", \"amber\"]": 15,
+      "[\"purple\", \"blue\", \"amber\"]": 1,
+      "[\"purple\", \"blue\", \"brown\", \"cyan\"]": 1,
+      "[\"purple\", \"blue\", \"brown\"]": 1,
+      "[\"purple\", \"blue\", \"green\", \"red\"]": 1,
+      "[\"purple\", \"blue\", \"green\"]": 2,
+      "[\"purple\", \"blue\", \"magenta\"]": 3,
+      "[\"purple\", \"blue\", \"olive\", \"amber\"]": 1,
+      "[\"purple\", \"blue\", \"olive\"]": 2,
+      "[\"purple\", \"blue\", \"orange\", \"magenta\", \"olive\"]": 1,
+      "[\"purple\", \"blue\", \"orange\"]": 2,
+      "[\"purple\", \"blue\", \"red\", \"orange\", \"magenta\"]": 1,
+      "[\"purple\", \"blue\", \"red\"]": 2,
+      "[\"purple\", \"blue\"]": 7,
+      "[\"purple\", \"brown\", \"blue\"]": 2,
+      "[\"purple\", \"brown\", \"cyan\", \"red\"]": 1,
+      "[\"purple\", \"brown\", \"cyan\"]": 1,
+      "[\"purple\", \"brown\", \"gray-blue\"]": 2,
+      "[\"purple\", \"brown\", \"magenta\"]": 3,
+      "[\"purple\", \"brown\", \"red\"]": 1,
+      "[\"purple\", \"brown\", \"teal\"]": 2,
+      "[\"purple\", \"brown\"]": 10,
+      "[\"purple\", \"cyan\", \"amber\", \"orange\"]": 1,
+      "[\"purple\", \"cyan\", \"amber\"]": 1,
+      "[\"purple\", \"cyan\", \"brown\"]": 1,
+      "[\"purple\", \"cyan\", \"gray-blue\"]": 2,
+      "[\"purple\", \"cyan\", \"green\"]": 2,
+      "[\"purple\", \"cyan\", \"magenta\"]": 2,
+      "[\"purple\", \"cyan\", \"olive\"]": 1,
+      "[\"purple\", \"cyan\", \"orange\"]": 1,
+      "[\"purple\", \"cyan\", \"teal\"]": 2,
+      "[\"purple\", \"cyan\"]": 5,
+      "[\"purple\", \"gray-blue\", \"amber\"]": 2,
+      "[\"purple\", \"gray-blue\", \"blue\"]": 2,
+      "[\"purple\", \"gray-blue\", \"cyan\"]": 2,
+      "[\"purple\", \"gray-blue\", \"magenta\", \"green\"]": 1,
+      "[\"purple\", \"gray-blue\", \"magenta\"]": 1,
+      "[\"purple\", \"gray-blue\", \"orange\"]": 1,
+      "[\"purple\", \"gray-blue\", \"red\"]": 4,
+      "[\"purple\", \"gray-blue\"]": 8,
+      "[\"purple\", \"green\", \"blue\"]": 2,
+      "[\"purple\", \"green\", \"brown\"]": 1,
+      "[\"purple\", \"green\", \"gray-blue\"]": 2,
+      "[\"purple\", \"green\", \"magenta\", \"red\"]": 1,
+      "[\"purple\", \"green\", \"magenta\"]": 2,
+      "[\"purple\", \"green\", \"olive\"]": 2,
+      "[\"purple\", \"green\", \"orange\"]": 1,
+      "[\"purple\", \"green\", \"red\"]": 2,
+      "[\"purple\", \"green\", \"teal\"]": 2,
+      "[\"purple\", \"green\"]": 7,
+      "[\"purple\", \"magenta\", \"amber\"]": 1,
+      "[\"purple\", \"magenta\", \"blue\", \"red\"]": 1,
+      "[\"purple\", \"magenta\", \"blue\", \"teal\"]": 1,
+      "[\"purple\", \"magenta\", \"cyan\"]": 2,
+      "[\"purple\", \"magenta\", \"gray-blue\"]": 1,
+      "[\"purple\", \"magenta\", \"olive\"]": 1,
+      "[\"purple\", \"magenta\", \"orange\"]": 1,
+      "[\"purple\", \"magenta\", \"teal\"]": 3,
+      "[\"purple\", \"magenta\"]": 7,
+      "[\"purple\", \"olive\", \"amber\"]": 1,
+      "[\"purple\", \"olive\", \"brown\", \"cyan\", \"teal\"]": 1,
+      "[\"purple\", \"olive\", \"brown\"]": 2,
+      "[\"purple\", \"olive\", \"gray-blue\", \"amber\"]": 1,
+      "[\"purple\", \"olive\", \"gray-blue\"]": 1,
+      "[\"purple\", \"olive\", \"green\", \"magenta\"]": 1,
+      "[\"purple\", \"olive\", \"green\"]": 3,
+      "[\"purple\", \"olive\", \"magenta\"]": 1,
+      "[\"purple\", \"olive\", \"orange\"]": 1,
+      "[\"purple\", \"olive\", \"red\"]": 1,
+      "[\"purple\", \"olive\", \"teal\", \"orange\"]": 1,
+      "[\"purple\", \"olive\"]": 12,
+      "[\"purple\", \"orange\", \"amber\"]": 3,
+      "[\"purple\", \"orange\", \"blue\"]": 3,
+      "[\"purple\", \"orange\", \"brown\"]": 1,
+      "[\"purple\", \"orange\", \"cyan\"]": 1,
+      "[\"purple\", \"orange\", \"gray-blue\"]": 1,
+      "[\"purple\", \"orange\", \"green\"]": 3,
+      "[\"purple\", \"orange\", \"magenta\"]": 1,
+      "[\"purple\", \"orange\", \"olive\"]": 1,
+      "[\"purple\", \"orange\", \"teal\"]": 1,
+      "[\"purple\", \"orange\"]": 4,
+      "[\"purple\", \"red\", \"amber\"]": 1,
+      "[\"purple\", \"red\", \"blue\"]": 3,
+      "[\"purple\", \"red\", \"brown\"]": 1,
+      "[\"purple\", \"red\", \"gray-blue\"]": 1,
+      "[\"purple\", \"red\", \"green\", \"magenta\"]": 1,
+      "[\"purple\", \"red\", \"magenta\"]": 1,
+      "[\"purple\", \"red\", \"orange\", \"green\"]": 1,
+      "[\"purple\", \"red\", \"teal\", \"cyan\"]": 1,
+      "[\"purple\", \"red\", \"teal\"]": 4,
+      "[\"purple\", \"red\"]": 12,
+      "[\"purple\", \"teal\", \"amber\"]": 1,
+      "[\"purple\", \"teal\", \"brown\"]": 1,
+      "[\"purple\", \"teal\", \"gray-blue\"]": 2,
+      "[\"purple\", \"teal\", \"green\"]": 2,
+      "[\"purple\", \"teal\", \"magenta\", \"cyan\"]": 1,
+      "[\"purple\", \"teal\", \"orange\", \"blue\", \"red\"]": 1,
+      "[\"purple\", \"teal\", \"orange\"]": 3,
+      "[\"purple\", \"teal\", \"red\"]": 1,
+      "[\"purple\", \"teal\"]": 8,
+      "[\"purple\"]": 6,
+      "[\"red\", \"amber\", \"blue\"]": 2,
+      "[\"red\", \"amber\", \"brown\"]": 2,
+      "[\"red\", \"amber\", \"cyan\"]": 3,
+      "[\"red\", \"amber\", \"olive\", \"green\"]": 1,
+      "[\"red\", \"amber\", \"olive\"]": 1,
