Datasets:
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
Error code: DatasetGenerationError
Exception: ValueError
Message: Expected object or value
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1816, in _prepare_split_single
for key, table in generator:
^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/json/json.py", line 248, in _generate_tables
examples = [ujson_loads(line) for line in batch.splitlines()]
^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/utils/json.py", line 20, in ujson_loads
return pd.io.json.ujson_loads(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: Expected object or value
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1348, in compute_config_parquet_and_info_response
parquet_operations = convert_to_parquet(builder)
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 980, in convert_to_parquet
builder.download_and_prepare(
File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 890, in download_and_prepare
self._download_and_prepare(
File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 951, in _download_and_prepare
self._prepare_split(split_generator, **prepare_split_kwargs)
File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1683, in _prepare_split
for job_id, done, content in self._prepare_split_single(
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1869, in _prepare_split_single
raise DatasetGenerationError("An error occurred while generating the dataset") from e
datasets.exceptions.DatasetGenerationError: An error occurred while generating the datasetNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
id string | task string | city string | difficulty string | route_id int64 | source_repo string | source_revision string | source_root string | source_paths dict | source_urls dict | instructions_text string | route_geojson dict | route_summary dict | prompt string | reference_note string |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Auckland_New_Zealand__easy__0003 | route_reversal | Auckland_New_Zealand | easy | 3 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/3/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/3/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/3/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/3/route.geojson"
} | Name: -, Instruction: Head north, Distance: 4.9 meters, Time: 3.5 seconds
Name: Saint Michael's Avenue, Instruction: Turn right onto Saint Michael's Avenue, Distance: 498.7 meters, Time: 359.1 seconds
Name: -, Instruction: Turn left, Distance: 226.6 meters, Time: 163.2 seconds
Name: Raymond Street, Instruction: Turn le... | {
"type": "FeatureCollection",
"bbox": [
174.698519,
-36.863716,
174.703274,
-36.859382
],
"features": [
{
"bbox": [
174.698519,
-36.863716,
174.703274,
-36.859382
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.698519,
-36.863716,
174.703274,
-36.859382
],
"geometry_type": "LineString",
"num_coordinates": 21,
"distance_m": 979.2,
"duration_s": 705.1,
"num_steps": 5,
"start_coordinate": [
174.698537,
-36.863716
],
"end_coordinate": [
174.699823,
-36.86031
]
... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0012 | route_reversal | Auckland_New_Zealand | easy | 12 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/12/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/12/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/12/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/12/route.geojson"
} | Name: -, Instruction: Head northwest, Distance: 79.8 meters, Time: 57.4 seconds
Name: -, Instruction: Keep left, Distance: 346.7 meters, Time: 249.6 seconds
Name: -, Instruction: Turn right, Distance: 9.0 meters, Time: 6.5 seconds
Name: Dorset Street, Instruction: Turn left onto Dorset Street, Distance: 27.6 meters, Ti... | {
"type": "FeatureCollection",
"bbox": [
174.717069,
-36.858063,
174.725332,
-36.854922
],
"features": [
{
"bbox": [
174.717069,
-36.858063,
174.725332,
-36.854922
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.717069,
-36.858063,
174.725332,
-36.854922
],
"geometry_type": "LineString",
"num_coordinates": 25,
"distance_m": 1044.4,
"duration_s": 752,
"num_steps": 9,
"start_coordinate": [
174.725332,
-36.855449
],
"end_coordinate": [
174.717308,
-36.858063
]
... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0030 | route_reversal | Auckland_New_Zealand | easy | 30 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/30/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/30/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/30/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/30/route.geojson"
} | Name: Lagoon Way, Instruction: Head east on Lagoon Way, Distance: 59.2 meters, Time: 42.6 seconds
Name: Lagoon Way, Instruction: Turn right onto Lagoon Way, Distance: 112.7 meters, Time: 81.2 seconds
Name: Excelsior Way, Instruction: Enter the roundabout and take the 1st exit onto Excelsior Way, Distance: 77.0 meters, ... | {
"type": "FeatureCollection",
"bbox": [
174.773817,
-36.334105,
174.7779,
-36.326572
],
"features": [
{
"bbox": [
174.773817,
-36.334105,
174.7779,
-36.326572
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.773817,
-36.334105,
174.7779,
-36.326572
],
"geometry_type": "LineString",
"num_coordinates": 29,
"distance_m": 1141.4,
"duration_s": 821.8,
"num_steps": 9,
"start_coordinate": [
174.777225,
-36.326572
],
"end_coordinate": [
174.774051,
-36.334105
]
... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0040 | route_reversal | Auckland_New_Zealand | easy | 40 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/40/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/40/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/40/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/40/route.geojson"
} | Name: -, Instruction: Head east, Distance: 7.2 meters, Time: 5.2 seconds
Name: -, Instruction: Turn left, Distance: 5.2 meters, Time: 3.8 seconds
Name: -, Instruction: Turn right, Distance: 21.5 meters, Time: 15.5 seconds
Name: -, Instruction: Turn left, Distance: 6.7 meters, Time: 4.8 seconds
Name: Roys Road, Instruct... | {
"type": "FeatureCollection",
"bbox": [
174.859545,
-37.05092,
174.861309,
-37.044535
],
"features": [
{
"bbox": [
174.859545,
-37.05092,
174.861309,
-37.044535
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.859545,
-37.05092,
174.861309,
-37.044535
],
"geometry_type": "LineString",
"num_coordinates": 13,
"distance_m": 849.5,
"duration_s": 611.7,
"num_steps": 11,
"start_coordinate": [
174.859648,
-37.05091
],
"end_coordinate": [
174.861309,
-37.044535
]
... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0048 | route_reversal | Auckland_New_Zealand | easy | 48 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/48/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/48/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/48/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/48/route.geojson"
} | Name: -, Instruction: Head northeast, Distance: 136.1 meters, Time: 98.0 seconds
Name: -, Instruction: Turn left, Distance: 107.6 meters, Time: 77.5 seconds
Name: -, Instruction: Turn left, Distance: 181.8 meters, Time: 130.9 seconds
Name: -, Instruction: Turn right, Distance: 2.0 meters, Time: 1.5 seconds
Name: -, Ins... | {
"type": "FeatureCollection",
"bbox": [
174.796706,
-36.829445,
174.802989,
-36.827972
],
"features": [
{
"bbox": [
174.796706,
-36.829445,
174.802989,
-36.827972
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.796706,
-36.829445,
174.802989,
-36.827972
],
"geometry_type": "LineString",
"num_coordinates": 33,
"distance_m": 817.4,
"duration_s": 588.5,
"num_steps": 9,
"start_coordinate": [
174.801672,
-36.829445
],
"end_coordinate": [
174.796706,
-36.827998
]... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0054 | route_reversal | Auckland_New_Zealand | easy | 54 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/54/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/54/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/54/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/54/route.geojson"
} | Name: -, Instruction: Head west, Distance: 5.9 meters, Time: 4.3 seconds
Name: -, Instruction: Turn left, Distance: 58.2 meters, Time: 41.9 seconds
Name: -, Instruction: Turn right, Distance: 77.5 meters, Time: 55.8 seconds
Name: -, Instruction: Keep left, Distance: 163.3 meters, Time: 117.6 seconds
Name: Flounder Road... | {
"type": "FeatureCollection",
"bbox": [
174.663609,
-36.798498,
174.667756,
-36.794571
],
"features": [
{
"bbox": [
174.663609,
-36.798498,
174.667756,
-36.794571
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.663609,
-36.798498,
174.667756,
-36.794571
],
"geometry_type": "LineString",
"num_coordinates": 46,
"distance_m": 1039.5,
"duration_s": 748.4,
"num_steps": 13,
"start_coordinate": [
174.667479,
-36.794961
],
"end_coordinate": [
174.667736,
-36.798498
... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0093 | route_reversal | Auckland_New_Zealand | easy | 93 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/93/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/93/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/93/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/93/route.geojson"
} | Name: Cascades Road, Instruction: Head northwest on Cascades Road, Distance: 8.5 meters, Time: 6.1 seconds
Name: Cascades Road, Instruction: Keep left onto Cascades Road, Distance: 77.0 meters, Time: 55.5 seconds
Name: Cascades Road, Instruction: Keep left onto Cascades Road, Distance: 33.5 meters, Time: 24.1 seconds
N... | {
"type": "FeatureCollection",
"bbox": [
174.885237,
-36.906713,
174.895252,
-36.903547
],
"features": [
{
"bbox": [
174.885237,
-36.906713,
174.895252,
-36.903547
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.885237,
-36.906713,
174.895252,
-36.903547
],
"geometry_type": "LineString",
"num_coordinates": 38,
"distance_m": 1326.1,
"duration_s": 954.8,
"num_steps": 12,
"start_coordinate": [
174.895252,
-36.904691
],
"end_coordinate": [
174.885237,
-36.906713
... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0094 | route_reversal | Auckland_New_Zealand | easy | 94 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/94/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/94/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/94/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/94/route.geojson"
} | Name: Cascades Road, Instruction: Head northwest on Cascades Road, Distance: 8.5 meters, Time: 6.1 seconds
Name: Cascades Road, Instruction: Keep right onto Cascades Road, Distance: 83.8 meters, Time: 60.4 seconds
Name: -, Instruction: Keep right, Distance: 24.8 meters, Time: 17.9 seconds
Name: Pakuranga Road, 5, Instr... | {
"type": "FeatureCollection",
"bbox": [
174.894704,
-36.904691,
174.90312,
-36.898901
],
"features": [
{
"bbox": [
174.894704,
-36.904691,
174.90312,
-36.898901
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.894704,
-36.904691,
174.90312,
-36.898901
],
"geometry_type": "LineString",
"num_coordinates": 60,
"distance_m": 1350.7,
"duration_s": 972.5,
"num_steps": 21,
"start_coordinate": [
174.895252,
