Datasets:
id stringlengths 26 89 | group stringclasses 2
values | arity stringclasses 3
values | n_flaws int64 1 3 | combination stringclasses 106
values | flaws listlengths 1 3 | library stringclasses 3
values | chart_type stringclasses 21
values | clean bool 1
class | source_id stringlengths 8 8 | ground_truth_code stringlengths 766 10.2k | table stringlengths 73 9.5k | summary stringlengths 1.16k 3.84k | flawed_code stringlengths 766 10.3k | flawed_image stringlengths 47 110 | ground_truth_image stringlengths 53 116 | initial_code stringlengths 766 10.3k | initial_image stringlengths 47 110 | track stringclasses 1
value | prompt_file stringclasses 106
values | prompt stringclasses 106
values |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
0d94c4af__axis-truncation__llm | A | single | 1 | axis-truncation | [
{
"category": "axis-truncation",
"description": "Clip one truth-backed plotted endpoint by limiting the value axis so the point is no longer visible."
}
] | seaborn | Violin Plot | false | 0d94c4af | import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
stages = [
"Homepage Visit",
"Product View",
"Add to Cart",
"Checkout Start",
"Payment Success"
]
metro_areas = [
"New York",
"Los Angeles",
"Chicago",
"San Francisco",
"Dallas",
"Houston",
"Bosto... | ```
Metro Area,Stage,Session Time (s)
New York,Homepage Visit,45
New York,Homepage Visit,50
New York,Homepage Visit,55
New York,Homepage Visit,60
New York,Homepage Visit,48
New York,Homepage Visit,52
New York,Homepage Visit,58
New York,Homepage Visit,62
New York,Homepage Visit,47
New York,Homepage Visit,53
New York,Pro... | This image is a violin plot titled "E-Commerce Session Time Distribution by Stage and Metro Area," which illustrates the distribution of session times (in seconds) across different stages of an e-commerce conversion process for twelve major metropolitan areas in the United States. The x-axis represents five distinct co... | import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
stages = [
"Homepage Visit",
"Product View",
"Add to Cart",
"Checkout Start",
"Payment Success"
]
metro_areas = [
"New York",
"Los Angeles",
"Chicago",
"San Francisco",
"Dallas",
"Houston",
"Bosto... | instances/0d94c4af__axis-truncation__llm/flawed.png | instances/0d94c4af__axis-truncation__llm/ground_truth.png | import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
stages = [
"Homepage Visit",
"Product View",
"Add to Cart",
"Checkout Start",
"Payment Success"
]
metro_areas = [
"New York",
"Los Angeles",
"Chicago",
"San Francisco",
"Dallas",
"Houston",
"Bosto... | instances/0d94c4af__axis-truncation__llm/flawed.png | A | prompts/06d4e5f3d897a6d5a2d7c4fda74bc12fe2a73b2006aa774fbb0001c1e0ce4a7b.txt | chart.py builds a chart from the data in table.csv and saves it as output.png.
The chart it currently produces is not faithful to that data: at least one thing
about it contradicts what table.csv contains or what summary.txt says the chart
shows. chart_current.png is how the chart looks right now.
Find what is wrong, ... |
90014ade__axis-truncation__llm | B | single | 1 | axis-truncation | [
{
"category": "axis-truncation",
"description": "Clip one truth-backed plotted endpoint by limiting the value axis so the point is no longer visible."
}
] | matplotlib | Histogram | false | 90014ade | import matplotlib.pyplot as plt
import numpy as np
education_data = {
'Country': ['Burkina Faso'] * 8 + ['Mali'] * 8 + ['Niger'] * 8,
'Gender': ['Male'] * 4 + ['Female'] * 4 + ['Male'] * 4 + ['Female'] * 4 + ['Male'] * 4 + ['Female'] * 4,
'Percentage': [
72.1, 68.4, 70.3, 65.8,
58.7, 45.9, ... | Country,Gender,Percentage
Burkina Faso,Male,65.8
Burkina Faso,Male,68.4
Burkina Faso,Male,70.3
Burkina Faso,Male,72.1
Burkina Faso,Female,40.5
Burkina Faso,Female,45.9
Burkina Faso,Female,52.1
Burkina Faso,Female,58.7
Mali,Male,35.9
Mali,Male,37.6
Mali,Male,40.2
Mali,Male,42.8
Mali,Female,25.6
Mali,Female,29.8
Mali,Fem... | The chart presented is a histogram that illustrates the distribution of primary education participation rates, specifically focusing on primary enrollment and dropout percentages across three countries and differentiated by gender. The title of the chart, "Distribution of Primary Education Participation Rates," is prom... | import matplotlib.pyplot as plt
import numpy as np
education_data = {
'Country': ['Burkina Faso'] * 8 + ['Mali'] * 8 + ['Niger'] * 8,
'Gender': ['Male'] * 4 + ['Female'] * 4 + ['Male'] * 4 + ['Female'] * 4 + ['Male'] * 4 + ['Female'] * 4,
'Percentage': [
72.1, 68.4, 70.3, 65.8,
58.7, 45.9, ... | instances/90014ade__axis-truncation__llm/flawed.png | instances/90014ade__axis-truncation__llm/ground_truth.png | import matplotlib.pyplot as plt
import numpy as np
education_data = {
'Country': ['Burkina Faso'] * 8 + ['Mali'] * 8 + ['Niger'] * 8,
'Gender': ['Male'] * 4 + ['Female'] * 4 + ['Male'] * 4 + ['Female'] * 4 + ['Male'] * 4 + ['Female'] * 4,
'Percentage': [
72.1, 68.4, 70.3, 65.8,
58.7, 45.9, ... | instances/90014ade__axis-truncation__llm/flawed.png | A | prompts/06d4e5f3d897a6d5a2d7c4fda74bc12fe2a73b2006aa774fbb0001c1e0ce4a7b.txt | chart.py builds a chart from the data in table.csv and saves it as output.png.