+      "[\"red\", \"amber\", \"purple\", \"cyan\"]": 1,
+      "[\"red\", \"amber\", \"teal\"]": 1,
+      "[\"red\", \"amber\"]": 13,
+      "[\"red\", \"blue\", \"amber\", \"orange\"]": 1,
+      "[\"red\", \"blue\", \"brown\"]": 1,
+      "[\"red\", \"blue\", \"gray-blue\"]": 1,
+      "[\"red\", \"blue\", \"green\", \"brown\"]": 1,
+      "[\"red\", \"blue\", \"green\"]": 2,
+      "[\"red\", \"blue\", \"magenta\", \"orange\"]": 1,
+      "[\"red\", \"blue\", \"magenta\"]": 1,
+      "[\"red\", \"blue\", \"orange\"]": 1,
+      "[\"red\", \"blue\", \"purple\"]": 1,
+      "[\"red\", \"blue\", \"teal\"]": 2,
+      "[\"red\", \"blue\"]": 16,
+      "[\"red\", \"brown\", \"amber\"]": 1,
+      "[\"red\", \"brown\", \"blue\"]": 2,
+      "[\"red\", \"brown\", \"gray-blue\"]": 1,
+      "[\"red\", \"brown\", \"green\"]": 1,
+      "[\"red\", \"brown\", \"magenta\"]": 1,
+      "[\"red\", \"brown\", \"olive\"]": 1,
+      "[\"red\", \"brown\", \"orange\"]": 1,
+      "[\"red\", \"brown\", \"teal\", \"blue\"]": 1,
+      "[\"red\", \"brown\"]": 6,
+      "[\"red\", \"cyan\", \"amber\"]": 1,
+      "[\"red\", \"cyan\", \"blue\"]": 1,
+      "[\"red\", \"cyan\", \"brown\"]": 2,
+      "[\"red\", \"cyan\", \"gray-blue\"]": 1,
+      "[\"red\", \"cyan\", \"green\"]": 4,
+      "[\"red\", \"cyan\", \"magenta\", \"green\"]": 1,
+      "[\"red\", \"cyan\", \"magenta\", \"purple\"]": 1,
+      "[\"red\", \"cyan\", \"magenta\"]": 1,
+      "[\"red\", \"cyan\", \"olive\"]": 1,
+      "[\"red\", \"cyan\", \"teal\"]": 2,
+      "[\"red\", \"cyan\"]": 13,
+      "[\"red\", \"gray-blue\", \"blue\"]": 3,
+      "[\"red\", \"gray-blue\", \"brown\", \"magenta\"]": 1,
+      "[\"red\", \"gray-blue\", \"cyan\", \"teal\"]": 1,
+      "[\"red\", \"gray-blue\", \"magenta\"]": 1,
+      "[\"red\", \"gray-blue\", \"olive\"]": 1,
+      "[\"red\", \"gray-blue\", \"purple\"]": 2,
+      "[\"red\", \"gray-blue\"]": 7,
+      "[\"red\", \"green\", \"amber\", \"blue\"]": 1,
+      "[\"red\", \"green\", \"blue\", \"purple\"]": 2,
+      "[\"red\", \"green\", \"blue\"]": 2,
+      "[\"red\", \"green\", \"brown\"]": 1,
+      "[\"red\", \"green\", \"cyan\"]": 1,
+      "[\"red\", \"green\", \"gray-blue\", \"magenta\"]": 1,
+      "[\"red\", \"green\", \"gray-blue\"]": 1,
+      "[\"red\", \"green\", \"magenta\"]": 2,
+      "[\"red\", \"green\", \"olive\", \"amber\", \"orange\"]": 1,
+      "[\"red\", \"green\", \"olive\", \"amber\"]": 1,
+      "[\"red\", \"green\", \"olive\"]": 1,
+      "[\"red\", \"green\", \"orange\", \"blue\"]": 1,
+      "[\"red\", \"green\", \"orange\"]": 1,
+      "[\"red\", \"green\", \"purple\"]": 1,
+      "[\"red\", \"green\", \"teal\", \"purple\"]": 1,
+      "[\"red\", \"green\"]": 10,
+      "[\"red\", \"magenta\", \"amber\", \"teal\"]": 1,
+      "[\"red\", \"magenta\", \"blue\", \"brown\", \"gray-blue\"]": 1,
+      "[\"red\", \"magenta\", \"gray-blue\"]": 1,
+      "[\"red\", \"magenta\", \"green\"]": 1,
+      "[\"red\", \"magenta\", \"olive\"]": 1,
+      "[\"red\", \"magenta\", \"orange\"]": 1,
+      "[\"red\", \"magenta\", \"purple\"]": 4,
+      "[\"red\", \"magenta\", \"teal\", \"gray-blue\"]": 1,
+      "[\"red\", \"magenta\"]": 6,
+      "[\"red\", \"olive\", \"blue\"]": 1,
+      "[\"red\", \"olive\", \"brown\"]": 1,
+      "[\"red\", \"olive\", \"cyan\"]": 1,
+      "[\"red\", \"olive\", \"gray-blue\", \"cyan\"]": 1,
+      "[\"red\", \"olive\", \"magenta\"]": 1,
+      "[\"red\", \"olive\", \"orange\"]": 1,
+      "[\"red\", \"olive\", \"teal\"]": 2,
+      "[\"red\", \"olive\"]": 9,
+      "[\"red\", \"orange\", \"amber\"]": 2,
+      "[\"red\", \"orange\", \"blue\"]": 2,
+      "[\"red\", \"orange\", \"brown\"]": 1,
+      "[\"red\", \"orange\", \"cyan\"]": 1,
+      "[\"red\", \"orange\", \"green\"]": 4,
+      "[\"red\", \"orange\", \"magenta\"]": 1,
+      "[\"red\", \"orange\", \"purple\", \"cyan\"]": 1,
+      "[\"red\", \"orange\", \"purple\"]": 2,
+      "[\"red\", \"orange\"]": 7,
+      "[\"red\", \"purple\", \"amber\", \"magenta\", \"olive\"]": 1,
+      "[\"red\", \"purple\", \"brown\", \"gray-blue\"]": 1,
+      "[\"red\", \"purple\", \"brown\"]": 1,
+      "[\"red\", \"purple\", \"cyan\"]": 1,
+      "[\"red\", \"purple\", \"magenta\"]": 4,
+      "[\"red\", \"purple\", \"orange\"]": 1,
+      "[\"red\", \"purple\"]": 9,
+      "[\"red\", \"teal\", \"amber\"]": 1,
+      "[\"red\", \"teal\", \"blue\"]": 2,
+      "[\"red\", \"teal\", \"cyan\", \"green\", \"gray-blue\"]": 1,
+      "[\"red\", \"teal\", \"cyan\"]": 2,
+      "[\"red\", \"teal\", \"gray-blue\"]": 3,
+      "[\"red\", \"teal\", \"green\", \"gray-blue\"]": 1,
+      "[\"red\", \"teal\", \"green\"]": 2,
+      "[\"red\", \"teal\", \"olive\"]": 2,
+      "[\"red\", \"teal\", \"purple\"]": 2,
+      "[\"red\", \"teal\"]": 3,
+      "[\"red\"]": 4,
+      "[\"teal\", \"amber\", \"blue\", \"magenta\"]": 1,
+      "[\"teal\", \"amber\", \"blue\"]": 1,
+      "[\"teal\", \"amber\", \"cyan\", \"orange\", \"purple\", \"green\"]": 1,
+      "[\"teal\", \"amber\", \"cyan\", \"purple\"]": 1,
+      "[\"teal\", \"amber\", \"gray-blue\"]": 1,
+      "[\"teal\", \"amber\", \"green\"]": 1,
+      "[\"teal\", \"amber\", \"magenta\", \"purple\"]": 1,
