maplebb commited on
Commit
0893185
·
1 Parent(s): 9987298

Update Leaderboard

Browse files
app.py CHANGED
@@ -1,3 +1,4 @@
 
1
  import gradio as gr
2
  import pandas as pd
3
  from pathlib import Path
@@ -47,18 +48,26 @@ def extract_categories_and_subs(df):
47
  i += 1
48
  return category_dict
49
 
50
- # 列过滤函数,保持固定列 + 用户选择列 + 顺序不变
51
- def filtered_leaderboard(df, selected_columns):
52
- selected_columns = selected_columns or []
53
- final_cols = FIXED_COLUMNS + [col for col in df.columns if col in selected_columns and col not in FIXED_COLUMNS]
54
- return df[final_cols]
55
-
56
- # Update functions
57
- def update_leaderboard_overall(selected_cols, df_overall):
58
- return filtered_leaderboard(df_overall, selected_cols)
59
-
60
- def update_leaderboard_cat(selected_cols, df_cat):
61
- return filtered_leaderboard(df_cat, selected_cols)
 
 
 
 
 
 
 
 
62
 
63
  # 初始化
64
  df = get_json_df()
@@ -69,7 +78,7 @@ categories = extract_categories_and_subs(df)
69
  optional_columns = [col for col in df.columns if col not in FIXED_COLUMNS]
70
 
71
  # Gradio interface
72
- demo = gr.Blocks(css=custom_css, title="UniGenBench Leaderboard")
73
 
74
  with demo:
75
  gr.HTML(TITLE)
@@ -84,20 +93,21 @@ with demo:
84
  label="Select additional columns to display",
85
  value=optional_columns
86
  )
 
87
  leaderboard_table = gr.Dataframe(
88
- value=df[ALL_COLUMNS_ORDERED],
89
- headers=list(df.columns),
90
- datatype=["html" if col in ["Model Name (clickable)","HF Model"] else "str" for col in df.columns],
91
  interactive=False,
92
  wrap=False
93
  )
94
  selected_columns_overall.change(
95
- fn=update_leaderboard_overall,
96
- inputs=[selected_columns_overall, gr.State(value=df)],
97
  outputs=leaderboard_table
98
  )
99
 
100
- # 每个大类 leaderboard
101
  for cat_name, info in categories.items():
102
  with gr.TabItem(f"🏆 {cat_name}", elem_id=f"tab-{cat_name}"):
103
  cat_cols = [info["overall"]] + info["subs"]
@@ -117,7 +127,7 @@ with demo:
117
  wrap=False
118
  )
119
  selected_columns_cat.change(
120
- fn=update_leaderboard_cat,
121
  inputs=[selected_columns_cat, gr.State(value=cat_df)],
122
  outputs=leaderboard_table_cat
123
  )
@@ -126,7 +136,7 @@ with demo:
126
  with gr.Row():
127
  with gr.Column():
128
  gr.Markdown("## 📙 Citation")
129
- gr.Markdown("If you use [UniGenBench]() in your research, please cite our work:")
130
  citation_textbox = gr.Textbox(
131
  value=CITATION_BUTTON_TEXT,
132
  elem_id="citation-textbox",
 
1
+
2
  import gradio as gr
3
  import pandas as pd
4
  from pathlib import Path
 
48
  i += 1
49
  return category_dict
50
 
51
+ def build_overall_tab_cols(df, categories):
52
+ # Overall 页面希望把 Real/Game 的 '-Overall' 放在子项之后
53
+ real = categories.get("Real-World", {"overall": None, "subs": []})
54
+ game = categories.get("Game-World", {"overall": None, "subs": []})
55
+
56
+ base = FIXED_COLUMNS.copy()
57
+ rest = [c for c in df.columns if c not in base]
58
+
59
+ ordered = base + ["Overall"]
60
+ # real: 子项 -> overall
61
+ ordered += [c for c in real.get("subs", []) if c in df.columns]
62
+ if real.get("overall") in df.columns:
63
+ ordered += [real["overall"]]
64
+ # game: 子项 -> overall
65
+ ordered += [c for c in game.get("subs", []) if c in df.columns]
66
+ if game.get("overall") in df.columns:
67
+ ordered += [game["overall"]]
68
+ # 其它剩余列
69
+ ordered += [c for c in rest if c not in ordered]
70
+ return ordered
71
 