The chart it currently produces is not faithful to that data: at least one thing
about it contradicts what table.csv contains or what summary.txt says the chart
shows. chart_current.png is how the chart looks right now.
Find what is wrong, ... |
443c654f__axis-truncation__llm | A | single | 1 | axis-truncation | [
{
"category": "axis-truncation",
"description": "Clip one truth-backed plotted endpoint by limiting the value axis so the point is no longer visible."
}
] | seaborn | Treemap | false | 443c654f | import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
data = {
'Country': ['China', 'India', 'Vietnam', 'South Korea', 'Japan',
'Thailand', 'Malaysia', 'Indonesia', 'Philippines', 'Bangladesh',
'Cambodia', 'Taiwan', 'Singapore', 'Pakistan', 'Sri Lanka'],
'Exp... | Country,Export_Value_USD_Billion,Sector
China,3450.2,Electronics
India,420.5,Textiles
Vietnam,354.8,Electronics
South Korea,644.3,Semiconductors
Japan,756.1,Automobiles
Thailand,285.6,Electronics
Malaysia,270.4,Electronics
Indonesia,230.7,Textiles
Philippines,102.3,Electronics
Bangladesh,52.8,Textiles
Cambodia,22.5,Tex... | The chart titled "Asia-Pacific Export Market Share by Country and Sector (2023)" is a horizontal bar chart that visualizes the export values of various countries in the Asia-Pacific region, segmented by industrial sectors. The x-axis represents the export value in USD billions, ranging from 0 to 3500 billion, while the... | import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
data = {
'Country': ['China', 'India', 'Vietnam', 'South Korea', 'Japan',
'Thailand', 'Malaysia', 'Indonesia', 'Philippines', 'Bangladesh',
'Cambodia', 'Taiwan', 'Singapore', 'Pakistan', 'Sri Lanka'],
'Exp... | instances/443c654f__axis-truncation__llm/flawed.png | instances/443c654f__axis-truncation__llm/ground_truth.png | import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
data = {
'Country': ['China', 'India', 'Vietnam', 'South Korea', 'Japan',
'Thailand', 'Malaysia', 'Indonesia', 'Philippines', 'Bangladesh',
'Cambodia', 'Taiwan', 'Singapore', 'Pakistan', 'Sri Lanka'],
'Exp... | instances/443c654f__axis-truncation__llm/flawed.png | A | prompts/06d4e5f3d897a6d5a2d7c4fda74bc12fe2a73b2006aa774fbb0001c1e0ce4a7b.txt | chart.py builds a chart from the data in table.csv and saves it as output.png.
The chart it currently produces is not faithful to that data: at least one thing
about it contradicts what table.csv contains or what summary.txt says the chart
shows. chart_current.png is how the chart looks right now.
Find what is wrong, ... |
8706e3a9__axis-truncation__llm | A | single | 1 | axis-truncation | [
{
"category": "axis-truncation",
"description": "Clip one truth-backed plotted endpoint by limiting the value axis so the point is no longer visible."