+      "[\"teal\", \"amber\"]": 8,
+      "[\"teal\", \"blue\", \"brown\", \"red\"]": 1,
+      "[\"teal\", \"blue\", \"brown\"]": 3,
+      "[\"teal\", \"blue\", \"gray-blue\"]": 1,
+      "[\"teal\", \"blue\", \"green\"]": 1,
+      "[\"teal\", \"blue\", \"olive\"]": 1,
+      "[\"teal\", \"blue\", \"orange\"]": 2,
+      "[\"teal\", \"blue\", \"red\"]": 1,
+      "[\"teal\", \"blue\"]": 3,
+      "[\"teal\", \"brown\", \"amber\"]": 1,
+      "[\"teal\", \"brown\", \"blue\"]": 2,
+      "[\"teal\", \"brown\", \"gray-blue\", \"purple\"]": 1,
+      "[\"teal\", \"brown\", \"gray-blue\"]": 1,
+      "[\"teal\", \"brown\", \"green\", \"amber\"]": 1,
+      "[\"teal\", \"brown\", \"magenta\"]": 3,
+      "[\"teal\", \"brown\", \"olive\", \"magenta\", \"gray-blue\"]": 1,
+      "[\"teal\", \"brown\", \"orange\", \"olive\", \"red\"]": 1,
+      "[\"teal\", \"brown\", \"orange\"]": 1,
+      "[\"teal\", \"brown\", \"purple\"]": 1,
+      "[\"teal\", \"brown\", \"red\"]": 2,
+      "[\"teal\", \"brown\"]": 6,
+      "[\"teal\", \"cyan\", \"blue\"]": 2,
+      "[\"teal\", \"cyan\", \"brown\"]": 1,
+      "[\"teal\", \"cyan\", \"gray-blue\", \"orange\"]": 1,
+      "[\"teal\", \"cyan\", \"green\", \"magenta\"]": 1,
+      "[\"teal\", \"cyan\", \"green\"]": 2,
+      "[\"teal\", \"cyan\", \"magenta\", \"orange\", \"brown\"]": 1,
+      "[\"teal\", \"cyan\", \"magenta\"]": 1,
+      "[\"teal\", \"cyan\", \"olive\"]": 1,
+      "[\"teal\", \"cyan\", \"purple\"]": 3,
+      "[\"teal\", \"cyan\", \"red\"]": 2,
+      "[\"teal\", \"cyan\"]": 8,
+      "[\"teal\", \"gray-blue\", \"blue\"]": 2,
+      "[\"teal\", \"gray-blue\", \"brown\"]": 1,
+      "[\"teal\", \"gray-blue\", \"cyan\"]": 1,
+      "[\"teal\", \"gray-blue\", \"green\"]": 4,
+      "[\"teal\", \"gray-blue\", \"magenta\", \"green\"]": 1,
+      "[\"teal\", \"gray-blue\", \"olive\"]": 1,
+      "[\"teal\", \"gray-blue\", \"purple\", \"olive\"]": 1,
+      "[\"teal\", \"gray-blue\"]": 7,
+      "[\"teal\", \"green\", \"amber\"]": 2,
+      "[\"teal\", \"green\", \"brown\", \"olive\"]": 1,
+      "[\"teal\", \"green\", \"brown\", \"orange\", \"cyan\"]": 1,
+      "[\"teal\", \"green\", \"brown\"]": 1,
+      "[\"teal\", \"green\", \"cyan\"]": 1,
+      "[\"teal\", \"green\", \"gray-blue\"]": 3,
+      "[\"teal\", \"green\", \"olive\"]": 1,
+      "[\"teal\", \"green\", \"orange\"]": 2,
+      "[\"teal\", \"green\", \"red\"]": 1,
+      "[\"teal\", \"green\"]": 8,
+      "[\"teal\", \"magenta\", \"amber\", \"gray-blue\"]": 1,
+      "[\"teal\", \"magenta\", \"blue\"]": 1,
+      "[\"teal\", \"magenta\", \"brown\"]": 1,
+      "[\"teal\", \"magenta\", \"gray-blue\"]": 5,
+      "[\"teal\", \"magenta\", \"green\", \"orange\", \"blue\"]": 1,
+      "[\"teal\", \"magenta\", \"green\"]": 2,
+      "[\"teal\", \"magenta\", \"olive\", \"cyan\"]": 1,
+      "[\"teal\", \"magenta\", \"purple\"]": 1,
+      "[\"teal\", \"magenta\"]": 7,
+      "[\"teal\", \"olive\", \"amber\", \"gray-blue\"]": 1,
+      "[\"teal\", \"olive\", \"amber\", \"orange\"]": 1,
+      "[\"teal\", \"olive\", \"amber\"]": 4,
+      "[\"teal\", \"olive\", \"brown\"]": 1,
+      "[\"teal\", \"olive\", \"cyan\"]": 2,
+      "[\"teal\", \"olive\", \"gray-blue\", \"amber\"]": 1,
+      "[\"teal\", \"olive\", \"gray-blue\"]": 4,
+      "[\"teal\", \"olive\", \"green\"]": 1,
+      "[\"teal\", \"olive\", \"magenta\", \"amber\"]": 1,
+      "[\"teal\", \"olive\", \"magenta\"]": 1,
+      "[\"teal\", \"olive\", \"orange\"]": 1,
+      "[\"teal\", \"olive\", \"purple\", \"brown\"]": 1,
+      "[\"teal\", \"olive\", \"purple\"]": 1,
+      "[\"teal\", \"olive\"]": 6,
+      "[\"teal\", \"orange\", \"blue\"]": 2,
+      "[\"teal\", \"orange\", \"brown\", \"gray-blue\"]": 1,
+      "[\"teal\", \"orange\", \"brown\"]": 1,
+      "[\"teal\", \"orange\", \"cyan\"]": 1,
+      "[\"teal\", \"orange\", \"gray-blue\"]": 1,
+      "[\"teal\", \"orange\", \"green\"]": 2,
+      "[\"teal\", \"orange\", \"olive\", \"gray-blue\", \"blue\"]": 1,
+      "[\"teal\", \"orange\", \"olive\"]": 2,
+      "[\"teal\", \"orange\", \"purple\"]": 3,
+      "[\"teal\", \"orange\"]": 5,
+      "[\"teal\", \"purple\", \"amber\", \"red\"]": 1,
+      "[\"teal\", \"purple\", \"blue\"]": 1,
+      "[\"teal\", \"purple\", \"brown\"]": 2,
+      "[\"teal\", \"purple\", \"cyan\", \"red\"]": 1,
+      "[\"teal\", \"purple\", \"cyan\"]": 2,
+      "[\"teal\", \"purple\", \"gray-blue\"]": 4,
+      "[\"teal\", \"purple\", \"green\", \"olive\"]": 1,
+      "[\"teal\", \"purple\", \"green\"]": 1,
+      "[\"teal\", \"purple\", \"magenta\"]": 1,
+      "[\"teal\", \"purple\", \"olive\"]": 1,
+      "[\"teal\", \"purple\", \"orange\"]": 3,
+      "[\"teal\", \"purple\", \"red\"]": 2,
+      "[\"teal\", \"purple\"]": 6,
+      "[\"teal\", \"red\", \"blue\", \"purple\"]": 1,
+      "[\"teal\", \"red\", \"blue\"]": 1,
+      "[\"teal\", \"red\", \"brown\", \"cyan\", \"green\"]": 1,
+      "[\"teal\", \"red\", \"brown\"]": 2,
+      "[\"teal\", \"red\", \"cyan\"]": 1,
+      "[\"teal\", \"red\", \"gray-blue\"]": 1,
+      "[\"teal\", \"red\", \"green\"]": 1,
+      "[\"teal\", \"red\", \"magenta\"]": 2,
+      "[\"teal\", \"red\", \"olive\"]": 2,