72
  # 初始化
73
  df = get_json_df()
 
78
  optional_columns = [col for col in df.columns if col not in FIXED_COLUMNS]
79
 
80
  # Gradio interface
81
+ demo = gr.Blocks(css=custom_css, title="UniREditBench Leaderboard")
82
 
83
  with demo:
84
  gr.HTML(TITLE)
 
93
  label="Select additional columns to display",
94
  value=optional_columns
95
  )
96
+ overall_cols = build_overall_tab_cols(df, categories)
97
  leaderboard_table = gr.Dataframe(
98
+ value=df[overall_cols],
99
+ headers=list(df[overall_cols].columns),
100
+ datatype=["html" if col in ["Model Name (clickable)","HF Model"] else "str" for col in df[overall_cols].columns],
101
  interactive=False,
102
  wrap=False
103
  )
104
  selected_columns_overall.change(
105
+ fn=lambda selected, d: d[[*FIXED_COLUMNS, *[c for c in d.columns if c in selected and c not in FIXED_COLUMNS]]],
106
+ inputs=[selected_columns_overall, gr.State(value=df[overall_cols])],
107
  outputs=leaderboard_table
108
  )
109
 
110
+ # 每个大类 leaderboard(保持 overall 在前,便于分组)
111
  for cat_name, info in categories.items():
112
  with gr.TabItem(f"🏆 {cat_name}", elem_id=f"tab-{cat_name}"):
113
  cat_cols = [info["overall"]] + info["subs"]
 
127
  wrap=False
128
  )
129
  selected_columns_cat.change(
130
+ fn=lambda selected, d: d[[*FIXED_COLUMNS, *[c for c in d.columns if c in selected and c not in FIXED_COLUMNS]]],
131
  inputs=[selected_columns_cat, gr.State(value=cat_df)],
132
  outputs=leaderboard_table_cat
133
  )
 
