File size: 31,629 Bytes
9cb15fd 20fb95e 9cb15fd d23511c 9cb15fd 44dcb13 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 900df7b 20fb95e 900df7b 20fb95e 900df7b 20fb95e 900df7b 20fb95e 900df7b 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 25a194d 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 3220848 20fb95e 9cb15fd 3220848 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 3220848 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 3220848 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 2ac4e00 83315e6 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e e519dd5 20fb95e 1e82f18 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd d23511c 9cb15fd 20fb95e 4d502ba 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd 20fb95e 9cb15fd d23511c 9cb15fd |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 |
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
LongBenchmark Results Visualization
"""
import json
import re
import pandas as pd
from pathlib import Path
import gradio as gr
import plotly.graph_objects as go
with open('./output/model_info.json', 'r', encoding='utf-8') as f:
MODLE_INFO_DICT = json.load(f)
def get_color(index):
"""Generate color based on index, using golden angle to ensure uniform and infinite color distribution"""
# Golden angle approx 137.508 degrees
hue = (index * 137.508) % 360
# Fixed saturation 70%, lightness 60%
return f"hsl({hue}, 70%, 60%)"
# Custom CSS
CUSTOM_CSS = """
/* Force title center */
h1 {
text-align: center;
display: block;
}
/* Header center */
#leaderboard_table th,
#leaderboard_table th button,
#leaderboard_table th span {
text-align: center !important;
justify-content: center !important;
}
/* Content column center: starting from 3rd column */
#leaderboard_table td:nth-child(n+3) {
text-align: center !important;
}
/* Make tab labels bold */
button[role="tab"] {
font-weight: bold !important;
}
"""
class ResultParser:
def __init__(self, output_dir: str):
self.output_dir = Path(output_dir)
self.results = []
def parse_filename(self, filename: str):
"""Parse filename to extract context length and thinking status"""
# Extract context length
context_match = re.search(r'context-(\d+)', filename)
context_length = int(context_match.group(1)) if context_match else 0
filename_lower = filename.lower()
# Check nonthinking
has_nonthinking = 'nonthinking' in filename_lower
# Check thinking
has_thinking = 'thinking' in filename_lower and not has_nonthinking
return context_length, has_thinking, has_nonthinking
def parse_result_file(self, model_name: str, file_path: Path):
"""Parse single result file"""
try:
with open(file_path, 'r', encoding='utf-8') as f:
data = json.load(f)
context_length, has_thinking, has_nonthinking = self.parse_filename(file_path.name)
# Use date field as evaluation date
eval_date = data.get('date', "Unknown")
# Extract BoN data
bon_data = {}
for bon_key in ['BoN-1', 'BoN-2', 'BoN-3']:
if bon_key in data and 'overall_metric' in data[bon_key]:
bon_data[bon_key] = data[bon_key]['overall_metric']
result = {
'model_name': model_name,
'eval_date': eval_date,
'context_length': context_length,
'has_thinking': has_thinking,
'has_nonthinking': has_nonthinking,
'overall_metric': data.get('average_overall_metric', 0.0),
'token_length_metrics': data.get('average_token_length_metric', {}),
'contextual_requirement': data.get('average_contextual_requirement_metric', {}),
'difficulty': data.get('average_difficulty_metric', {}),
'primary_task': data.get('average_primary_task_metric', {}),
'language': data.get('average_language_metric', {}),
'bon_data': bon_data, # Store BoN-1, BoN-2, BoN-3 overall_metric
'pass_at_k': {
'Pass@1': data.get('pass@1'),
'Pass@2': data.get('pass@2'),
'Pass@3': data.get('pass@3')
}
}
return result
except Exception as e:
print(f"Error parsing file {file_path}: {e}")
return None
def scan_all_results(self):
"""Scan all model result files"""
self.results = []
if not self.output_dir.exists():
print(f"Output directory does not exist: {self.output_dir}")
return