}
] | matplotlib | Stem Plot | false | 8706e3a9 | import matplotlib.pyplot as plt
import pandas as pd
data = {
'Country': ['Germany', 'France', 'Italy', 'Peru', 'Spain', 'Thailand',
'Brazil', 'Mexico', 'India', 'Japan', 'Canada', 'South Korea',
'Australia', 'Indonesia', 'Argentina'],
'Largest City Impact (%)': [
-12.4, ... | Country,Largest City Impact (%),Urban Agglomeration Impact (%)
Germany,-12.4,15.2
France,-14.1,16.8
Italy,-13.5,18.3
Peru,-37.2,29.6
Spain,-15.7,22.1
Thailand,-28.9,14.5
Brazil,-21.3,19.4
Mexico,-23.8,17.9
India,-11.8,13.2
Japan,-17.5,20.3
Canada,-14.9,16.7
South Korea,-18.6,15.9
Australia,-9.2,12.3
Indonesia,-22.7,21.... | The chart presented is a stem plot that illustrates the population impact of the largest city versus urban agglomeration across various countries between the years 2003 and 2006. The title of the chart, "Population Impact: Largest City vs Urban Agglomeration (2003-2006)," clearly indicates its focus. The x-axis is labe... | import matplotlib.pyplot as plt
import pandas as pd
data = {
'Country': ['Germany', 'France', 'Italy', 'Peru', 'Spain', 'Thailand',
'Brazil', 'Mexico', 'India', 'Japan', 'Canada', 'South Korea',
'Australia', 'Indonesia', 'Argentina'],
'Largest City Impact (%)': [
-12.4, ... | instances/8706e3a9__axis-truncation__llm/flawed.png | instances/8706e3a9__axis-truncation__llm/ground_truth.png | import matplotlib.pyplot as plt
import pandas as pd
data = {
'Country': ['Germany', 'France', 'Italy', 'Peru', 'Spain', 'Thailand',
'Brazil', 'Mexico', 'India', 'Japan', 'Canada', 'South Korea',
'Australia', 'Indonesia', 'Argentina'],
'Largest City Impact (%)': [
-12.4, ... | instances/8706e3a9__axis-truncation__llm/flawed.png | A | prompts/06d4e5f3d897a6d5a2d7c4fda74bc12fe2a73b2006aa774fbb0001c1e0ce4a7b.txt | chart.py builds a chart from the data in table.csv and saves it as output.png.
The chart it currently produces is not faithful to that data: at least one thing
about it contradicts what table.csv contains or what summary.txt says the chart
shows. chart_current.png is how the chart looks right now.
Find what is wrong, ... |
2b83df4a__axis-truncation__llm | B | single | 1 | axis-truncation | [
{
"category": "axis-truncation",
"description": "Clip one truth-backed plotted endpoint by limiting the value axis so the point is no longer visible."
}
] | seaborn | Tornado Chart | false | 2b83df4a | import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
countries = ['Singapore', 'Malaysia', 'Thailand', 'Vietnam', 'Indonesia', 'Philippines', 'Brunei']
categories = [
'Business Setup Ease (Index)',
'Operational Cost Efficiency (Index)',
'Approval Efficiency (Index)',... | Category,Negative_Impact,Positive_Impact
Business Setup Ease (Index),-22,0
Physical Infrastructure (Index),-20,0
Transparency & Governance (Index),-18,3
Digital Infrastructure (Index),-17,5
Approval Efficiency (Index),-21,2
Operational Cost Efficiency (Index),-20,2 | The chart presented is a tornado chart titled "ASEAN Business Environment: Negative vs Positive Impacts by Category," which visually compares the negative and positive impacts on various aspects of the business environment across ASEAN regions. This horizontal bar chart features six distinct categories along the vertic... | import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
countries = ['Singapore', 'Malaysia', 'Thailand', 'Vietnam', 'Indonesia', 'Philippines', 'Brunei']
categories = [
'Business Setup Ease (Index)',
'Operational Cost Efficiency (Index)',
'Approval Efficiency (Index)',... | instances/2b83df4a__axis-truncation__llm/flawed.png | instances/2b83df4a__axis-truncation__llm/ground_truth.png | import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
countries = ['Singapore', 'Malaysia', 'Thailand', 'Vietnam', 'Indonesia', 'Philippines', 'Brunei']
categories = [
'Business Setup Ease (Index)',
'Operational Cost Efficiency (Index)',
'Approval Efficiency (Index)',... | instances/2b83df4a__axis-truncation__llm/flawed.png | A | prompts/06d4e5f3d897a6d5a2d7c4fda74bc12fe2a73b2006aa774fbb0001c1e0ce4a7b.txt | chart.py builds a chart from the data in table.csv and saves it as output.png.
The chart it currently produces is not faithful to that data: at least one thing
about it contradicts what table.csv contains or what summary.txt says the chart
shows. chart_current.png is how the chart looks right now.
Find what is wrong, ... |
9394b893__axis-truncation__llm | B | single | 1 | axis-truncation | [
{
"category": "axis-truncation",
"description": "Clip one truth-backed plotted endpoint by limiting the value axis so the point is no longer visible."