+      "[\"teal\", \"red\", \"orange\", \"purple\"]": 1,
+      "[\"teal\", \"red\", \"orange\"]": 1,
+      "[\"teal\", \"red\", \"purple\"]": 2,
+      "[\"teal\", \"red\"]": 6,
+      "[\"teal\"]": 8
+    },
+    "lines_at_target": {
+      "1": 93,
+      "2": 1090,
+      "3": 1506,
+      "4": 247,
+      "5": 63,
+      "6": 1
+    },
+    "reached": {
+      "[\"A\", \"B\", \"C\", \"D\"]": 1,
+      "[\"A\", \"B\", \"C\"]": 111,
+      "[\"A\", \"B\", \"D\"]": 101,
+      "[\"A\", \"B\", \"E\"]": 37,
+      "[\"A\", \"B\", \"F\"]": 11,
+      "[\"A\", \"B\"]": 201,
+      "[\"A\", \"C\", \"D\", \"E\"]": 1,
+      "[\"A\", \"C\", \"D\"]": 98,
+      "[\"A\", \"C\", \"E\"]": 30,
+      "[\"A\", \"C\", \"F\"]": 8,
+      "[\"A\", \"C\"]": 202,
+      "[\"A\", \"D\", \"E\"]": 26,
+      "[\"A\", \"D\", \"F\"]": 16,
+      "[\"A\", \"D\"]": 191,
+      "[\"A\", \"E\", \"F\"]": 10,
+      "[\"A\", \"E\"]": 114,
+      "[\"A\", \"F\"]": 43,
+      "[\"A\"]": 107,
+      "[\"B\", \"C\", \"D\", \"E\"]": 1,
+      "[\"B\", \"C\", \"D\"]": 103,
+      "[\"B\", \"C\", \"E\"]": 31,
+      "[\"B\", \"C\", \"F\"]": 7,
+      "[\"B\", \"C\"]": 205,
+      "[\"B\", \"D\", \"E\"]": 35,
+      "[\"B\", \"D\", \"F\"]": 12,
+      "[\"B\", \"D\"]": 198,
+      "[\"B\", \"E\", \"F\"]": 13,
+      "[\"B\", \"E\"]": 82,
+      "[\"B\", \"F\"]": 46,
+      "[\"B\"]": 118,
+      "[\"C\", \"D\", \"E\"]": 38,
+      "[\"C\", \"D\", \"F\"]": 14,
+      "[\"C\", \"D\"]": 192,
+      "[\"C\", \"E\", \"F\"]": 14,
+      "[\"C\", \"E\"]": 93,
+      "[\"C\", \"F\"]": 37,
+      "[\"C\"]": 102,
+      "[\"D\", \"E\", \"F\"]": 10,
+      "[\"D\", \"E\"]": 105,
+      "[\"D\", \"F\"]": 40,
+      "[\"D\"]": 94,
+      "[\"E\", \"F\"]": 36,
+      "[\"E\"]": 46,
+      "[\"F\"]": 20
+    },
+    "target": {
+      "A": 609,
+      "B": 582,
+      "C": 630,
+      "D": 636,
+      "E": 397,
+      "F": 146
+    },
+    "target_degree": {
+      "1": 93,
+      "2": 1090,
+      "3": 1506,
+      "4": 247,
+      "5": 63,
+      "6": 1
+    },
+    "unreached": {
+      "[\"A\", \"B\", \"C\", \"D\"]": 9,
+      "[\"A\", \"B\", \"C\", \"E\"]": 10,
+      "[\"A\", \"B\", \"C\", \"F\"]": 10,
+      "[\"A\", \"B\", \"C\"]": 33,
+      "[\"A\", \"B\", \"D\", \"E\"]": 9,
+      "[\"A\", \"B\", \"D\", \"F\"]": 12,
+      "[\"A\", \"B\", \"D\"]": 45,
+      "[\"A\", \"B\", \"E\", \"F\"]": 9,
+      "[\"A\", \"B\", \"E\"]": 53,
+      "[\"A\", \"B\", \"F\"]": 28,
+      "[\"A\", \"B\"]": 106,
+      "[\"A\", \"C\", \"D\", \"E\"]": 9,
+      "[\"A\", \"C\", \"D\", \"F\"]": 10,
+      "[\"A\", \"C\", \"D\"]": 44,
+      "[\"A\", \"C\", \"E\", \"F\"]": 10,
+      "[\"A\", \"C\", \"E\"]": 56,
+      "[\"A\", \"C\", \"F\"]": 44,
+      "[\"A\", \"C\"]": 95,
+      "[\"A\", \"D\", \"E\", \"F\"]": 14,
+      "[\"A\", \"D\", \"E\"]": 44,
+      "[\"A\", \"D\", \"F\"]": 28,
+      "[\"A\", \"D\"]": 106,
+      "[\"A\", \"E\", \"F\"]": 26,
+      "[\"A\", \"E\"]": 64,
+      "[\"A\", \"F\"]": 19,
+      "[\"A\"]": 190,
+      "[\"B\", \"C\", \"D\", \"E\"]": 12,
+      "[\"B\", \"C\", \"D\", \"F\"]": 7,
+      "[\"B\", \"C\", \"D\"]": 46,
+      "[\"B\", \"C\", \"E\", \"F\"]": 7,
+      "[\"B\", \"C\", \"E\"]": 41,
+      "[\"B\", \"C\", \"F\"]": 38,
+      "[\"B\", \"C\"]": 112,
+      "[\"B\", \"D\", \"E\", \"F\"]": 17,
+      "[\"B\", \"D\", \"E\"]": 43,
+      "[\"B\", \"D\", \"F\"]": 36,
+      "[\"B\", \"D\"]": 93,
+      "[\"B\", \"E\", \"F\"]": 36,
+      "[\"B\", \"E\"]": 79,
+      "[\"B\", \"F\"]": 20,
+      "[\"B\"]": 194,
+      "[\"C\", \"D\", \"E\", \"F\"]": 7,
+      "[\"C\", \"D\", \"E\"]": 45,
+      "[\"C\", \"D\", \"F\"]": 22,
+      "[\"C\", \"D\"]": 116,
+      "[\"C\", \"E\", \"F\"]": 26,
+      "[\"C\", \"E\"]": 77,
+      "[\"C\", \"F\"]": 12,
+      "[\"C\"]": 184,
+      "[\"D\", \"E\", \"F\"]": 30,
+      "[\"D\", \"E\"]": 91,
+      "[\"D\", \"F\"]": 14,
+      "[\"D\"]": 179,
+      "[\"E\", \"F\"]": 21,
+      "[\"E\"]": 51,
+      "[]": 261
+    }
+  },
+  "connected_label_pairs_out_of_canonical_15": {
+    "2": 4,
+    "3": 96,
+    "4": 405,
+    "5": 809,
+    "6": 755,
+    "7": 484,
+    "8": 301,
+    "9": 119,
+    "10": 26,
+    "11": 1
+  },
+  "constant_answer_baselines": {
+    "answer": 0.913,
+    "colors": 0.006666666666666667,
+    "lines_at_target": 0.502,
+    "reached": 0.06833333333333333,
+    "target": 0.212,
+    "target_degree": 0.502,
+    "unreached": 0.087
+  },
+  "derivation_mismatches": [],
+  "fallback_count": 404,
+  "fields_over_60_percent": {
+    "answer": 0.913
+  },
+  "issues": [],
+  "no_answer_leakage": true,
+  "no_coordinate_prompts": true,
+  "no_existing_question_paraphrase": true,
+  "png_recoverability": {
+    "passed": 3000,
+    "total": 3000
+  },
+  "preferred_degree_2_or_3": 2596,
+  "public_columns": [
+    "question_id",
+    "image",
+    "prompt"
+  ],
+  "question_count": 3000,
+  "selection_distribution": {
+    "target": {
+      "A": 609,
+      "B": 582,
+      "C": 630,
+      "D": 636,
+      "E": 397,
+      "F": 146
+    },
+    "target_degree": {
+      "1": 93,
+      "2": 1090,
+      "3": 1506,
+      "4": 247,
+      "5": 63,
+      "6": 1
+    }
+  }
+}