# Traverse all model directories
for model_dir in self.output_dir.iterdir():
if not model_dir.is_dir():
continue
model_name = model_dir.name
print(f"Scanning model: {model_name}")
# Find all _summary.json files
for file_path in model_dir.glob("*_summary.json"):
print(f" Parsing file: {file_path.name}")
result = self.parse_result_file(model_name, file_path)
if result:
self.results.append(result)
print(f"Total parsed {len(self.results)} result files")
def get_leaderboard_data(self):
"""Get leaderboard data"""
if not self.results:
return pd.DataFrame()
# Aggregate data by model name
model_groups = {}
for result in self.results:
model_name = result['model_name']
if model_name not in model_groups:
model_groups[model_name] = {
'dates': [],
'contexts': [],
'thinking_scores': [],
'non_thinking_scores': []
}
group = model_groups[model_name]
group['dates'].append(result['eval_date'])
group['contexts'].append(result['context_length'])
score = result['overall_metric']
if result['has_thinking']:
group['thinking_scores'].append(score)
else:
group['non_thinking_scores'].append(score)
leaderboard_data = []
for model_name, group in model_groups.items():
# Get latest date
valid_dates = [d for d in group['dates'] if d != "Unknown"]
latest_date = max(valid_dates) if valid_dates else "Unknown"
# Get max Context Window
max_context = max(group['contexts']) if group['contexts'] else 0
# Format truncated length
if max_context >= 1000000:
context_str = f"{max_context/1000000:.0f}M" if max_context % 1000000 == 0 else f"{max_context/1000000:.1f}M"
elif max_context >= 1000:
context_str = f"{max_context/1000:.0f}k" if max_context % 1000 == 0 else f"{max_context/1000:.1f}k"
else:
context_str = str(max_context)
# Get model type and context length
model_context = "-"
model_url = ""
if model_name in MODLE_INFO_DICT:
model_info = MODLE_INFO_DICT[model_name]
if isinstance(model_info, dict):
model_type = model_info.get("type", "Unknown")
model_context = model_info.get("context_length", "-")
model_url = model_info.get("url", "")
else:
model_type = str(model_info)
else:
model_type = "Unknown"
# Handle model name link and icon
display_model_name = model_name
if model_url:
display_model_name = f"[{display_model_name}]({model_url})"
# Calculate average score
nt_score_val = 0
nt_score_str = "-"
if group['non_thinking_scores']:
nt_score_val = sum(group['non_thinking_scores']) / len(group['non_thinking_scores'])
nt_score_str = f"{nt_score_val * 100:.2f}"
t_score_val = 0
t_score_str = "-"
if group['thinking_scores']:
t_score_val = sum(group['thinking_scores']) / len(group['thinking_scores'])
t_score_str = f"{t_score_val * 100:.2f}"
leaderboard_data.append({
'Model Name': display_model_name,
'Model Type': model_type,
'Context Length': model_context,
'Truncated Length': context_str,
'Non-Thinking Score': nt_score_str,
'Thinking Score': t_score_str,
'_sort_score': max(nt_score_val, t_score_val)
})
df = pd.DataFrame(leaderboard_data)
# Sort by highest score descending
if not df.empty:
df = df.sort_values('_sort_score', ascending=False).drop(columns=['_sort_score']).reset_index(drop=True)
return df
def get_display_name_for_result(result):
"""Get display name for model (append suffix based on thinking/nonthinking)"""
if result.get('has_nonthinking'):
return f"{result['model_name']}_nonthinking"
elif result.get('has_thinking'):
return f"{result['model_name']}_thinking"
else:
return result['model_name']
def get_model_color_index(model_name, all_models):
"""Get model index in color list"""
try:
return all_models.index(model_name)
except ValueError:
return 0
def create_contextual_requirement_chart(results, selected_models):
"""Create contextual requirement comparison bar chart"""
if not selected_models:
return go.Figure()
# Collect data
chart_data = {}
for result in results:
display_name = get_display_name_for_result(result)
if display_name in selected_models:
model_name = display_name
contextual_requirement = result['contextual_requirement']
# Store each model's result directly
if model_name not in chart_data:
chart_data[model_name] = {}