}
] | matplotlib | Radar Chart | false | 9394b893 | import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
regions = ["North America", "Western Europe", "East Asia", "Middle East", "Latin America"]
metrics = {
"Undergrad Enrollment": {
"North America": 4.5,
"Western Europe": -9.2,
"East Asia": 2.5,
"Middle East": 1.5,... | Region,Undergrad Enrollment,Graduate Enrollment,Research Funding,Faculty Hiring,Campus Infrastructure,International Collaboration
North America,4.5,8.2,48.0,-4.5,38.0,12.0
Western Europe,-9.2,3.5,50.0,2.0,34.0,15.0
East Asia,2.5,12.5,40.0,-4.2,43.0,9.0
Middle East,1.5,6.2,36.5,6.8,42.5,7.0
Latin America,-1.5,5.3,28.0,-... | The image is a radar chart titled "University Performance Metrics by Region: Average Growth and Investment," which visually compares six key performance metrics across five different global regions. The chart uses a circular layout with six axes radiating from the center, each representing one of the following metrics:... | import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
regions = ["North America", "Western Europe", "East Asia", "Middle East", "Latin America"]
metrics = {
"Undergrad Enrollment": {
"North America": 4.5,
"Western Europe": -9.2,
"East Asia": 2.5,
"Middle East": 1.5,... | instances/9394b893__axis-truncation__llm/flawed.png | instances/9394b893__axis-truncation__llm/ground_truth.png | import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
regions = ["North America", "Western Europe", "East Asia", "Middle East", "Latin America"]
metrics = {
"Undergrad Enrollment": {
"North America": 4.5,
"Western Europe": -9.2,
"East Asia": 2.5,
"Middle East": 1.5,... | instances/9394b893__axis-truncation__llm/flawed.png | A | prompts/06d4e5f3d897a6d5a2d7c4fda74bc12fe2a73b2006aa774fbb0001c1e0ce4a7b.txt | chart.py builds a chart from the data in table.csv and saves it as output.png.
The chart it currently produces is not faithful to that data: at least one thing
about it contradicts what table.csv contains or what summary.txt says the chart
shows. chart_current.png is how the chart looks right now.
Find what is wrong, ... |
e1e7c85a__axis-truncation__llm | B | single | 1 | axis-truncation | [
{
"category": "axis-truncation",
"description": "Clip one truth-backed plotted endpoint by limiting the value axis so the point is no longer visible."
}
] | seaborn | Area Chart | false | e1e7c85a | import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
data = {
"Year": [2015, 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023],
"High School or Below": [35.1, 34.2, 33.5, 32.1, 30.8, 29.6, 28.9, 28.5, 28.3],
"Diploma/Associate": [20.5, 20.8, 20.6, 20.1, 19.9, 19.7, 19.8, 19.8, 19.8],
... | Year,High School or Below,Diploma/Associate,Bachelor's Degree (STEM),Bachelor's Degree (Non-STEM),Master's Degree (STEM),Master's Degree (Non-STEM),Professional Degree (Law),Professional Degree (Medicine),Doctorate (STEM),Doctorate (Non-STEM)
2015,35.1,20.5,12.3,11.2,4.1,3.2,1.5,1.2,2.4,2.5
2016,34.2,20.8,13.0,11.5,4.5... | The image presents an area chart titled "Trends in Educational Attainment in the Saudi Labor Force (2015-2023)," illustrating how the distribution of educational levels within the labor force has evolved over a nine-year period. The chart features the percentage of the labor force on the vertical axis, ranging from 0 t... | import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
data = {
"Year": [2015, 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023],
"High School or Below": [35.1, 34.2, 33.5, 32.1, 30.8, 29.6, 28.9, 28.5, 28.3],
"Diploma/Associate": [20.5, 20.8, 20.6, 20.1, 19.9, 19.7, 19.8, 19.8, 19.8],
... | instances/e1e7c85a__axis-truncation__llm/flawed.png | instances/e1e7c85a__axis-truncation__llm/ground_truth.png | import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
data = {
"Year": [2015, 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023],
"High School or Below": [35.1, 34.2, 33.5, 32.1, 30.8, 29.6, 28.9, 28.5, 28.3],
"Diploma/Associate": [20.5, 20.8, 20.6, 20.1, 19.9, 19.7, 19.8, 19.8, 19.8],
... | instances/e1e7c85a__axis-truncation__llm/flawed.png | A | prompts/06d4e5f3d897a6d5a2d7c4fda74bc12fe2a73b2006aa774fbb0001c1e0ce4a7b.txt | chart.py builds a chart from the data in table.csv and saves it as output.png.
The chart it currently produces is not faithful to that data: at least one thing
about it contradicts what table.csv contains or what summary.txt says the chart
shows. chart_current.png is how the chart looks right now.
Find what is wrong, ... |
78a12e5d__axis-truncation__llm | B | single | 1 | axis-truncation | [
{
"category": "axis-truncation",
"description": "Clip one truth-backed plotted endpoint by limiting the value axis so the point is no longer visible."