Dataset/route_dataset_3000/open_validation_report.txt

@@ -0,0 +1,12 @@
+ROUTE OPEN-QUESTION VALIDATION REPORT
+=====================================
+Questions: 3000
+Derivation mismatches: 0
+PNG-recoverable items: 3000/3000
+Selected target degrees: {1: 93, 2: 1090, 3: 1506, 4: 247, 5: 63, 6: 1}
+Connected-pair diagnostic (of canonical 15): {2: 4, 3: 96, 4: 405, 5: 809, 6: 755, 7: 484, 8: 301, 9: 119, 10: 26, 11: 1}
+Constant baselines: {'target': 0.212, 'target_degree': 0.502, 'lines_at_target': 0.502, 'colors': 0.006666666666666667, 'reached': 0.06833333333333333, 'unreached': 0.087, 'answer': 0.913}
+Fields over 60%: {'answer': 0.913}
+Issues: 0
+None
+Summary: PASS

Dataset/route_dataset_3000/validate_open_questions.py

@@ -0,0 +1,263 @@
+"""Independently validate route open questions, answers, derivations, and PNG evidence."""
+from __future__ import annotations
+
+import csv
+import json
+import math
+import random
+import re
+from collections import Counter
+from pathlib import Path
+
+from PIL import Image
+
+ROOT = Path(__file__).resolve().parent
+PUBLIC_COLUMNS = ["question_id", "image", "prompt"]
+ANSWER_COLUMNS = [
+    "question_id", "image", "target", "target_position", "target_degree",
+    "lines_at_target", "colors", "reached", "unreached", "answer",
+]
+COLOR_HEX = {
+    "teal": "#009E9A", "orange": "#F05A17", "blue": "#0878BD", "green": "#149B43",
+    "red": "#D7192D", "purple": "#9256C2", "brown": "#7A4938", "amber": "#C77A00",
+    "magenta": "#D51A78", "cyan": "#00A8C2", "olive": "#799514", "gray-blue": "#557383",
+}
+
+
+def read_csv(path: Path) -> list[dict[str, str]]:
+    with path.open(encoding="utf-8-sig", newline="") as handle:
+        return list(csv.DictReader(handle))
+
+
+def endpoints(record: dict) -> dict[str, dict]:
+    width, height = record["canvas_size"]
+    margin = 64
+    inner_width, inner_height = width - 2 * margin, height - 2 * margin
+    perimeter = 2 * (inner_width + inner_height)
+    result = {}
+    for index, letter in enumerate(record["endpoint_letters"]):
+        distance = (inner_width / 2 + index * perimeter / record["num_endpoints"]) % perimeter
+        if distance < inner_width:
+            anchor, side = [round(margin + distance), margin], "top"
+        elif distance < inner_width + inner_height:
+            anchor, side = [width - margin, round(margin + distance - inner_width)], "right"
+        elif distance < 2 * inner_width + inner_height:
+            anchor, side = [round(width - margin - (distance - inner_width - inner_height)), height - margin], "bottom"
+        else:
+            anchor, side = [margin, round(height - margin - (distance - 2 * inner_width - inner_height))], "left"
+        result[letter] = {"anchor": anchor, "side": side}
+    return result
+
+
+def position_phrase(endpoint: dict, canvas: list[int]) -> str:
+    x, y = endpoint["anchor"]
+    width, height = canvas
+    side = endpoint["side"]
+    if side in {"top", "bottom"}:
+        horizontal = " left" if x < width * .38 else (" right" if x > width * .62 else "")
+        return f"at the {side}{horizontal}"
+    vertical = "upper " if y < height * .38 else ("lower " if y > height * .62 else "")
+    return f"on the {vertical}{side}" if vertical else f"on the {side}"
+
+
+def derive(record: dict) -> dict:
+    degrees = {
+        letter: sum(letter in (route["start"], route["end"]) for route in record["routes"])
+        for letter in record["endpoint_letters"]
+    }
+    candidates = [letter for letter in record["endpoint_letters"] if degrees[letter] in (2, 3)]
+    if not candidates:
+        positive = [value for value in degrees.values() if value >= 1]
+        if not positive:
+            raise ValueError("no positive-degree target")
+        minimum = min(positive)
+        candidates = [letter for letter in record["endpoint_letters"] if degrees[letter] == minimum]
+    target = random.Random(int(record["seed"])).choice(candidates)
+    incident = [route for route in record["routes"] if target in (route["start"], route["end"])]
+    far_ends = [route["end"] if route["start"] == target else route["start"] for route in incident]
+    reached = sorted(set(far_ends) - {target})
+    unreached = sorted(set(record["endpoint_letters"]) - {target} - set(reached))
+    return {
+        "target": target,
+        "target_position": position_phrase(endpoints(record)[target], record["canvas_size"]),
+        "target_degree": len(incident),
+        "lines_at_target": len(incident),
+        "colors": [route["color"] for route in incident],
+        "reached": reached,
+        "unreached": unreached,
+        "answer": "connected to every other labelled side" if not unreached else "some labelled sides remain unreached",
+        "incident": incident,
+        "degrees": degrees,
+    }
+
+
+def rgb(hex_value: str) -> tuple[int, int, int]:
+    value = hex_value.lstrip("#")
+    return tuple(int(value[index:index + 2], 16) for index in (0, 2, 4))
+
+
+def close(pixel: tuple[int, ...], target: tuple[int, int, int], tolerance: int = 85) -> bool:
+    return max(abs(pixel[index] - target[index]) for index in range(3)) <= tolerance
+
+
+def color_near(image: Image.Image, point: list[int], target: tuple[int, int, int], radius: int = 4) -> bool:
+    x0, y0 = point
+    return any(
+        close(image.getpixel((x, y)), target)
+        for y in range(max(0, y0 - radius), min(image.height, y0 + radius + 1))
+        for x in range(max(0, x0 - radius), min(image.width, x0 + radius + 1))
+    )
+
+
+def endpoint_segment_visible(image: Image.Image, points: list[list[int]], target: tuple[int, int, int], at_start: bool) -> bool:
+    endpoint, inward = (points[0], points[1]) if at_start else (points[-1], points[-2])
+    dx, dy = inward[0] - endpoint[0], inward[1] - endpoint[1]
+    length = max(1.0, math.hypot(dx, dy))
+    probes = [
+        [round(endpoint[0] + dx * distance / length), round(endpoint[1] + dy * distance / length)]
+        for distance in (0, 2, 5, 10, 16, 24, 35)
+    ]
+    return any(color_near(image, point, target, 6) for point in probes)
+
+
+def label_center(endpoint: dict) -> tuple[int, int]:
+    x, y = endpoint["anchor"]
+    side = endpoint["side"]
+    if side == "top":
+        return x, y - 39
+    if side == "bottom":
+        return x, y + 38
+    if side == "left":
+        return x - 39, y
+    return x + 39, y
+
+
+def label_visible(image: Image.Image, endpoint: dict) -> bool:
+    x0, y0 = label_center(endpoint)
+    ink = (16, 23, 25)
+    return any(
+        close(image.getpixel((x, y)), ink, 55)
+        for y in range(max(0, y0 - 25), min(image.height, y0 + 26))
+        for x in range(max(0, x0 - 25), min(image.width, x0 + 26))
+    )
+
+
+def distribution(values) -> dict[str, int]:
+    return dict(sorted(Counter(json.dumps(value, ensure_ascii=False, sort_keys=True) if isinstance(value, (list, dict)) else str(value) for value in values).items()))
+
+
+def baseline(counts: dict[str, int]) -> float:
+    return max(counts.values()) / sum(counts.values()) if counts else 0.0
+
+
+def main() -> None:
+    records = [json.loads(line) for line in (ROOT / "annotations.jsonl").read_text(encoding="utf-8").splitlines() if line]
+    public = read_csv(ROOT / "open_questions.csv")
+    answers = read_csv(ROOT / "open_answer_key.csv")