136
  with gr.Row():
137
  with gr.Column():
138
  gr.Markdown("## 📙 Citation")
139
+ gr.Markdown("If you use UniREditBench in your research, please cite our work:")
140
  citation_textbox = gr.Textbox(
141
  value=CITATION_BUTTON_TEXT,
142
  elem_id="citation-textbox",
leaderboard_data.json CHANGED
@@ -1,2438 +1,312 @@
1
  {
2
  "leaderboard": [
3
  {
4
- "model": "wan2.5-t2i-preview",
5
- "link": "https://help.aliyun.com/zh/model-studio/text-to-image-v2-api-reference",
6
  "hf": "-",
7
  "open_source": false,
8
- "release_date": "2025-09",
9
-
10
- "Overall": 78.17,
11
- "Style": 93.15,
12
- "World Knowledge": 95.22,
13
-
14
- "Attribute-Overall": 81.06,
15
- "Quantity": 75.00,
16
- "Expression": 67.95,
17
- "Material": 91.04,
18
- "Size": 85.29,
19
- "Shape": 77.50,
20
- "Color": 87.50,
21
-
22
- "Action-Overall": 74.23,
23
- "Hand": 61.18,
24
- "Full body": 75.00,
25
- "Animal": 76.47,
26
- "Non Contact": 75.00,
27
- "Contact": 72.02,
28
- "State": 82.55,
29
-
30
- "Relationship-Overall": 82.23,
31
- "Composition": 85.14,
32
- "Similarity": 75.00,
33
- "Inclusion": 82.07,
34
- "Comparison": 85.94,
35
-
36
- "Compound-Overall": 76.23,
37
- "Imagination": 79.38,
38
- "Feature matching": 73.04,
39
-
40
- "Grammar-Overall": 73.59,
41
- "Pronoun Reference": 84.07,
42
- "Consistency": 73.15,
43
- "Negation": 63.08,
44
-
45
- "Layout-Overall": 77.61,
46
- "2D": 75.74,
47
- "3D": 79.55,
48
-
49
- "Logical Reasoning": 56.36,
50
-
51
- "Text": 71.97
52
- },
53
- {
54
- "model": "Echo-4o",
55
- "link": "https://arxiv.org/pdf/2508.09987",
56
- "hf": "https://huggingface.co/Yejy53/Echo-4o",
57
- "open_source": true,
58
- "release_date": "2025-8",
59
-
60
- "Overall": 69.12,
61
- "Style": 92.20,
62
- "World Knowledge": 90.51,
63
-
64
- "Attribute-Overall": 79.06,
65
- "Quantity": 70.14,
66
- "Expression": 71.15,
67
- "Material": 84.91,
68
- "Size": 83.33,
69
- "Shape": 68.75,
70
- "Color": 98.33,
71
-
72
- "Action-Overall": 68.92,
73
- "Hand": 66.03,
74
- "Full body": 66.30,
75
- "Animal": 77.94,
76
- "Non Contact": 67.86,
77
- "Contact": 59.52,
78
- "State": 75.94,
79
-
80
- "Relationship-Overall": 76.52,
81
- "Composition": 81.76,
82
- "Similarity": 70.56,
83
- "Inclusion": 77.72,
84
- "Comparison": 71.09,
85
-
86
- "Compound-Overall": 71.78,
87
- "Imagination": 76.79,
88
- "Feature matching": 66.67,
89
-
90
- "Grammar-Overall": 75.13,
91
- "Pronoun Reference": 80.51,
92
- "Consistency": 74.54,
93
- "Negation": 70.00,
94
-
95
- "Layout-Overall": 82.28,
96
- "2D": 87.13,
97
- "3D": 77.27,
98
-
99
- "Logical Reasoning": 44.77,
100
-
101
- "Text": 10.06
102
- },
103
- {
104
- "model": "UniWorld-V1",
105
- "link": "https://arxiv.org/pdf/2506.03147",
106
- "hf": "https://huggingface.co/LanguageBind/UniWorld-V1",
107
- "open_source": true,
108
- "release_date": "2025-06",
109
-
110
- "Overall": 63.11,
111
- "Style": 91.10,
112
- "World Knowledge": 82.91,
113
-
114
- "Attribute-Overall": 70.62,
115
- "Quantity": 70.14,
116
- "Expression": 64.74,
117
- "Material": 61.32,
118
- "Size": 72.22,
119
- "Shape": 66.25,
120
- "Color": 99.17,
121
-
122
- "Action-Overall": 67.21,
123
- "Hand": 55.13,
124
- "Full body": 72.28,
125
- "Animal": 73.53,
126
- "Non Contact": 63.78,
127
- "Contact": 61.90,
128
- "State": 75.00,
129
-
130
- "Relationship-Overall": 67.13,
131
- "Composition": 72.30,
132
- "Similarity": 63.33,
133
- "Inclusion": 64.67,
134
- "Comparison": 64.06,
135
-
136
- "Compound-Overall": 54.51,
137
- "Imagination": 58.16,
138
- "Feature matching": 50.78,
139
-
140
- "Grammar-Overall": 63.77,
141
- "Pronoun Reference": 74.26,
142
- "Consistency": 64.35,
143