-36.904691
],
"end_coordinate": [
174.90312,
-36.898958
]... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0103 | route_reversal | Auckland_New_Zealand | easy | 103 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/103/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/103/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/103/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/103/route.geojson"
} | Name: -, Instruction: Head west, Distance: 99.3 meters, Time: 71.5 seconds
Name: Grey Street, Instruction: Turn right onto Grey Street, Distance: 610.0 meters, Time: 439.2 seconds
Name: Wade Avenue, Instruction: Turn sharp right onto Wade Avenue, Distance: 168.2 meters, Time: 121.1 seconds
Name: -, Instruction: Turn sh... | {
"type": "FeatureCollection",
"bbox": [
174.787779,
-36.918263,
174.795132,
-36.916809
],
"features": [
{
"bbox": [
174.787779,
-36.918263,
174.795132,
-36.916809
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.787779,
-36.918263,
174.795132,
-36.916809
],
"geometry_type": "LineString",
"num_coordinates": 72,
"distance_m": 940,
"duration_s": 676.8,
"num_steps": 5,
"start_coordinate": [
174.795132,
-36.917093
],
"end_coordinate": [
174.787779,
-36.916839
]
} | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0104 | route_reversal | Auckland_New_Zealand | easy | 104 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/104/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/104/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/104/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/104/route.geojson"
} | Name: -, Instruction: Head west, Distance: 99.3 meters, Time: 71.5 seconds
Name: Grey Street, Instruction: Turn right onto Grey Street, Distance: 717.4 meters, Time: 516.5 seconds
Name: Cameron Street, Instruction: Turn right onto Cameron Street, Distance: 49.1 meters, Time: 35.3 seconds
Name: -, Instruction: Turn left... | {
"type": "FeatureCollection",
"bbox": [
174.786498,
-36.918371,
174.795132,
-36.917092
],
"features": [
{
"bbox": [
174.786498,
-36.918371,
174.795132,
-36.917092
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.786498,
-36.918371,
174.795132,
-36.917092
],
"geometry_type": "LineString",
"num_coordinates": 71,
"distance_m": 890,
"duration_s": 640.8,
"num_steps": 6,
"start_coordinate": [
174.795132,
-36.917093
],
"end_coordinate": [
174.786498,
-36.917873
]
} | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0110 | route_reversal | Auckland_New_Zealand | easy | 110 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/110/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/110/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/110/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/110/route.geojson"
} | Name: Kingdale Road, Instruction: Head southeast on Kingdale Road, Distance: 13.8 meters, Time: 10.0 seconds
Name: -, Instruction: Turn left, Distance: 4.2 meters, Time: 3.0 seconds
Name: -, Instruction: Turn slight right, Distance: 351.8 meters, Time: 253.3 seconds
Name: -, Instruction: Turn slight left, Distance: 178... | {
"type": "FeatureCollection",
"bbox": [
174.62525,
-36.862016,
174.629445,
-36.860066
],
"features": [
{
"bbox": [
174.62525,
-36.862016,
174.629445,
-36.860066
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.62525,
-36.862016,
174.629445,
-36.860066
],
"geometry_type": "LineString",
"num_coordinates": 31,
"distance_m": 757.3,
"duration_s": 545.2,
"num_steps": 8,
"start_coordinate": [
174.62525,
-36.861553
],
"end_coordinate": [
174.62721,
-36.860066
]
} | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0113 | route_reversal | Auckland_New_Zealand | easy | 113 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/113/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/113/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/113/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/113/route.geojson"
} | Name: Liberty Crescent, Instruction: Head east on Liberty Crescent, Distance: 8.2 meters, Time: 5.9 seconds
Name: -, Instruction: Turn left, Distance: 14.8 meters, Time: 10.6 seconds
Name: -, Instruction: Turn left, Distance: 108.5 meters, Time: 78.2 seconds
Name: -, Instruction: Turn left, Distance: 17.4 meters, Time:... | {
"type": "FeatureCollection",
"bbox": [
175.010222,
-36.888255,
175.01297,
-36.885033
],
"features": [
{
"bbox": [
175.010222,
-36.888255,
175.01297,
-36.885033
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
175.010222,
-36.888255,
175.01297,
-36.885033
],
"geometry_type": "LineString",
"num_coordinates": 31,
"distance_m": 561.4,
"duration_s": 404.2,
"num_steps": 7,
"start_coordinate": [
175.010377,
-36.88824
],
"end_coordinate": [
175.01297,
-36.885063
]
} | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0117 | route_reversal | Auckland_New_Zealand | easy | 117 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/117/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/117/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/117/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/117/route.geojson"
} | Name: -, Instruction: Head northeast, Distance: 75.8 meters, Time: 54.5 seconds
Name: -, Instruction: Keep right, Distance: 29.0 meters, Time: 20.9 seconds
Name: -, Instruction: Turn right, Distance: 55.9 meters, Time: 40.2 seconds
Name: Bearing Parade, Instruction: Turn left onto Bearing Parade, Distance: 302.3 meters... | {
"type": "FeatureCollection",
"bbox": [
174.744944,
-36.691294,
174.749963,
-36.687994
],
"features": [
{
"bbox": [
174.744944,
-36.691294,
174.749963,
-36.687994