}
] | matplotlib | Rose Chart | false | 78a12e5d | import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
data = {
"Region": [
"East Asia", "East Asia", "East Asia",
"Central Europe", "Central Europe", "Central Europe",
"Southeast Europe", "Southeast Europe", "Southeast Europe",
"Latin America", "Latin America", "Lat... | Region,Reform_Type,Completion_Rate
Central Europe,Fiscal,60.0
Central Europe,Trade,52.0
Central Europe,Monetary,80.0
East Africa,Fiscal,30.0
East Africa,Trade,22.0
East Africa,Monetary,42.0
East Asia,Fiscal,55.0
East Asia,Trade,48.0
East Asia,Monetary,72.0
Latin America,Fiscal,42.0
Latin America,Trade,35.0
Latin Americ... | The image is a rose chart, also known as a polar area chart, that illustrates the regional macroeconomic reform completion rates by type over the period from 2020 to 2022. The chart features three categories of reforms: Fiscal, Trade, and Monetary, each positioned along the circular axis at equal intervals. The radial ... | import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
data = {
"Region": [
"East Asia", "East Asia", "East Asia",
"Central Europe", "Central Europe", "Central Europe",
"Southeast Europe", "Southeast Europe", "Southeast Europe",
"Latin America", "Latin America", "Lat... | instances/78a12e5d__axis-truncation__llm/flawed.png | instances/78a12e5d__axis-truncation__llm/ground_truth.png | import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
data = {
"Region": [
"East Asia", "East Asia", "East Asia",
"Central Europe", "Central Europe", "Central Europe",
"Southeast Europe", "Southeast Europe", "Southeast Europe",
"Latin America", "Latin America", "Lat... | instances/78a12e5d__axis-truncation__llm/flawed.png | A | prompts/06d4e5f3d897a6d5a2d7c4fda74bc12fe2a73b2006aa774fbb0001c1e0ce4a7b.txt | chart.py builds a chart from the data in table.csv and saves it as output.png.
The chart it currently produces is not faithful to that data: at least one thing
about it contradicts what table.csv contains or what summary.txt says the chart
shows. chart_current.png is how the chart looks right now.
Find what is wrong, ... |
aa83f6b3__axis-truncation__llm | A | single | 1 | axis-truncation | [
{
"category": "axis-truncation",
"description": "Clip one truth-backed plotted endpoint by limiting the value axis so the point is no longer visible."
}
] | seaborn | Histogram | false | aa83f6b3 | import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
data = {
'Region': [
'Metro North', 'Metro North', 'Metro North', 'Metro North', 'Metro North', 'Metro North', 'Metro North',
'Metro South', 'Metro South', 'Metro South', 'Metro South', 'Metro South', 'Metro South', 'Metro So... | Region,FundingPerMonthMillions,NumberOfProjects
Metro North,0.7-0.9,1
Metro South,0.7-0.9,1
Rural West,0.7-0.9,1
Suburban East,0.7-0.9,1
Metro Central,0.7-0.9,0
Suburban North,0.7-0.9,0
Rural North,0.7-0.9,1
Metro North,0.9-1.1,3
Metro South,0.9-1.1,2
Suburban East,0.9-1.1,2
Rural West,0.9-1.1,1
Metro Central,0.9-1.1,1... | This image is a stacked histogram that illustrates the distribution of monthly funding across various regions, measured in millions of U.S. dollars. The chart is titled "Distribution of Monthly Funding Across Regions," with the x-axis labeled "Monthly Funding (Millions USD)" and the y-axis labeled "Number of Projects."... | import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
data = {
'Region': [
'Metro North', 'Metro North', 'Metro North', 'Metro North', 'Metro North', 'Metro North', 'Metro North',
'Metro South', 'Metro South', 'Metro South', 'Metro South', 'Metro South', 'Metro South', 'Metro So... | instances/aa83f6b3__axis-truncation__llm/flawed.png | instances/aa83f6b3__axis-truncation__llm/ground_truth.png | import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
data = {
'Region': [
'Metro North', 'Metro North', 'Metro North', 'Metro North', 'Metro North', 'Metro North', 'Metro North',
'Metro South', 'Metro South', 'Metro South', 'Metro South', 'Metro South', 'Metro South', 'Metro So... | instances/aa83f6b3__axis-truncation__llm/flawed.png | A | prompts/06d4e5f3d897a6d5a2d7c4fda74bc12fe2a73b2006aa774fbb0001c1e0ce4a7b.txt | chart.py builds a chart from the data in table.csv and saves it as output.png.
The chart it currently produces is not faithful to that data: at least one thing
about it contradicts what table.csv contains or what summary.txt says the chart
shows. chart_current.png is how the chart looks right now.
Find what is wrong, ... |
30127d7e__axis-truncation__llm | B | single | 1 | axis-truncation | [
{
"category": "axis-truncation",
"description": "Clip one truth-backed plotted endpoint by limiting the value axis so the point is no longer visible."