for req_type, score in contextual_requirement.items():
chart_data[model_name][req_type] = score * 100 # multiply by 100
# Create chart
fig = go.Figure()
# Get all requirement types
all_req_types = []
for result in results:
display_name = get_display_name_for_result(result)
if display_name in selected_models:
contextual_requirement = result['contextual_requirement']
for req_type in contextual_requirement.keys():
if req_type not in all_req_types:
all_req_types.append(req_type)
for model_name in selected_models:
if model_name in chart_data:
scores = [chart_data[model_name].get(req_type, 0) for req_type in all_req_types]
color_index = get_model_color_index(model_name, selected_models)
fig.add_trace(go.Bar(
name=model_name,
x=all_req_types,
y=scores,
marker_color=get_color(color_index),
text=[f"{score:.2f}" for score in scores], # keep 2 decimal places
textposition='auto'
))
fig.update_layout(
title='Performance Comparison on Different Context Requirements',
xaxis_title='Context Requirement Type',
yaxis_title='Average Score',
barmode='group',
autosize=True, # auto size
legend=dict(
orientation="h",
yanchor="top",
y=-0.25, # adjust lower
xanchor="center",
x=0.5
),
margin=dict(b=100) # increase bottom margin
)
return fig
def create_primary_task_radar_chart(results, selected_models):
"""Create primary task radar chart (aggregate by prefix)"""
if not selected_models:
return go.Figure()
# Collect all model task prefixes
prefix_order = []
# Map prefix -> [scores] for each model
model_prefix_scores = {}
for result in results:
display_name = get_display_name_for_result(result)
if display_name not in selected_models:
continue
primary_task = result.get('primary_task', {})
if display_name not in model_prefix_scores:
model_prefix_scores[display_name] = {}
for task_key, score in primary_task.items():
prefix = task_key.split('.')[0].strip() if isinstance(task_key, str) else str(task_key)
if prefix not in prefix_order:
prefix_order.append(prefix)
if prefix not in model_prefix_scores[display_name]:
model_prefix_scores[display_name][prefix] = []
model_prefix_scores[display_name][prefix].append(score * 100)
# Take first 11 prefixes
categories = prefix_order[:11]
# Create radar chart
fig = go.Figure()
for model_name in selected_models:
if model_name not in model_prefix_scores:
continue
# Mean aggregation for each prefix
values = []
for prefix in categories:
scores = model_prefix_scores[model_name].get(prefix, [])
if scores:
values.append(sum(scores) / len(scores))
else:
values.append(0)
# Close polygon
r_values = values + ([values[0]] if values else [])
theta_values = categories + ([categories[0]] if categories else [])
color_index = get_model_color_index(model_name, selected_models)
fig.add_trace(go.Scatterpolar(
r=r_values,
theta=theta_values,
mode='lines+markers',
name=model_name,
line=dict(color=get_color(color_index), width=3),
marker=dict(size=6),
fill='toself'
))
fig.update_layout(
title='Performance Comparison on Different Primary Tasks',
polar=dict(
radialaxis=dict(visible=True, range=[0, 100])
),
legend=dict(
orientation="h",
yanchor="top",
y=-0.2,
xanchor="center",
x=0.5
),
margin=dict(b=100)
)
return fig
def create_language_chart(results, selected_models):
"""Create language comparison bar chart"""
if not selected_models:
return go.Figure()
# Collect data
chart_data = {}
for result in results:
display_name = get_display_name_for_result(result)
if display_name in selected_models:
model_name = display_name
language = result['language']
# Store each model's result directly
if model_name not in chart_data:
chart_data[model_name] = {}
for lang_type, score in language.items():
chart_data[model_name][lang_type] = score * 100 # multiply by 100
# Create chart
fig = go.Figure()
# Get all language types
all_lang_types = []
for result in results:
display_name = get_display_name_for_result(result)
if display_name in selected_models:
language = result['language']
for lang_type in language.keys():
if lang_type not in all_lang_types:
all_lang_types.append(lang_type)
for model_name in selected_models:
if model_name in chart_data:
scores = [chart_data[model_name].get(lang_type, 0) for lang_type in all_lang_types]
color_index = get_model_color_index(model_name, selected_models)
fig.add_trace(go.Bar(
name=model_name,
x=all_lang_types,
y=scores,
marker_color=get_color(color_index),
text=[f"{score:.2f}" for score in scores], # keep 2 decimal places
textposition='auto'
))
fig.update_layout(
title='Performance Comparison on Different Languages',
xaxis_title='Language Type',
yaxis_title='Average Score',
barmode='group',
autosize=True, # auto size
legend=dict(
orientation="h",
yanchor="top",
y=-0.25, # adjust lower
xanchor="center",
x=0.5
),
margin=dict(b=100) # increase bottom margin
)
return fig
def create_difficulty_chart(results, selected_models):
"""Create difficulty comparison bar chart"""
if not selected_models:
return go.Figure()
# Collect data
chart_data = {}
for result in results:
display_name = get_display_name_for_result(result)
if display_name in selected_models:
model_name = display_name
difficulty = result['difficulty']
# Store each model's result directly
if model_name not in chart_data:
chart_data[model_name] = {}
for diff_type, score in difficulty.items():
chart_data[model_name][diff_type] = score * 100 # multiply by 100
# Create chart
fig = go.Figure()
# Get all difficulty types
all_diff_types = []
for result in results:
display_name = get_display_name_for_result(result)
if display_name in selected_models:
difficulty = result['difficulty']
for diff_type in difficulty.keys():
if diff_type not in all_diff_types:
all_diff_types.append(diff_type)
for model_name in selected_models:
if model_name in chart_data:
scores = [chart_data[model_name].get(diff_type, 0) for diff_type in all_diff_types]
color_index = get_model_color_index(model_name, selected_models)
fig.add_trace(go.Bar(
name=model_name,
x=all_diff_types,
y=scores,
marker_color=get_color(color_index),
text=[f"{score:.2f}" for score in scores], # keep 2 decimal places
textposition='auto'
))
fig.update_layout(
title='Performance Comparison on Different Difficulties',
xaxis_title='Difficulty Type',
yaxis_title='Average Score',
barmode='group',
autosize=True, # auto size
legend=dict(
orientation="h",
yanchor="top",
y=-0.25, # adjust lower
xanchor="center",
x=0.5
),
margin=dict(b=100) # increase bottom margin
)
return fig
def create_length_heatmap(results, selected_models):
"""Create length heatmap"""
if not selected_models:
return go.Figure()
# Standard context lengths
standard_lengths = [8000, 16000, 32000, 64000, 128000, 256000]
standard_length_keys = ['8k', '16k', '32k', '64k', '128k', '256k']
# Map results by name
result_map = {get_display_name_for_result(r): r for r in results}
# Prepare heatmap data
heatmap_data = []
model_names = []
for model_name in selected_models:
if model_name in result_map:
model_names.append(model_name)
result = result_map[model_name]
# Get data from token_length_metrics
token_length_metrics = result.get('token_length_metrics', {})
row_data = []
for key in standard_length_keys:
if key in token_length_metrics:
row_data.append(token_length_metrics[key] * 100) # multiply by 100
else:
row_data.append(None) # No data point
heatmap_data.append(row_data)
# Create heatmap
fig = go.Figure(data=go.Heatmap(
z=heatmap_data,
x=[f"{length//1000}k" for length in standard_lengths], # x axis labels
y=model_names, # y axis labels
colorscale='RdYlBu_r', # Red is low, Blue is high
showscale=True,
text=[[f"{val:.2f}" if val is not None else "N/A" for val in row] for row in heatmap_data], # show values
texttemplate="%{text}",
textfont={"size": 10},
hoverongaps=False
))
fig.update_layout(
title='Performance Heatmap on Different Sample Lengths',
xaxis_title='Sample Length (tokens)',
yaxis_title='Model Name',
autosize=True,
height=max(400, len(model_names) * 50), # adjust height based on model count
margin=dict(l=150, r=50, t=80, b=80) # adjust margins
)
return fig
def create_bon_chart(results, selected_models):
"""Create BoN 1-3 line chart"""
if not selected_models:
return go.Figure()
# BoN labels
bon_labels = ['BoN-1', 'BoN-2', 'BoN-3']