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.744944,
-36.691294,
174.749963,
-36.687994
],
"geometry_type": "LineString",
"num_coordinates": 42,
"distance_m": 877.7,
"duration_s": 632,
"num_steps": 9,
"start_coordinate": [
174.744944,
-36.68856
],
"end_coordinate": [
174.749294,
-36.691294
]
} | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0119 | route_reversal | Auckland_New_Zealand | easy | 119 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/119/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/119/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/119/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/119/route.geojson"
} | Name: -, Instruction: Head east, Distance: 25.1 meters, Time: 18.1 seconds
Name: William Pickering Drive, Instruction: Turn right onto William Pickering Drive, Distance: 133.4 meters, Time: 96.0 seconds
Name: -, Instruction: Turn left, Distance: 20.0 meters, Time: 14.4 seconds
Name: -, Instruction: Turn right, Distance... | {
"type": "FeatureCollection",
"bbox": [
174.69956,
-36.753774,
174.706983,
-36.748382
],
"features": [
{
"bbox": [
174.69956,
-36.753774,
174.706983,
-36.748382
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.69956,
-36.753774,
174.706983,
-36.748382
],
"geometry_type": "LineString",
"num_coordinates": 62,
"distance_m": 1156.9,
"duration_s": 832.9,
"num_steps": 17,
"start_coordinate": [
174.69956,
-36.748438
],
"end_coordinate": [
174.706983,
-36.753774
]... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0136 | route_reversal | Auckland_New_Zealand | easy | 136 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/136/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/136/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/136/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/136/route.geojson"
} | Name: -, Instruction: Head east, Distance: 67.9 meters, Time: 48.9 seconds
Name: -, Instruction: Turn right, Distance: 6.8 meters, Time: 4.9 seconds
Name: Kopuru Road, Instruction: Turn left onto Kopuru Road, Distance: 21.5 meters, Time: 15.5 seconds
Name: -, Instruction: Turn right, Distance: 6.9 meters, Time: 5.0 sec... | {
"type": "FeatureCollection",
"bbox": [
174.608526,
-36.791354,
174.610131,
-36.788101
],
"features": [
{
"bbox": [
174.608526,
-36.791354,
174.610131,
-36.788101
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.608526,
-36.791354,
174.610131,
-36.788101
],
"geometry_type": "LineString",
"num_coordinates": 50,
"distance_m": 609.3,
"duration_s": 438.7,
"num_steps": 15,
"start_coordinate": [
174.608526,
-36.788101
],
"end_coordinate": [
174.608849,
-36.790903
... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0157 | route_reversal | Auckland_New_Zealand | easy | 157 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/157/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/157/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/157/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/157/route.geojson"
} | Name: Boyle Crescent, Instruction: Head east on Boyle Crescent, Distance: 89.1 meters, Time: 64.2 seconds
Name: -, Instruction: Turn right, Distance: 174.4 meters, Time: 125.6 seconds
Name: -, Instruction: Turn left, Distance: 288.5 meters, Time: 207.7 seconds
Name: -, Instruction: Arrive at your destination, straight ... | {
"type": "FeatureCollection",
"bbox": [
174.769603,
-36.86481,
174.774123,
-36.862583
],
"features": [
{
"bbox": [
174.769603,
-36.86481,
174.774123,
-36.862583
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.769603,
-36.86481,
174.774123,
-36.862583
],
"geometry_type": "LineString",
"num_coordinates": 19,
"distance_m": 552.1,
"duration_s": 397.5,
"num_steps": 4,
"start_coordinate": [
174.769603,
-36.862583
],
"end_coordinate": [
174.774123,
-36.86481
]
} | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0163 | route_reversal | Auckland_New_Zealand | easy | 163 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/163/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/163/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/163/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/163/route.geojson"
} | Name: -, Instruction: Head north, Distance: 7.6 meters, Time: 5.5 seconds
Name: -, Instruction: Turn left, Distance: 22.0 meters, Time: 15.8 seconds
Name: -, Instruction: Turn left, Distance: 15.4 meters, Time: 11.1 seconds
Name: -, Instruction: Turn right, Distance: 86.8 meters, Time: 62.5 seconds
Name: Glenmall Place... | {
"type": "FeatureCollection",
"bbox": [
174.648363,
-36.912079,
174.652957,
-36.909188
],
"features": [
{
"bbox": [
174.648363,
-36.912079,
174.652957,
-36.909188
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.648363,
-36.912079,
174.652957,
-36.909188
],
"geometry_type": "LineString",
"num_coordinates": 37,
"distance_m": 636.7,
"duration_s": 458.4,
"num_steps": 11,
"start_coordinate": [
174.652945,
-36.912079
],
"end_coordinate": [
174.64838,
-36.909188
]... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0173 | route_reversal | Auckland_New_Zealand | easy | 173 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/173/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/173/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/173/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/173/route.geojson"
} | Name: -, Instruction: Head northeast, Distance: 144.3 meters, Time: 103.9 seconds
Name: -, Instruction: Turn left, Distance: 27.3 meters, Time: 19.6 seconds
Name: -, Instruction: Turn left, Distance: 417.8 meters, Time: 300.8 seconds