}
] | matplotlib | Box Plot | false | 30127d7e | import matplotlib.pyplot as plt
data = {
'Region': [
'Sub-Saharan Africa', 'Sub-Saharan Africa', 'Sub-Saharan Africa',
'South Asia', 'South Asia', 'South Asia',
'Southeast Asia & Pacific', 'Southeast Asia & Pacific', 'Southeast Asia & Pacific',
'Latin America & Caribbean', 'Latin Am... | Region,Min,Q1,Median,Q3,Max
Sub-Saharan Africa,40.1,40.1,42.5,45.8,45.8
South Asia,37.6,37.6,39.2,41.9,41.9
Southeast Asia & Pacific,50.2,50.2,53.4,56.7,56.7
Latin America & Caribbean,81.0,81.0,83.3,85.2,85.2
East Asia,65.8,65.8,67.2,69.5,69.5
Developed Nations,78.3,78.3,80.1,82.4,82.4
Europe & Central Asia,74.5,74.5,7... | The image is a box plot titled "Distribution of Urban Population Percentage by Region (2022)," which visually represents the spread and central tendency of urban population percentages across various global regions. The x-axis is labeled "Region" and lists nine distinct regions: Sub-Saharan Africa, South Asia, Southeas... | import matplotlib.pyplot as plt
data = {
'Region': [
'Sub-Saharan Africa', 'Sub-Saharan Africa', 'Sub-Saharan Africa',
'South Asia', 'South Asia', 'South Asia',
'Southeast Asia & Pacific', 'Southeast Asia & Pacific', 'Southeast Asia & Pacific',
'Latin America & Caribbean', 'Latin Am... | instances/30127d7e__axis-truncation__llm/flawed.png | instances/30127d7e__axis-truncation__llm/ground_truth.png | import matplotlib.pyplot as plt
data = {
'Region': [
'Sub-Saharan Africa', 'Sub-Saharan Africa', 'Sub-Saharan Africa',
'South Asia', 'South Asia', 'South Asia',
'Southeast Asia & Pacific', 'Southeast Asia & Pacific', 'Southeast Asia & Pacific',
'Latin America & Caribbean', 'Latin Am... | instances/30127d7e__axis-truncation__llm/flawed.png | A | prompts/06d4e5f3d897a6d5a2d7c4fda74bc12fe2a73b2006aa774fbb0001c1e0ce4a7b.txt | chart.py builds a chart from the data in table.csv and saves it as output.png.
The chart it currently produces is not faithful to that data: at least one thing
about it contradicts what table.csv contains or what summary.txt says the chart
shows. chart_current.png is how the chart looks right now.
Find what is wrong, ... |
753c8334__axis-truncation__llm | B | single | 1 | axis-truncation | [
{
"category": "axis-truncation",
"description": "Clip one truth-backed plotted endpoint by limiting the value axis so the point is no longer visible."
}
] | seaborn | Line Chart | false | 753c8334 | import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
data = {
'Quarter': ['Q1-2023', 'Q2-2023', 'Q3-2023', 'Q4-2023', 'Q1-2024', 'Q2-2024'],
'Health Services': [4.2, 4.5, 4.8, 5.1, 5.3, 5.6],
'Defense Operations': [2.1, 2.3, 2.2, 2.4, 2.5, 2.7],
'Public Education': [3.5, 3.7, 3.9, ... | Quarter,Health Services,Defense Operations,Public Education,Transportation,Energy Sector,Tech Innovation,Infrastructure
Q1-2023,4.2,2.1,3.5,3.1,1.8,5.0,3.0
Q2-2023,4.5,2.3,3.7,3.4,1.9,4.8,3.2
Q3-2023,4.8,2.2,3.9,3.6,2.0,5.2,3.5
Q4-2023,5.1,2.4,4.1,3.8,2.2,5.4,3.7
Q1-2024,5.3,2.5,4.3,3.9,2.3,5.6,3.9
Q2-2024,5.6,2.7,4.4,... | The image is a line chart titled "Quarterly Budget Impact Trends by Department (2023-2024)," which illustrates the budget impact in billions of dollars for various government departments over a span of six quarters, from Q1 2023 to Q2 2024. The x-axis represents the quarters, while the y-axis indicates the budget impac... | import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
data = {
'Quarter': ['Q1-2023', 'Q2-2023', 'Q3-2023', 'Q4-2023', 'Q1-2024', 'Q2-2024'],
'Health Services': [4.2, 4.5, 4.8, 5.1, 5.3, 5.6],
'Defense Operations': [2.1, 2.3, 2.2, 2.4, 2.5, 2.7],
'Public Education': [3.5, 3.7, 3.9, ... | instances/753c8334__axis-truncation__llm/flawed.png | instances/753c8334__axis-truncation__llm/ground_truth.png | import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
data = {
'Quarter': ['Q1-2023', 'Q2-2023', 'Q3-2023', 'Q4-2023', 'Q1-2024', 'Q2-2024'],
'Health Services': [4.2, 4.5, 4.8, 5.1, 5.3, 5.6],
'Defense Operations': [2.1, 2.3, 2.2, 2.4, 2.5, 2.7],
'Public Education': [3.5, 3.7, 3.9, ... | instances/753c8334__axis-truncation__llm/flawed.png | A | prompts/06d4e5f3d897a6d5a2d7c4fda74bc12fe2a73b2006aa774fbb0001c1e0ce4a7b.txt | chart.py builds a chart from the data in table.csv and saves it as output.png.