+    annotations = [json.loads(line) for line in (ROOT / "open_annotations.jsonl").read_text(encoding="utf-8").splitlines() if line]
+    issues, derivation_mismatches = [], []
+    png_pass = 0
+    selected_degrees, targets, connected_pairs = Counter(), Counter(), Counter()
+    values = {key: [] for key in ["target", "target_degree", "lines_at_target", "colors", "reached", "unreached", "answer"]}
+    if len(records) != len(public) or len(public) != len(answers) or len(answers) != len(annotations):
+        issues.append("row-count mismatch among source/open files")
+    if public and list(public[0]) != PUBLIC_COLUMNS:
+        issues.append(f"public columns {list(public[0])} != {PUBLIC_COLUMNS}")
+    if answers and list(answers[0]) != ANSWER_COLUMNS:
+        issues.append(f"answer columns {list(answers[0])} != {ANSWER_COLUMNS}")
+    for position, (record, question, answer, annotation) in enumerate(zip(records, public, answers, annotations), 1):
+        expected = derive(record)
+        qid = f"{record['id']}_open_q1"
+        parsed = {
+            "target": answer["target"], "target_position": answer["target_position"],
+            "target_degree": int(answer["target_degree"]), "lines_at_target": int(answer["lines_at_target"]),
+            "colors": json.loads(answer["colors"]), "reached": json.loads(answer["reached"]),
+            "unreached": json.loads(answer["unreached"]), "answer": answer["answer"],
+        }
+        for key in parsed:
+            if parsed[key] != expected[key]:
+                derivation_mismatches.append(f"{record['id']}/{key}")
+        if question["question_id"] != qid or answer["question_id"] != qid or annotation["question_id"] != qid:
+            issues.append(f"{record['id']}: question_id mismatch")
+        if question["image"] != Path(record["image_path"]).name or answer["image"] != question["image"]:
+            issues.append(f"{record['id']}: image mismatch")
+        prompt = question["prompt"]
+        if re.search(r"\(\s*-?\d+\s*,\s*-?\d+\s*\)", prompt) or "coordinate" in prompt.lower() or "axis" in prompt.lower():
+            issues.append(f"{record['id']}: prompt contains coordinates")
+        named_colors = {name for name in COLOR_HEX if re.search(rf"\b{re.escape(name)}\b", prompt.lower())}
+        if not named_colors <= set(record["colors_used"]):
+            issues.append(f"{record['id']}: prompt names absent color")
+        if any(prompt.strip() == old["question_text"].strip() for old in record["questions"]):
+            issues.append(f"{record['id']}: exact existing-question paraphrase")
+        if not all(fragment in prompt.lower() for fragment in ["far end", "distinguished each line by colour", "confidence score"]):
+            issues.append(f"{record['id']}: open task lacks required multi-part reasoning")
+        if "most routes" in prompt.lower() or "maximum endpoint degree" in prompt.lower():
+            issues.append(f"{record['id']}: reduces to degree-ranking task")
+        leaked_color = any(re.search(rf"\b{re.escape(color)}\b", prompt.lower()) for color in expected["colors"])
+        leaked_label = any(
+            re.search(rf"\b{re.escape(label)}\b", prompt)
+            for label in set(expected["reached"] + expected["unreached"]) - {expected["target"]}
+        )
+        if leaked_color or leaked_label:
+            issues.append(f"{record['id']}: prompt leaks an answer-specific color/label")
+        if annotation.get("target") != expected["target"] or annotation.get("derivation", {}).get("degrees") != expected["degrees"]:
+            issues.append(f"{record['id']}: annotation derivation mismatch")
+        selected_degrees[expected["target_degree"]] += 1
+        targets[expected["target"]] += 1
+        pairs = {"".join(sorted((route["start"], route["end"]))) for route in record["routes"] if route["start"] != route["end"]}
+        connected_pairs[len(pairs)] += 1
+        for key in values:
+            values[key].append(expected[key])
+        image_path = ROOT / record["image_path"]
+        with Image.open(image_path) as source:
+            image = source.convert("RGB")
+            image.load()
+        layout = endpoints(record)
+        png_ok = image.size == tuple(record["canvas_size"]) and all(label_visible(image, layout[label]) for label in record["endpoint_letters"])
+        for route in expected["incident"]:
+            target_color = rgb(route.get("hex") or COLOR_HEX[route["color"]])
+            if not endpoint_segment_visible(image, route["points"], target_color, True) or not endpoint_segment_visible(image, route["points"], target_color, False):
+                png_ok = False
+        if png_ok:
+            png_pass += 1
+        else:
+            issues.append(f"{record['id']}: PNG target/path/endpoint recovery")
+        if position % 500 == 0:
+            print(f"Validated {position}/{len(records)}", flush=True)
+    if derivation_mismatches:
+        issues.append(f"independent derivation mismatches: {len(derivation_mismatches)}")
+    distributions = {key: distribution(field_values) for key, field_values in values.items()}
+    baselines = {key: baseline(counts) for key, counts in distributions.items()}
+    metrics = {
+        "question_count": len(public),
+        "selection_distribution": {"target": dict(sorted(targets.items())), "target_degree": dict(sorted(selected_degrees.items()))},
+        "preferred_degree_2_or_3": sum(selected_degrees[value] for value in (2, 3)),
+        "fallback_count": len(records) - sum(selected_degrees[value] for value in (2, 3)),
+        "connected_label_pairs_out_of_canonical_15": dict(sorted(connected_pairs.items())),
+        "answer_distributions": distributions,
+        "constant_answer_baselines": baselines,
+        "fields_over_60_percent": {key: value for key, value in baselines.items() if value > .60},
+        "derivation_mismatches": derivation_mismatches,
+        "png_recoverability": {"passed": png_pass, "total": len(records)},
+        "public_columns": PUBLIC_COLUMNS,
+        "no_coordinate_prompts": not any("prompt contains coordinates" in issue for issue in issues),
+        "no_answer_leakage": not any("prompt leaks" in issue for issue in issues),
+        "no_existing_question_paraphrase": not any("paraphrase" in issue or "degree-ranking" in issue for issue in issues),
+        "issues": issues,
+    }
+    (ROOT / "open_validation_metrics.json").write_text(json.dumps(metrics, indent=2, ensure_ascii=False, sort_keys=True) + "\n", encoding="utf-8")
+    report = [
+        "ROUTE OPEN-QUESTION VALIDATION REPORT", "=" * 37,
+        f"Questions: {len(public)}", f"Derivation mismatches: {len(derivation_mismatches)}",
+        f"PNG-recoverable items: {png_pass}/{len(records)}", f"Selected target degrees: {dict(sorted(selected_degrees.items()))}",
+        f"Connected-pair diagnostic (of canonical 15): {dict(sorted(connected_pairs.items()))}",
+        f"Constant baselines: {baselines}", f"Fields over 60%: {metrics['fields_over_60_percent']}",
+        f"Issues: {len(issues)}", *(issues if issues else ["None"]), f"Summary: {'PASS' if not issues else 'FAIL'}",
+    ]
+    (ROOT / "open_validation_report.txt").write_text("\n".join(report) + "\n", encoding="utf-8")
+    print("\n".join(report))
+    raise SystemExit(1 if issues else 0)
+
+
+if __name__ == "__main__":
+    main()