Name: Karapapa Road, Instruction: Turn right onto Karapapa Road, Distance: 204.5 meter... | {
"type": "FeatureCollection",
"bbox": [
174.645718,
-36.614009,
174.651301,
-36.610153
],
"features": [
{
"bbox": [
174.645718,
-36.614009,
174.651301,
-36.610153
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.645718,
-36.614009,
174.651301,
-36.610153
],
"geometry_type": "LineString",
"num_coordinates": 55,
"distance_m": 793.9,
"duration_s": 571.6,
"num_steps": 5,
"start_coordinate": [
174.645718,
-36.614009
],
"end_coordinate": [
174.651301,
-36.611457
]... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0182 | route_reversal | Auckland_New_Zealand | easy | 182 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/182/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/182/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/182/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/182/route.geojson"
} | Name: Haumako Place, Instruction: Head south on Haumako Place, Distance: 39.4 meters, Time: 28.4 seconds
Name: Haumako Place, Instruction: Turn left onto Haumako Place, Distance: 91.0 meters, Time: 65.5 seconds
Name: Albert Street, Instruction: Turn right onto Albert Street, Distance: 104.7 meters, Time: 75.4 seconds
N... | {
"type": "FeatureCollection",
"bbox": [
174.849909,
-36.939494,
174.857276,
-36.937765
],
"features": [
{
"bbox": [
174.849909,
-36.939494,
174.857276,
-36.937765
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.849909,
-36.939494,
174.857276,
-36.937765
],
"geometry_type": "LineString",
"num_coordinates": 39,
"distance_m": 884.4,
"duration_s": 636.8,
"num_steps": 10,
"start_coordinate": [
174.849909,
-36.937975
],
"end_coordinate": [
174.857276,
-36.937765
... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0187 | route_reversal | Auckland_New_Zealand | easy | 187 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/187/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/187/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/187/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/187/route.geojson"
} | Name: Wainoni Heights, Instruction: Head northwest on Wainoni Heights, Distance: 34.5 meters, Time: 24.8 seconds
Name: -, Instruction: Turn left, Distance: 4.7 meters, Time: 3.4 seconds
Name: -, Instruction: Turn right, Distance: 99.7 meters, Time: 71.8 seconds
Name: -, Instruction: Turn left, Distance: 246.9 meters, T... | {
"type": "FeatureCollection",
"bbox": [
174.674734,
-36.77365,
174.678957,
-36.771342
],
"features": [
{
"bbox": [
174.674734,
-36.77365,
174.678957,
-36.771342
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.674734,
-36.77365,
174.678957,
-36.771342
],
"geometry_type": "LineString",
"num_coordinates": 25,
"distance_m": 606.9,
"duration_s": 436.9,
"num_steps": 9,
"start_coordinate": [
174.678957,
-36.77365
],
"end_coordinate": [
174.674734,
-36.771342
]
} | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0197 | route_reversal | Auckland_New_Zealand | easy | 197 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/197/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/197/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/197/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/197/route.geojson"
} | Name: -, Instruction: Head southeast, Distance: 170.1 meters, Time: 122.5 seconds
Name: Parsons Road, Instruction: Turn sharp left onto Parsons Road, Distance: 95.1 meters, Time: 68.5 seconds
Name: Gowing Drive, Instruction: Turn right onto Gowing Drive, Distance: 131.5 meters, Time: 94.7 seconds
Name: Rutherford Terra... | {
"type": "FeatureCollection",
"bbox": [
174.831233,
-36.87312,
174.836634,
-36.869159
],
"features": [
{
"bbox": [
174.831233,
-36.87312,
174.836634,
-36.869159
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.831233,
-36.87312,
174.836634,
-36.869159
],
"geometry_type": "LineString",
"num_coordinates": 31,
"distance_m": 924.4,
"duration_s": 665.6,
"num_steps": 8,
"start_coordinate": [
174.831294,
-36.869159
],
"end_coordinate": [
174.835804,
-36.873099
]
... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0201 | route_reversal | Auckland_New_Zealand | easy | 201 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/201/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/201/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/201/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/201/route.geojson"
} | Name: -, Instruction: Head west, Distance: 3.7 meters, Time: 2.6 seconds
Name: -, Instruction: Turn left, Distance: 164.4 meters, Time: 118.3 seconds
Name: -, Instruction: Turn right, Distance: 4.4 meters, Time: 3.2 seconds
Name: Waihorotiu Path, Instruction: Keep left onto Waihorotiu Path, Distance: 319.9 meters, Time... | {
"type": "FeatureCollection",
"bbox": [
174.764618,
-36.852138,
174.766713,
-36.844819
],
"features": [
{
"bbox": [
174.764618,
-36.852138,
174.766713,
-36.844819
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.764618,
-36.852138,
174.766713,
-36.844819
],
"geometry_type": "LineString",
"num_coordinates": 44,
"distance_m": 901.3,
"duration_s": 648.9,
"num_steps": 13,
"start_coordinate": [
174.766713,
-36.844819
],
"end_coordinate": [
174.764957,
-36.852138
... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0207 | route_reversal | Auckland_New_Zealand | easy | 207 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/207/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/207/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/207/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/207/route.geojson"