bon_indices = [1, 2, 3]
# Prepare data for each model
model_data = {}
for result in results:
display_name = get_display_name_for_result(result)
if display_name in selected_models:
if display_name not in model_data:
model_data[display_name] = {}
# Get data from bon_data
bon_data = result.get('bon_data', {})
for bon_key in bon_labels:
if bon_key in bon_data:
bon_index = bon_labels.index(bon_key) + 1
model_data[display_name][bon_index] = bon_data[bon_key] * 100 # multiply by 100
# Create chart
fig = go.Figure()
for model_name in selected_models:
if model_name not in model_data:
continue
data = model_data[model_name]
if not data:
continue
# Prepare data for each BoN
x_values = []
y_values = []
text_values = []
for bon_index in bon_indices:
x_values.append(bon_index)
if bon_index in data:
y_values.append(data[bon_index])
text_values.append(f"{data[bon_index]:.2f}")
else:
y_values.append(None)
text_values.append("")
# Get model color index
color_index = get_model_color_index(model_name, selected_models)
fig.add_trace(go.Scatter(
x=x_values,
y=y_values,
mode='lines+markers',
name=model_name,
line=dict(color=get_color(color_index), width=3),
marker=dict(size=10),
text=text_values,
textposition='top center',
connectgaps=False
))
# Set x axis
fig.update_layout(
title='Performance Comparison on Different Best-of-N',
xaxis_title='N',
yaxis_title='Average Score',
autosize=True,
xaxis=dict(
tickmode='array',
tickvals=bon_indices,
ticktext=bon_labels,
tickangle=0
),
legend=dict(
orientation="h",
yanchor="top",
y=-0.25,
xanchor="center",
x=0.5
),
margin=dict(b=100)
)
return fig
def create_pass_k_chart(results, selected_models):
"""Create Pass@N line chart"""
if not selected_models:
return go.Figure()
# Pass@K labels
k_labels = ['Pass@1', 'Pass@2', 'Pass@3']
k_indices = [1, 2, 3]
# Prepare data for each model
model_data = {}
for result in results:
display_name = get_display_name_for_result(result)
if display_name in selected_models:
if display_name not in model_data:
model_data[display_name] = {}
# Get data from pass_at_k
pass_data = result.get('pass_at_k', {})
for i, k_key in enumerate(k_labels):
val = pass_data.get(k_key)
if val is not None:
k_index = k_indices[i]
model_data[display_name][k_index] = val * 100 # multiply by 100
# Create chart
fig = go.Figure()
for model_name in selected_models:
if model_name not in model_data:
continue
data = model_data[model_name]
if not data:
continue
# Prepare data for each Pass@K
x_values = []
y_values = []
text_values = []
for k_index in k_indices:
x_values.append(k_index)
if k_index in data:
y_values.append(data[k_index])
text_values.append(f"{data[k_index]:.2f}")
else:
y_values.append(None)
text_values.append("")
# Get model color index
color_index = get_model_color_index(model_name, selected_models)
fig.add_trace(go.Scatter(
x=x_values,
y=y_values,
mode='lines+markers',
name=model_name,
line=dict(color=get_color(color_index), width=3),
marker=dict(size=10),
text=text_values,
textposition='top center',
connectgaps=False
))
# Set x axis
fig.update_layout(
title='Performance Comparison on Different Pass@N',
xaxis_title='N',
yaxis_title='Pass@N (%)',
autosize=True,
xaxis=dict(
tickmode='array',
tickvals=k_indices,
ticktext=k_labels,
tickangle=0
),
legend=dict(
orientation="h",
yanchor="top",
y=-0.25,
xanchor="center",
x=0.5
),
margin=dict(b=100)
)
return fig
def create_gradio_interface(parser: ResultParser):
"""Create Gradio interface"""
def refresh_data():
"""Refresh data"""
parser.scan_all_results()
return parser.get_leaderboard_data()
def get_model_choices():
"""Get model choices (distinguish by suffix for thinking/nonthinking)"""
if not parser.results:
return []
display_names = set()
for r in parser.results:
name = get_display_name_for_result(r)
display_names.add(name)
models = sorted(list(display_names))
return models
def update_charts(selected_models):
"""Update all charts"""
if not selected_models:
return None, None, None, None, None, None, None
length_heatmap = create_length_heatmap(parser.results, selected_models)