The chart it currently produces is not faithful to that data: at least one thing
about it contradicts what table.csv contains or what summary.txt says the chart
shows. chart_current.png is how the chart looks right now.
Find what is wrong, ... |
c13a4570__axis-truncation__llm | A | single | 1 | axis-truncation | [
{
"category": "axis-truncation",
"description": "Clip one truth-backed plotted endpoint by limiting the value axis so the point is no longer visible."
}
] | matplotlib | Swarm Plot | false | c13a4570 | import matplotlib.pyplot as plt
import pandas as pd
data = {
'EducationStage': ['Primary', 'Primary', 'Lower Secondary', 'Lower Secondary',
'Upper Secondary', 'Upper Secondary', 'Tertiary', 'Tertiary',
'Primary', 'Primary', 'Lower Secondary', 'Lower Secondary',
... | Year,EducationStage,IncomeGroup,EnrollmentRateGap
2015,Primary,High Income,-1.2
2019,Primary,High Income,-1.8
2023,Primary,High Income,-2.1
2015,Primary,Low Income,-45.3
2019,Primary,Low Income,-49.0
2023,Primary,Low Income,-52.4
2016,Lower Secondary,High Income,-7.8
2020,Lower Secondary,High Income,-8.5
2023,Lower Sec... | The chart presented is a scatter plot that illustrates enrollment rate gaps by income group and education stage from 2015 to 2023. The x-axis represents the enrollment rate gap in percentage, ranging from -90% to 0%, where negative values indicate lower enrollment rates compared to a reference standard, while the y-axi... | import matplotlib.pyplot as plt
import pandas as pd
data = {
'EducationStage': ['Primary', 'Primary', 'Lower Secondary', 'Lower Secondary',
'Upper Secondary', 'Upper Secondary', 'Tertiary', 'Tertiary',
'Primary', 'Primary', 'Lower Secondary', 'Lower Secondary',
... | instances/c13a4570__axis-truncation__llm/flawed.png | instances/c13a4570__axis-truncation__llm/ground_truth.png | import matplotlib.pyplot as plt
import pandas as pd
data = {
'EducationStage': ['Primary', 'Primary', 'Lower Secondary', 'Lower Secondary',
'Upper Secondary', 'Upper Secondary', 'Tertiary', 'Tertiary',
'Primary', 'Primary', 'Lower Secondary', 'Lower Secondary',
... | instances/c13a4570__axis-truncation__llm/flawed.png | A | prompts/06d4e5f3d897a6d5a2d7c4fda74bc12fe2a73b2006aa774fbb0001c1e0ce4a7b.txt | chart.py builds a chart from the data in table.csv and saves it as output.png.
The chart it currently produces is not faithful to that data: at least one thing
about it contradicts what table.csv contains or what summary.txt says the chart
shows. chart_current.png is how the chart looks right now.
Find what is wrong, ... |
7b1daade__axis-truncation__llm | A | single | 1 | axis-truncation | [
{
"category": "axis-truncation",
"description": "Clip one truth-backed plotted endpoint by limiting the value axis so the point is no longer visible."