Dataset/route_dataset_3000/validation_metrics.json

@@ -48,7 +48,7 @@
       "p95": 12.0
     }
   },
-  "dataset_version": "route-2.0.0",
+  "dataset_version": "route-3.0.0",
   "degeneracy_flags": {},
   "features_at_v_ge_0_10": {
     "crossing_count": {

Dataset/verify_open_question_release.py

@@ -0,0 +1,119 @@
+"""Final invariants and release report for the route/hex supplementary open sets."""
+from __future__ import annotations
+
+import csv
+import hashlib
+import json
+import subprocess
+from pathlib import Path
+
+
+REPO = Path(__file__).resolve().parents[1]
+DATASETS = REPO / "Dataset"
+DOMAINS = ("route_dataset_3000", "hex_pathfinding_dataset_3000")
+OUT = DATASETS / "open_question_release_report.json"
+
+
+def sha256(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as handle:
+        for chunk in iter(lambda: handle.read(1024 * 1024), b""):
+            digest.update(chunk)
+    return digest.hexdigest()
+
+
+def head_bytes(relative: str) -> bytes:
+    raw = subprocess.check_output(["git", "show", f"HEAD:{relative}"], cwd=REPO)
+    if raw.startswith(b"version https://git-lfs.github.com/spec/v1"):
+        raw = subprocess.run(
+            ["git", "lfs", "smudge"], cwd=REPO, input=raw,
+            stdout=subprocess.PIPE, check=True,
+        ).stdout
+    return raw
+
+
+def annotations_without_version(raw: bytes) -> list[dict]:
+    records = []
+    for line in raw.decode("utf-8-sig").splitlines():
+        if line:
+            record = json.loads(line)
+            record.pop("dataset_version", None)
+            records.append(record)
+    return records
+
+
+def csv_count(path: Path) -> int:
+    with path.open(encoding="utf-8-sig", newline="") as handle:
+        return sum(1 for _ in csv.DictReader(handle))
+
+
+def main() -> None:
+    before = json.loads((DATASETS / "open_question_prechange_hashes.json").read_text(encoding="utf-8"))
+    domains: dict[str, dict] = {}
+    for name in DOMAINS:
+        folder = DATASETS / name
+        expected = before[name]
+        unchanged_hashes = {
+            "question_set.csv": sha256(folder / "question_set.csv") == expected["question_set_sha256"],
+            "answer_key.csv": sha256(folder / "answer_key.csv") == expected["answer_key_sha256"],
+            "dataset_final.csv": sha256(folder / "dataset_final.csv") == expected["dataset_final_csv_sha256"],
+            "dataset_final.jsonl": sha256(folder / "dataset_final.jsonl") == expected["dataset_final_jsonl_sha256"],
+        }
+        relative_annotations = (Path("Dataset") / name / "annotations.jsonl").as_posix()
+        current_annotations = (folder / "annotations.jsonl").read_bytes()
+        annotation_payload_unchanged = (
+            annotations_without_version(head_bytes(relative_annotations))
+            == annotations_without_version(current_annotations)
+        )
+        image_status = subprocess.check_output(
+            ["git", "status", "--porcelain", "--", str(Path("Dataset") / name / "images")],
+            cwd=REPO, text=True,
+        ).strip()
+        metrics = json.loads((folder / "open_validation_metrics.json").read_text(encoding="utf-8"))
+        if not all(unchanged_hashes.values()) or not annotation_payload_unchanged or image_status:
+            raise RuntimeError(f"{name}: protected existing content changed")
+        if metrics["issues"] or metrics["derivation_mismatches"]:
+            raise RuntimeError(f"{name}: open validation did not pass")
+        domains[name] = {
+            "dataset_version": json.loads((folder / "build_manifest.json").read_text(encoding="utf-8"))["dataset_version"],
+            "open_questions": csv_count(folder / "open_questions.csv"),
+            "open_answers": csv_count(folder / "open_answer_key.csv"),
+            "derivation_mismatches": len(metrics["derivation_mismatches"]),
+            "png_recoverability": metrics["png_recoverability"],
+            "constant_answer_baselines": metrics["constant_answer_baselines"],
+            "fields_over_60_percent": metrics["fields_over_60_percent"],
+            "protected_file_hashes_unchanged": unchanged_hashes,
+            "annotation_payload_unchanged_except_version": annotation_payload_unchanged,
+            "images_unchanged": not bool(image_status),
+        }
+        if name.startswith("route"):
+            domains[name]["selection_distribution"] = metrics["selection_distribution"]
+            domains[name]["connected_label_pairs_out_of_canonical_15"] = metrics[
+                "connected_label_pairs_out_of_canonical_15"
+            ]
+        else:
+            domains[name]["boundary_vs_interior"] = metrics["boundary_vs_interior"]
+            domains[name]["template_choice"] = metrics["template_choice"]
+            domains[name]["answer_distributions"] = metrics["answer_distributions"]
+
+    combined = json.loads((DATASETS / "combined_open_question_report.json").read_text(encoding="utf-8"))
+    suite = json.loads((DATASETS / "combined_suite_rebuild_report.json").read_text(encoding="utf-8"))["actual_totals"]
+    report = {
+        "status": "PASS",
+        "scope": list(DOMAINS),
+        "domains": domains,
+        "combined_open": combined,
+        "five_level_suite": {
+            "datasets": 34,
+            "images": suite["images"],
+            "questions": suite["questions"],
+            "answers": suite["answers"],
+        },
+        "protected_content": "images, renderers, existing five-level question text, and existing answers unchanged",
+    }
+    OUT.write_text(json.dumps(report, indent=2) + "\n", encoding="utf-8")
+    print(json.dumps(report, indent=2))
+
+
+if __name__ == "__main__":
+    main()

README.md

@@ -48,7 +48,7 @@ This repository generates datasets and ground truth. It does not run models, sco
 