contextual_chart = create_contextual_requirement_chart(parser.results, selected_models)
primary_task_radar_chart = create_primary_task_radar_chart(parser.results, selected_models)
language_chart = create_language_chart(parser.results, selected_models)
difficulty_chart = create_difficulty_chart(parser.results, selected_models)
bon_chart = create_bon_chart(parser.results, selected_models)
pass_k_chart = create_pass_k_chart(parser.results, selected_models)
return length_heatmap, contextual_chart, primary_task_radar_chart, language_chart, difficulty_chart, bon_chart, pass_k_chart
# Create interface
with gr.Blocks(title="LongBench Pro Results Visualization", theme=gr.themes.Soft(), css=CUSTOM_CSS) as demo:
gr.Markdown("# LongBench Pro Results Visualization")
gr.HTML("""
<div style="text-align: center; display: flex; justify-content: center; gap: 10px; margin-bottom: 20px;">
<a href="https://huggingface.co/datasets/caskcsg/LongBench-Pro" target="_blank"><img src="https://img.shields.io/badge/Dataset-yellow?logo=huggingface&logoColor=yellow&labelColor=white" alt="Dataset"></a>
<a href="https://github.com/caskcsg/longcontext/tree/main/LongBench-Pro" target="_blank"><img src="https://img.shields.io/badge/Code-181717?logo=github&logoColor=181717&labelColor=white" alt="Code"></a>
<a href="#" target="_blank"><img src="https://img.shields.io/badge/Paper-red?logo=arxiv&logoColor=B31B1B&labelColor=white" alt="Paper"></a>
<a href="https://huggingface.co/spaces/caskcsg/LongBench-Pro-Leaderboard" target="_blank"><img src="https://img.shields.io/badge/π-Leaderboard-blue?labelColor=white" alt="Leaderboard"></a>
</div>
""")
# Leaderboard area
gr.Markdown("## π Overall Performance Leaderboard")
gr.Markdown("""
- *Thinking scores for Thinking and Mixed-Thinking models use their own thinking capabilities (Non-Thinking Prompt)*
- *Thinking scores for Instruct models are obtained using thinking prompts (Thinking Prompt)*
""")
leaderboard_df = gr.Dataframe(
headers=["Model Name", "Model Type", "Context Length", "Truncation Length", "Non-Thinking Score", "Thinking Score"],
datatype=["markdown", "str", "str", "str", "str", "str"],
interactive=False,
wrap=True,
show_row_numbers=True,
show_search="filter",
max_height=800,
column_widths=["250px", "100px", "100px", "100px", "120px", "120px"],
elem_id="leaderboard_table"
)
# Model selection and chart area
gr.HTML("<br>")
gr.Markdown("## π Specific Dimension Comparison")
with gr.Row():
with gr.Column(scale=4):
model_selector = gr.Dropdown(
choices=[],
label="Select Models",
value=[],
multiselect=True,
interactive=True
)
with gr.Column(scale=1):
update_charts_btn = gr.Button("Update Charts", variant="primary", size="lg")
with gr.Tabs():
with gr.TabItem("Language"):
language_plot = gr.Plot(show_label=False)
with gr.TabItem("Difficulty"):
difficulty_plot = gr.Plot(show_label=False)
with gr.TabItem("Sample Length"):
length_heatmap = gr.Plot(show_label=False)
with gr.TabItem("Primary Task"):
primary_task_radar_plot = gr.Plot(show_label=False)
with gr.TabItem("Context Requirement"):
contextual_plot = gr.Plot(show_label=False)
with gr.TabItem("Best-of-N"):
bon_plot = gr.Plot(show_label=False)
with gr.TabItem("Pass@N"):
pass_k_plot = gr.Plot(show_label=False)
# Add bottom spacer
gr.HTML("<div style='height: 100px;'></div>")
# Event handling
def update_model_choices():
models = get_model_choices()
return gr.Dropdown(choices=models, value=[])
update_charts_btn.click(
fn=update_charts,
inputs=[model_selector],
outputs=[length_heatmap, contextual_plot, primary_task_radar_plot, language_plot, difficulty_plot, bon_plot, pass_k_plot]
)
# Initialize
demo.load(
fn=refresh_data,
outputs=[leaderboard_df]
).then(
fn=update_model_choices,
outputs=[model_selector]
)
return demo
def main():
"""Main function"""
output_dir = "./output"
print("Initializing result parser...")
parser = ResultParser(output_dir)
print("Scanning result files...")
parser.scan_all_results()
print("Creating Gradio interface...")
demo = create_gradio_interface(parser)
print("Starting server...")
demo.launch()
if __name__ == "__main__":
main()
|