}
] | seaborn | Scatter Plot | false | 7b1daade | import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
data = {
'Country': ['Germany', 'Germany', 'Germany', 'France', 'France', 'France', 'Spain', 'Spain', 'Spain',
'Brazil', 'Brazil', 'Brazil', 'Italy', 'Italy', 'Italy', 'India', 'India', 'India',
'Canada', 'Can... | Country,Region,Avg_GDP_Growth,Avg_Debt_to_GDP,GDP_Per_Capita
Germany,Europe,0.033333,65.933333,46788
France,Europe,0.366667,108.366667,38114
Spain,Europe,-0.966667,111.300000,26410
Brazil,Americas,0.900000,84.133333,7674
Italy,Europe,-0.533333,147.133333,31167
India,Asia,1.800000,82.033333,2094
Canada,Americas,0.400000... | This chart is a bubble scatter plot titled *Economic Performance: GDP Growth vs Debt-to-GDP Ratio (2019-2021)*, which examines the relationship between average GDP growth and average debt-to-GDP ratios across various countries from 2019 to 2021. The x-axis represents the average GDP growth in percentage, ranging approx... | import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
data = {
'Country': ['Germany', 'Germany', 'Germany', 'France', 'France', 'France', 'Spain', 'Spain', 'Spain',
'Brazil', 'Brazil', 'Brazil', 'Italy', 'Italy', 'Italy', 'India', 'India', 'India',
'Canada', 'Can... | instances/7b1daade__axis-truncation__llm/flawed.png | instances/7b1daade__axis-truncation__llm/ground_truth.png | import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
data = {
'Country': ['Germany', 'Germany', 'Germany', 'France', 'France', 'France', 'Spain', 'Spain', 'Spain',
'Brazil', 'Brazil', 'Brazil', 'Italy', 'Italy', 'Italy', 'India', 'India', 'India',
'Canada', 'Can... | instances/7b1daade__axis-truncation__llm/flawed.png | A | prompts/06d4e5f3d897a6d5a2d7c4fda74bc12fe2a73b2006aa774fbb0001c1e0ce4a7b.txt | chart.py builds a chart from the data in table.csv and saves it as output.png.
The chart it currently produces is not faithful to that data: at least one thing
about it contradicts what table.csv contains or what summary.txt says the chart
shows. chart_current.png is how the chart looks right now.
Find what is wrong, ... |
End of preview. Expand in Data Studio
ChartRepairBench v1.4
ChartRepairBench evaluates agents that identify and repair errors in chart
plotting code using a chart image, code, table and summary.
Protocol: abc-uncapped-v1.
| Track | Tasks | Instruction |
|---|---|---|
| A | 1,900 | Repair the chart with its flaw categories disclosed. |
| B | 1,900 | Identify and repair the flaws in the chart. |
| C | 2,200 | Determine whether repair is needed; preserve correct charts. |
A and B contain 1,900 flawed tasks. C contains the same flawed tasks plus 300 clean reference tasks. All tracks provide all four input modalities. Episode step and wall-time limits are disabled; individual HTTP, Python-tool and scoring timeouts remain enabled.
Dataset
| Flawed subset | Tasks | Scenarios |
|---|---|---|
| Single | 400 | 10 |
| Double | 900 | 36 |
| Triple | 600 | 60 |
Charts use matplotlib, seaborn and plotly. The flawed set draws on 854
distinct ChartNet core_permissive source charts. Multiple tasks can share a
source. The group labels A/B identify stratified subsets, independently of
evaluation tracks; they are not source-disjoint. Clean tasks have no group.
dataset.json and clean_dataset.json contain the instance catalogs;
tracks/*.json define track membership and reference the exact task prompts.
suite.json defines the protocol. base_bindings.json, files.sha256.json
and package_checksums.json record file identities and hashes.
Load a track
from datasets import load_dataset
ds = load_dataset("aydamdonelly0/chartrepairbench-v1.3", "B", split="test", revision="<commit-sha>")
initial_code = ds[0]["initial_code"]
Specify A, B or C and pin the repository revision. initial_code and
initial_image select the starting chart. flawed_code and flawed_image
are null for clean tasks. Image fields are repository-relative PNG paths.
data.jsonl contains the 1,900-task flawed catalog.
Dataset rows include reference artifacts and answer metadata for scoring.
Give the agent only the track prompt and its four input files: chart.py,
chart_current.png, table.csv, summary.txt. Keep reference answers, clean
status, flaw metadata and category-bearing IDs outside the agent workspace.
Track A's prompt discloses the diagnosis; Track C uses the same prompt for
clean and flawed tasks.
Run and score
See RUNNING.md for installation and run_suite.py --help for launch options.
The isolated runner requires macOS and a separately obtained, pinned sandbox
dependency. Exact task instructions are under prompts/; see PROMPT.md.
python scoring/score_submission.py --instance instances/<flawed-id> \
--submission repaired.py --category "<combination>"
python scoring/score_submission.py --instance instances/<clean-id> \
--submission repaired.py --clean
The scorer reports executability, recovery, preservation and data fidelity, with binary variants where defined and separate strict/lenient comparisons. There is no weighted overall score. Recovery is undefined for clean charts; data fidelity can also be undefined. Exclude undefined values from their own means and report metric denominators and infrastructure errors.
Full recovery means every evaluated flaw property satisfies the scorer's rules.
Preservation compares figure objects and can penalize visually equivalent
redesigns or library changes. Use Python 3.12 and the pinned environment in
scoring/env-lock.txt for comparable captures. Record dataset, protocol and
runner revisions with evaluation results.
Licence
Derived from ChartNet core_permissive under CDLA-Permissive-2.0.
The external sandbox is a separately obtained dependency governed by its own
licence.
- Downloads last month
- 171