LiveCVEBench-io/chart_output/generate_boxplot.py

#!/usr/bin/env python3
"""
Plot boxplot for agent duration distributions.
"""

import pandas as pd
import matplotlib.pyplot as plt
import matplotlib.font_manager as fm
import numpy as np

# Use Liberation Serif (Times New Roman compatible)
try:
    fm.fontManager.addfont('/usr/share/fonts/truetype/liberation/LiberationSerif-Regular.ttf')
    fm.fontManager.addfont('/usr/share/fonts/truetype/liberation/LiberationSerif-Bold.ttf')
    plt.rcParams['font.family'] = 'Liberation Serif'
except:
    plt.rcParams['font.family'] = 'serif'

# Bold font for labels
bold_font = fm.FontProperties(fname='/usr/share/fonts/truetype/liberation/LiberationSerif-Bold.ttf', size=20)

# Style settings - consistent with other charts
plt.rcParams.update({
    'font.size': 20,
    'axes.labelsize': 24,
    'xtick.labelsize': 20,
    'ytick.labelsize': 21,
    'figure.dpi': 300,
    'savefig.dpi': 300,
})

# Read data
df = pd.read_excel('zsq_stats.xlsx')

# Agent names
agents = ['analyzer', 'generator', 'builder', 'validator', 'solver', 'checker']
agent_labels = ['Analyzer', 'Generator', 'Builder', 'Validator', 'Solver', 'Checker']

# Duration filter thresholds (in minutes)
duration_thresholds = {
    'analyzer': 30,
    'generator': 20,
    'builder': 60,
    'validator': 60,
    'solver': 60,
    'checker': 60
}

# Collect filtered duration data for each agent
duration_data = []
for agent in agents:
    duration_col = f'{agent}_duration_ms'
    duration_min = df[duration_col].dropna() / 1000 / 60
    threshold = duration_thresholds[agent]
    duration_filtered = duration_min[duration_min <= threshold]
    duration_data.append(duration_filtered.values)

# Create figure
fig, ax = plt.subplots(figsize=(10, 7))

# Create boxplot (hide outliers with showfliers=False, narrower boxes)
bp = ax.boxplot(duration_data, tick_labels=agent_labels, patch_artist=True, showfliers=False,
                widths=0.6)

# Colors - slightly darker (稍深的淡绿、淡橙、淡蓝、淡紫、淡红、浅绿)
colors = ['#7AB08A', '#D49A70', '#8AB4D8', '#A0A0C8', '#E0AAAA', '#A8D098']
edge_colors = ['#5A8868', '#B07848', '#6888A8', '#787898', '#C08080', '#80A870']

for patch, color, edge in zip(bp['boxes'], colors, edge_colors):
    patch.set_facecolor(color)
    patch.set_edgecolor(edge)
    patch.set_linewidth(2.5)
    patch.set_alpha(0.9)

# Style whiskers, caps, medians
for whisker in bp['whiskers']:
    whisker.set_color('#666666')
    whisker.set_linewidth(2)

for cap in bp['caps']:
    cap.set_color('#666666')
    cap.set_linewidth(2)

for median in bp['medians']:
    median.set_color('#333333')
    median.set_linewidth(2.5)

# Add mean values as text above each box (at the top whisker position)
for i, data in enumerate(duration_data):
    mean_val = np.mean(data)
    # Get the top whisker position (Q3 + 1.5*IQR or max within that range)
    q1 = np.percentile(data, 25)
    q3 = np.percentile(data, 75)
    iqr = q3 - q1
    whisker_top = min(np.max(data), q3 + 1.5 * iqr)
    ax.text(i + 1, whisker_top + 1.5, f'{mean_val:.1f}', ha='center', va='bottom',
            fontproperties=bold_font, color='#222222')

# Style the plot
ax.set_ylabel('Duration (min)', fontsize=24)
ax.tick_params(axis='x', labelsize=20)
ax.tick_params(axis='y', labelsize=21)

# Remove top and right spines
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)

plt.tight_layout()
plt.savefig('agent_duration_boxplot.png', dpi=300, bbox_inches='tight')
plt.savefig('agent_duration_boxplot.pdf', bbox_inches='tight')
print('Saved: agent_duration_boxplot.png')
print('Saved: agent_duration_boxplot.pdf')
plt.close()