} | Name: -, Instruction: Head west, Distance: 70.3 meters, Time: 50.6 seconds
Name: -, Instruction: Turn right, Distance: 44.5 meters, Time: 32.0 seconds
Name: -, Instruction: Turn left, Distance: 75.5 meters, Time: 54.4 seconds
Name: -, Instruction: Turn right, Distance: 90.3 meters, Time: 65.0 seconds
Name: Cleek Road, ... | {
"type": "FeatureCollection",
"bbox": [
174.823374,
-36.95905,
174.829603,
-36.957572
],
"features": [
{
"bbox": [
174.823374,
-36.95905,
174.829603,
-36.957572
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.823374,
-36.95905,
174.829603,
-36.957572
],
"geometry_type": "LineString",
"num_coordinates": 35,
"distance_m": 762.6,
"duration_s": 549,
"num_steps": 10,
"start_coordinate": [
174.829603,
-36.957665
],
"end_coordinate": [
174.823374,
-36.95807
]
} | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0215 | route_reversal | Auckland_New_Zealand | easy | 215 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/215/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/215/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/215/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/215/route.geojson"
} | Name: -, Instruction: Head northwest, Distance: 10.0 meters, Time: 7.2 seconds
Name: -, Instruction: Turn slight left, Distance: 21.5 meters, Time: 15.5 seconds
Name: -, Instruction: Turn sharp left, Distance: 17.9 meters, Time: 12.9 seconds
Name: -, Instruction: Turn right, Distance: 7.0 meters, Time: 5.0 seconds
Name... | {
"type": "FeatureCollection",
"bbox": [
174.762016,
-36.853655,
174.765643,
-36.851035
],
"features": [
{
"bbox": [
174.762016,
-36.853655,
174.765643,
-36.851035
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.762016,
-36.853655,
174.765643,
-36.851035
],
"geometry_type": "LineString",
"num_coordinates": 36,
"distance_m": 491.5,
"duration_s": 353.8,
"num_steps": 13,
"start_coordinate": [
174.765643,
-36.851189
],
"end_coordinate": [
174.762016,
-36.853655
... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0217 | route_reversal | Auckland_New_Zealand | easy | 217 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/217/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/217/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/217/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/217/route.geojson"
} | Name: -, Instruction: Head west, Distance: 37.1 meters, Time: 26.7 seconds
Name: -, Instruction: Turn right, Distance: 142.7 meters, Time: 102.8 seconds
Name: -, Instruction: Turn slight right, Distance: 94.5 meters, Time: 68.1 seconds
Name: -, Instruction: Turn left, Distance: 13.8 meters, Time: 9.9 seconds
Name: -, I... | {
"type": "FeatureCollection",
"bbox": [
174.649343,
-36.854437,
174.651157,
-36.848783
],
"features": [
{
"bbox": [
174.649343,
-36.854437,
174.651157,
-36.848783
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.649343,
-36.854437,
174.651157,
-36.848783
],
"geometry_type": "LineString",
"num_coordinates": 45,
"distance_m": 801.7,
"duration_s": 577.2,
"num_steps": 11,
"start_coordinate": [
174.650156,
-36.854343
],
"end_coordinate": [
174.651101,
-36.848783
... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0236 | route_reversal | Auckland_New_Zealand | easy | 236 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/236/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/236/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/236/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/236/route.geojson"
} | Name: Nautilus Drive, Instruction: Head north on Nautilus Drive, Distance: 68.4 meters, Time: 49.3 seconds
Name: -, Instruction: Turn left, Distance: 6.4 meters, Time: 4.6 seconds
Name: -, Instruction: Turn right, Distance: 20.3 meters, Time: 14.6 seconds
Name: -, Instruction: Turn left, Distance: 43.8 meters, Time: 31... | {
"type": "FeatureCollection",
"bbox": [
174.787856,
-36.616802,
174.793362,
-36.614069
],
"features": [
{
"bbox": [
174.787856,
-36.616802,
174.793362,
-36.614069
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.787856,
-36.616802,
174.793362,
-36.614069
],
"geometry_type": "LineString",
"num_coordinates": 50,
"distance_m": 878.7,
"duration_s": 632.6,
"num_steps": 10,
"start_coordinate": [
174.792527,
-36.616802
],
"end_coordinate": [
174.787856,
-36.615488
... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0256 | route_reversal | Auckland_New_Zealand | easy | 256 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/256/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/256/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/256/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/256/route.geojson"
} | Name: Calypso Way, Instruction: Head west on Calypso Way, Distance: 64.4 meters, Time: 46.4 seconds
Name: Calypso Way, Instruction: Enter the roundabout and take the 2nd exit onto Calypso Way, Distance: 108.2 meters, Time: 77.9 seconds
Name: Calypso Way, Instruction: Continue straight onto Calypso Way, Distance: 15.3 m... | {
"type": "FeatureCollection",
"bbox": [
174.711816,
-36.757249,
174.719305,
-36.754731
],
"features": [
{
"bbox": [
174.711816,
-36.757249,
174.719305,
-36.754731
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.711816,
-36.757249,
174.719305,
-36.754731
],
"geometry_type": "LineString",
"num_coordinates": 64,
"distance_m": 1055.7,
"duration_s": 760.1,
"num_steps": 17,
"start_coordinate": [
174.719305,
-36.757113
],
"end_coordinate": [
174.711816,
-36.756905
... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0258 | route_reversal | Auckland_New_Zealand | easy | 258 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/258/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/258/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/258/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/258/route.geojson"
} | Name: -, Instruction: Head northwest, Distance: 27.4 meters, Time: 19.8 seconds
Name: -, Instruction: Turn right, Distance: 27.9 meters, Time: 20.1 seconds
Name: -, Instruction: Turn slight right, Distance: 2.1 meters, Time: 1.5 seconds
Name: -, Instruction: Turn left, Distance: 19.8 meters, Time: 14.2 seconds
Name: -,... | {
"type": "FeatureCollection",
"bbox": [
174.59434,
-36.816618,
174.598614,
-36.813184
],
"features": [
{
"bbox": [
174.59434,
-36.816618,
174.598614,
-36.813184
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.59434,
-36.816618,
174.598614,
-36.813184
],
"geometry_type": "LineString",
"num_coordinates": 46,
"distance_m": 907.9,
"duration_s": 653.7,
"num_steps": 17,
"start_coordinate": [
174.59442,
-36.814938
],
"end_coordinate": [
174.597585,
-36.816618
]
... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0264 | route_reversal | Auckland_New_Zealand | easy | 264 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/264/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/264/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/264/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/264/route.geojson"
} | Name: -, Instruction: Head southeast, Distance: 6.0 meters, Time: 4.4 seconds
Name: -, Instruction: Turn left, Distance: 261.2 meters, Time: 188.1 seconds
Name: -, Instruction: Turn right, Distance: 13.0 meters, Time: 9.3 seconds
Name: -, Instruction: Turn left, Distance: 207.1 meters, Time: 149.1 seconds
Name: -, Inst... | {
"type": "FeatureCollection",
"bbox": [
174.687771,
-36.791361,
174.693662,
-36.790014
],
"features": [
{
"bbox": [
174.687771,
-36.791361,
174.693662,
-36.790014
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.687771,
-36.791361,
174.693662,
-36.790014
],
"geometry_type": "LineString",
"num_coordinates": 26,
"distance_m": 653.1,
"duration_s": 470.2,
"num_steps": 9,
"start_coordinate": [
174.687771,
-36.791329
],
"end_coordinate": [
174.693662,
-36.791272
]... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
Auckland_New_Zealand__easy__0268 | route_reversal | Auckland_New_Zealand | easy | 268 | bghjmn32/EMNLP2025_Turnback | main | 36kroutes | {
"instructions": "36kroutes/Auckland_New_Zealand/easy/268/instructions.txt",
"route_geojson": "36kroutes/Auckland_New_Zealand/easy/268/route.geojson"
} | {
"instructions": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/268/instructions.txt",
"route_geojson": "https://raw.githubusercontent.com/bghjmn32/EMNLP2025_Turnback/main/36kroutes/Auckland_New_Zealand/easy/268/route.geojson"
} | Name: -, Instruction: Head east, Distance: 8.8 meters, Time: 6.3 seconds
Name: Musick Point Road, Instruction: Turn right onto Musick Point Road, Distance: 439.7 meters, Time: 316.6 seconds
Name: -, Instruction: Turn left, Distance: 26.1 meters, Time: 18.8 seconds
Name: -, Instruction: Turn left, Distance: 38.8 meters,... | {
"type": "FeatureCollection",
"bbox": [
174.900185,
-36.855481,
174.901603,
-36.851564
],
"features": [
{
"bbox": [
174.900185,
-36.855481,
174.901603,
-36.851564
],
"type": "Feature",
"properties": {
"segments": [
{
... | {
"bbox": [
174.900185,
-36.855481,
174.901603,
-36.851564
],
"geometry_type": "LineString",
"num_coordinates": 14,
"distance_m": 513.4,
"duration_s": 369.7,
"num_steps": 5,
"start_coordinate": [
174.900185,
-36.851564
],
"end_coordinate": [
174.901603,
-36.85529
]
... | You are evaluating route-level geospatial cognition.
Given the following forward pedestrian navigation instructions, generate reverse navigation instructions that guide a pedestrian from the destination back to the original starting point. Preserve street names when useful and write executable step-by-step navigation ... | For TurnBack route reversal, the optimal reverse route is the original route traversed backwards. Geometry-based evaluation should execute a model's reverse instructions with Path Builder and compare the recovered route against this route_geojson reference. |
🧭 TurnBack 10% Stratified Sample
A deterministic route-reversal subset for lightweight geospatial cognition experiments
TurnBack 10% is a balanced, reproducible subset of the public TurnBack route corpus for route-level reverse-navigation generation and evaluation.
Quick Start · At a Glance · Files · Sampling · Citation
This is a 10% stratified sample of the public raw route corpus from
bghjmn32/EMNLP2025_Turnback, not the full TurnBack benchmark. It is intended for lightweight route-reversal experiments, prompt iteration, and reproducibility checks before full-scale evaluation.
✨ Why This Dataset?
TurnBack evaluates route-level geospatial cognition: a model receives forward pedestrian navigation instructions and must generate reverse instructions that guide the pedestrian back to the original start. The reference geometry is the original route traversed backwards, and downstream evaluation should execute the generated reverse instructions with Path Builder before scoring recovered geometry.
This subset is designed for questions like:
- Can a model reverse pedestrian navigation instructions across cities and difficulty levels?
- How does route-reversal quality change from easy to hard routes?
- Can a geospatial evaluator run quickly on a balanced sample before full TurnBack?
- Can every sampled route be traced back to official raw instructions and route geometry?