 | Dataset | Geometry class | Version | Images | Questions | Validation | Key skill tested |
 |---|---|---|---:|---:|---|---|
-| [route](Dataset/route_dataset_3000/) | Topology / Graph Theory | route-2.0.0 | 3,000 | 15,000 | PASS — 0 mismatches | Trace colored routes and reason over endpoint connectivity |
+| [route](Dataset/route_dataset_3000/) | Topology / Graph Theory | route-3.0.0 | 3,000 | 15,000 + 3,000 open | PASS — 0 mismatches | Trace colored routes and reason over endpoint connectivity |
 | [nested_squares](Dataset/nested_squares_dataset_3000/) | Transformational Geometry | nested-squares-8.0.0 | 3,000 | 15,000 | PASS — 0 mismatches | Measure center drift, size ratio, four-fold rotation, and geometric visibility |
 | [nested_triangles](Dataset/nested_triangles_dataset_3000/) | Transformational Geometry | nested-triangles-8.0.0 | 3,000 | 15,000 | PASS — 0 mismatches | Generalize drift, size, and three-fold rotation reasoning to triangles |
 | [nested_hexagons](Dataset/nested_hexagons_dataset_3000/) | Transformational Geometry | nested-hexagons-8.0.0 | 3,000 | 15,000 | PASS — 0 mismatches | Generalize drift, size, and six-fold rotation reasoning to hexagons |
@@ -79,7 +79,7 @@ This repository generates datasets and ground truth. It does not run models, sco
 | [gauge_reading](Dataset/gauge_reading_dataset_3000/) | Physical / Mechanical Reasoning | legacy-current | 3,000 | 15,000 | PASS — 0 mismatches | Interpolate single-needle readings across varied ranges and reason about thresholds and projected values |
 | [optical_illusion](Dataset/optical_illusion_dataset_3000/) | Plane Geometry / Visual Perception | optical-illusion-3.0.0 | 3,000 | 15,000 | PASS — 0 mismatches | Separate true pixel geometry from misleading contextual size cues |
 | [compass_bearing](Dataset/compass_bearing_dataset_3000/) | Analytic Geometry / Navigation | compass-bearing-2.0.0 | 3,000 | 15,000 | PASS — 0 mismatches | Compute compass bearings, turns, and counterfactual destinations |
-| [hex_pathfinding](Dataset/hex_pathfinding_dataset_3000/) | Topology / Graph Theory | hex-pathfinding-2.0.0 | 3,000 | 15,000 | PASS — 0 mismatches | Find and count shortest paths through obstructed hex grids |
+| [hex_pathfinding](Dataset/hex_pathfinding_dataset_3000/) | Topology / Graph Theory | hex-pathfinding-3.0.0 | 3,000 | 15,000 + 3,000 open | PASS — 0 mismatches | Find and count shortest paths through obstructed hex grids |
 | [laser_mirror](Dataset/laser_mirror_dataset_3000/) | Plane Geometry / Physical Optics | legacy-current | 3,000 | 15,000 | PASS — 0 mismatches | Trace multi-bounce reflections and counterfactual mirror rotations |
 | [projectile_motion](Dataset/projectile_motion_dataset_1000/) | Physical / Mechanical Reasoning | projectile-motion-1.0.0 | 1,000 | 5,000 | PASS — 0 mismatches | Apply ideal projectile kinematics and reason about obstacle clearance and angle optimization |
 | **Total** | **34 sub-benchmarks** | — | **100,000** | **500,000** | **34/34 PASS** | **Broad visual, spatial, geometric, topological, analytic, inductive, optical, and mechanical reasoning** |
@@ -118,7 +118,7 @@ Every image has exactly five questions in increasing order of difficulty:
 
 The exact question templates differ by category, but this progression and the `difficulty_level: 1..5` annotation contract are shared by all 34 datasets. Every Level 5 operation is independently recomputable from raw stored geometry or scene metadata. Flattened CSV/JSONL files contain one row per question, yielding 15,000 rows per core dataset and 5,000 for the focused projectile-motion category.
 
-The suite-level files in `combined/` are rebuilt from dataset directories discovered by `build_manifest.json`, not from a hardcoded list. The current combined files include all 34 datasets: 33 datasets at 15,000 questions each plus `projectile_motion_dataset_1000` at 5,000 questions, for 500,000 questions and 500,000 answers. The Hub's `default` configuration loads the sharded answer Parquet view, including embedded image bytes, prompt, ground truth, and answer format. The separate `annotations` configuration contains one row per image with the combined scene metadata. The original question and answer CSVs remain available for non-viewer workflows.
+The suite-level files in `combined/` are rebuilt from dataset directories discovered by `build_manifest.json`, not from a hardcoded list. The current five-level combined files include all 34 datasets: 33 datasets at 15,000 questions each plus `projectile_motion_dataset_1000` at 5,000 questions, for 500,000 questions and 500,000 answers. Route and hex pathfinding additionally provide 3,000 supplementary open-ended questions each. Those 6,000 questions and answers are deliberately kept in the separate `all_open_questions_combined.csv` and `all_open_answers_combined.csv` files; they do not change the five-level totals or structure. The Hub's `default` configuration loads the sharded answer Parquet view, including embedded image bytes, prompt, ground truth, and answer format. The separate `annotations` configuration contains one row per image with the combined scene metadata. The original question and answer CSVs remain available for non-viewer workflows.
 
 Free-body-diagram Level 4 questions introduce the structured scoring declaration `{"type":"numeric_tolerance","tolerance_percent":2}` for real-valued mechanics answers. This field appears in annotations, answer keys, and the published answer view, but not in question-only CSVs; it records grading precision explicitly.
 
@@ -191,7 +191,7 @@ The `orthographic_dataset_3000` category implements classical orthographic/third
 
 GRIP is a published open benchmark, so the Hub's `default` configuration intentionally exposes ground truth and `answer_format` alongside each question. Its `image_bytes` column is a Hugging Face `Image` feature with the PNG bytes embedded in Parquet; `image` retains the original filename and `image_path` retains the repository-relative source path. The `annotations` configuration exposes scene metadata for inspection and analysis, with its `image` column stored as the same embedded `Image` feature. Complex list and object metadata is losslessly JSON-encoded in individual columns so the heterogeneous 34-domain schema remains representable in one table.
 
-The per-dataset `question_set.csv` files and `combined/all_questions_combined.csv` remain the question-only model-facing artifacts. Per-dataset `answer_key.csv` files, `combined/all_answers_combined.csv`, and the answer Parquet view include published reference answers. Raw `annotations.jsonl` and the annotations Parquet view can reveal the exact scene geometry and quantities from which answers are derived. **Do not provide annotations to a model under evaluation: doing so leaks answer-generating metadata and invalidates the measurement.**
+The per-dataset `question_set.csv` files, `combined/all_questions_combined.csv`, per-domain `open_questions.csv`, and `combined/all_open_questions_combined.csv` are question-only model-facing artifacts. Per-dataset `answer_key.csv`, per-domain `open_answer_key.csv`, and their combined answer files include published reference answers. Raw `annotations.jsonl`, `open_annotations.jsonl`, and the annotations Parquet view can reveal the exact scene geometry and quantities from which answers are derived. **Do not provide annotations to a model under evaluation: doing so leaks answer-generating metadata and invalidates the measurement.**
 
 ### Validation methodology
 
@@ -253,6 +253,8 @@ geomstry/
 └── combined/
     ├── all_questions_combined.csv
     ├── all_answers_combined.csv
+    ├── all_open_questions_combined.csv
+    ├── all_open_answers_combined.csv
     ├── all_answers_combined-*.parquet
     └── all_annotations_combined-*.parquet
 ```

combined/all_annotations_combined-00002-of-00006.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7f0ab80f528d212f13daafed1ff9fdaf8199c4712aefecda0fa6a277acabdb1a
+oid sha256:4c97331bc89cc149a33a913b3ae4b9a3fc4b218afedc6e3152c13392bbacf501
 size 475340110

combined/all_annotations_combined-00005-of-00006.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d1d7c11d83c31a47bd3c4aa9d83ebdb69da5232565d5e589ebb2200c43b5655f
+oid sha256:7f689b00770b112976d3e68e53fbe508c438e63b5ade64663671b7f5c1039651
 size 475831815

combined/all_answers_combined-00002-of-00006.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e293b34987f7cbc4c2edd6089d9d5df6ee244b7e3f96e30c155eeb81a34bba10
+oid sha256:c45521d87370c59e3acb8f5abaf1a0428dbf56b48e42852065dceec5296267e3
 size 475030840

combined/all_answers_combined-00005-of-00006.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:456713bb5fbecd32dd2a4c3daf54e6edb2c4b1466aacf82ad1845af743c18328
+oid sha256:2e2952d9ee1edfcbe773cb9fea5190e7c0516d774b2820a4c4d33ddac1796eb4
 size 473915751

combined/all_answers_combined.csv

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:51ccbaaf21696118e8daf22e14fe657bd3bf159208293505bb82b5ebec444da7
+oid sha256:efb6249ee39248f80c0a9de16a0a847b7b7426f53d859e68d3e5c18d6c16de85
 size 159805944

combined/all_questions_combined.csv

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8313f4c0a2f3af48f39c2ac949b651ca6b6db80d6c4845d2f65016c608ab305d
+oid sha256:af9455738f222bf117d46007f253bbe75d28126508642d43516cadaa64e2e7c3
 size 146250639