📦 Dataset at a Glance
| 3,900 sampled routes |
13 public city directories |
3 difficulty levels |
100 routes per stratum |
| 1,300 easy routes |
1,300 medium routes |
1,300 hard routes |
20260606 sampling seed |
Covered cities
Auckland_New_Zealand · Cairo_Egypt · Cape_Town_South_Africa · Denver_Colorado_USA · London_UK · Mexico_City_Mexico · Munich_Germany · Paris_France · Rio_de_Janeiro_Brazil · Singapore_Singapore · Sydney_Australia · Tokyo_23_wards · Toronto_Canada
Evaluation shape
| Component | Description |
|---|---|
| Input | Forward pedestrian navigation instructions from the official raw corpus. |
| Output | Model-generated reverse instructions back to the original start. |
| Reference | Original route geometry traversed backwards. |
| Recommended scorer | Execute generated instructions with Path Builder, then compare recovered route geometry. |
| Interaction pattern | One prompt / response per sampled route. |
🚀 Quick Start
Load the JSONL dataset:
from datasets import load_dataset
ds = load_dataset(
"json",
data_files="https://huggingface.co/datasets/zhangdw/TurnBack-10pct/resolve/main/data/turnback_10pct.jsonl",
split="train",
)
print(len(ds))
print(ds[0].keys())
Inspect one route-reversal prompt:
row = ds[0]
print(row["city"], row["difficulty"], row["route_id"])
print(row["instructions_text"][:1000])
print(row["prompt"][:1000])
Download the full subset, metadata, and regeneration script:
hf download zhangdw/TurnBack-10pct \
--type dataset \
--local-dir data/turnback-10pct
🗂️ Files
| Path | Description |
|---|---|
data/turnback_10pct.jsonl |
One JSON object per sampled route. |
metadata/counts.json |
Sampling settings, city/difficulty counts, and validation summary. |
metadata/sample_ids.json |
Sampled route IDs for every city × difficulty cell. |
scripts/build_turnback_10pct_dataset.py |
Exact script used to regenerate this dataset from the official source. |
Each JSONL record contains:
| Field | Description |
|---|---|
id, city, difficulty, route_id |
Stable route identifiers and strata. |
instructions_text |
Forward navigation instructions from the official raw corpus. |
route_geojson |
Reference route geometry from the official raw corpus. |
route_summary |
Distance, duration, step count, coordinate count, bbox, and start/end coordinates. |
prompt |
A simple route-reversal prompt built from the forward instructions. |
source_paths, source_urls |
Provenance back to the official GitHub raw files. |
🧭 Example Workflows
Count routes by city and difficulty
import pandas as pd
df = ds.to_pandas()
print(df.groupby(["city", "difficulty"]).size())
print(df.groupby("difficulty").size())
Read the sampling manifest
import json
from huggingface_hub import hf_hub_download
path = hf_hub_download(
repo_id="zhangdw/TurnBack-10pct",
repo_type="dataset",
filename="metadata/counts.json",
)
with open(path, "r", encoding="utf-8") as f:
counts = json.load(f)
print(counts["total_sampled_routes"])
print(counts["counts_by_difficulty"])
Build a simple model input
messages = [
{
"role": "system",
"content": "You reverse pedestrian navigation instructions while preserving route geometry.",
},
{"role": "user", "content": row["prompt"]},
]
🧪 Sampling Policy
| Setting | Value |
|---|---|
| Source repository | bghjmn32/EMNLP2025_Turnback |
| Source revision | main |
| Source root | 36kroutes/ |
| Sampling seed | 20260606 |
| Sampling fraction | 0.1 |
| Stratification | city × difficulty |
| Sample per cell | 100 routes |
| Route ID universe per cell | 0..999 |
| Total sampled routes | 3,900 |
| Created at | 2026-06-06T18:31:38.193396+00:00 |
The original paper describes a 36,000-route benchmark over 12 metropolitan areas. The currently published raw 36kroutes/ snapshot in the official GitHub repository contains an additional public city directory, Paris_France. The official public code can enumerate, load, and score that city, so this subset includes all 13 public city directories. To preserve a paper-style balanced layout, it samples from the canonical route-id universe 0..999 in each city × difficulty cell and excludes overflow route directories that appear in some raw-snapshot cells.
📊 Counts by Stratum
Every city contributes 300 routes: 100 easy, 100 medium, and 100 hard.
Full city × difficulty table
| City | Easy | Medium | Hard | Total |
|---|---|---|---|---|
| Auckland_New_Zealand | 100 | 100 | 100 | 300 |
| Cairo_Egypt | 100 | 100 | 100 | 300 |
| Cape_Town_South_Africa | 100 | 100 | 100 | 300 |
| Denver_Colorado_USA | 100 | 100 | 100 | 300 |
| London_UK | 100 | 100 | 100 | 300 |
| Mexico_City_Mexico | 100 | 100 | 100 | 300 |
| Munich_Germany | 100 | 100 | 100 | 300 |
| Paris_France | 100 | 100 | 100 | 300 |
| Rio_de_Janeiro_Brazil | 100 | 100 | 100 | 300 |
| Singapore_Singapore | 100 | 100 | 100 | 300 |
| Sydney_Australia | 100 | 100 | 100 | 300 |
| Tokyo_23_wards | 100 | 100 | 100 | 300 |
| Toronto_Canada | 100 | 100 | 100 | 300 |
| Total | 1,300 | 1,300 | 1,300 | 3,900 |
✅ Intended Use
This subset is useful for:
- quick route-reversal prompt and evaluator iteration;
- balanced city × difficulty smoke tests;
- geospatial instruction-following experiments;
- provenance-preserving route-level error analysis;
- validating code before running the full TurnBack corpus.
It is not a replacement for full TurnBack benchmark reporting.
⚖️ License
This subset follows the MIT license metadata of the upstream public TurnBack release. Check the official TurnBack repository and paper for source-specific requirements.
📚 Citation
If you use this dataset, cite the original TurnBack paper and official code/data release:
@inproceedings{luo2025turnback,
title = {TurnBack: A Geospatial Route Cognition Benchmark for Large Language Models through Reverse Route},
author = {Luo, Hongyi and Cheng, Qing and Matos, Daniel and Gadi, Hari Krishna and Zhang, Yanfeng and Liu, Lu and Wang, Yongliang and Zeller, Niclas and Cremers, Daniel and Meng, Liqiu},
booktitle = {Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing},
year = {2025}
}
You may also cite this deterministic subset when the 10% sampling policy matters for reproducibility:
@misc{turnback_10pct2026,
title = {TurnBack 10% Stratified Sample},
author = {Dawei Zhang},
year = {2026},
howpublished = {Hugging Face Dataset},
url = {https://huggingface.co/datasets/zhangdw/TurnBack-10pct}
}
A compact route-reversal sample for balanced geospatial cognition experiments.
- Downloads last month
- 43