- "Negation": 52.31,
144
-
145
- "Layout-Overall": 69.03,
146
- "2D": 73.90,
147
- "3D": 64.02,
148
-
149
- "Logical Reasoning": 38.41,
150
-
151
- "Text": 26.44
152
- },
153
- {
154
- "model": "SD-3.5-Medium",
155
- "link": "https://stability.ai/news/introducing-stable-diffusion-3-5",
156
- "hf": "stabilityai/stable-diffusion-3.5-medium",
157
- "open_source": true,
158
- "release_date": "2024-10",
159
-
160
- "Overall": 60.71,
161
- "Style": 89.80,
162
- "World Knowledge": 84.34,
163
-
164
- "Attribute-Overall": 66.99,
165
- "Quantity": 59.72,
166
- "Expression": 51.92,
167
- "Material": 67.92,
168
- "Size": 70.83,
169
- "Shape": 63.75,
170
- "Color": 93.33,
171
-
172
- "Action-Overall": 60.65,
173
- "Hand": 50.00,
174
- "Full body": 63.04,
175
- "Animal": 69.12,
176
- "Non Contact": 55.61,
177
- "Contact": 52.98,
178
- "State": 71.70,
179
-
180
- "Relationship-Overall": 68.78,
181
- "Composition": 74.66,
182
- "Similarity": 61.67,
183
- "Inclusion": 73.37,
184
- "Comparison": 58.59,
185
-
186
- "Compound-Overall": 53.35,
187
- "Imagination": 58.16,
188
- "Feature matching": 48.44,
189
-
190
- "Grammar-Overall": 59.89,
191
- "Pronoun Reference": 73.53,
192
- "Consistency": 61.57,
193
- "Negation": 44.23,
194
-
195
- "Layout-Overall": 70.34,
196
- "2D": 72.06,
197
- "3D": 68.56,
198
-
199
- "Logical Reasoning": 37.73,
200
-
201
- "Text": 15.23
202
- },
203
- {
204
- "model": "Lumina-DiMOO",
205
- "link": "https://synbol.github.io/Lumina-DiMOO/",
206
- "hf": "https://huggingface.co/Alpha-VLLM/Lumina-DiMOO",
207
- "open_source": true,
208
- "release_date": "2025-09",
209
-
210
- "Overall": 71.12,
211
- "Style": 89.70,
212
- "World Knowledge": 90.03,
213
-
214
- "Attribute-Overall": 81.62,
215
- "Quantity": 69.44,
216
- "Expression": 85.90,
217
- "Material": 81.60,
218
- "Size": 76.39,
219
- "Shape": 80.00,
220
- "Color": 99.17,
221
-
222
- "Action-Overall": 73.76,
223
- "Hand": 64.10,
224
- "Full body": 78.80,
225
- "Animal": 75.74,
226
- "Non Contact": 73.98,
227
- "Contact": 64.88,
228
- "State": 82.08,
229
-
230
- "Relationship-Overall": 78.43,
231
- "Composition": 83.45,
232
- "Similarity": 74.44,
233
- "Inclusion": 81.52,
234
- "Comparison": 67.97,
235
-
236
- "Compound-Overall": 73.32,
237
- "Imagination": 78.83,
238
- "Feature matching": 67.71,
239
-
240
- "Grammar-Overall": 70.45,
241
- "Pronoun Reference": 81.99,
242
- "Consistency": 77.78,
243
- "Negation": 52.31,
244
-
245
- "Layout-Overall": 82.84,
246
- "2D": 84.93,
247
- "3D": 80.68,
248
-
249
- "Logical Reasoning": 45.45,
250
-
251
- "Text": 25.57
252
- },
253
- {
254
- "model": "MMaDA",
255
- "link": "https://arxiv.org/pdf/2505.15809",
256
- "hf": "https://huggingface.co/Gen-Verse/MMaDA-8B-MixCoT",
257
- "open_source": true,
258
  "release_date": "2025-05",
259
-
260
- "Overall": 41.35,
261
- "Style": 82.40,
262
- "World Knowledge": 56.65,
263
-
264
- "Attribute-Overall": 48.93,
265
- "Quantity": 45.83,
266
- "Expression": 29.49,
267
- "Material": 54.25,
268
- "Size": 49.31,
269
- "Shape": 44.38,
270
- "Color": 74.17,
271
-
272
- "Action-Overall": 37.83,
273
- "Hand": 15.38,
274
- "Full body": 40.22,
275
- "Animal": 52.94,
276
- "Non Contact": 33.16,
277
- "Contact": 25.60,
278
- "State": 56.60,
279
-
280
- "Relationship-Overall": 50.25,
281
- "Composition": 55.07,
282
- "Similarity": 57.22,
283
- "Inclusion": 47.28,
284
- "Comparison": 33.59,
285
-
286
- "Compound-Overall": 32.35,
287
- "Imagination": 40.56,
288
- "Feature matching": 23.96,
289
-
290
- "Grammar-Overall": 55.75,
291
- "Pronoun Reference": 59.19,
292
- "Consistency": 40.28,
293
- "Negation": 65.00,
294
-
295
- "Layout-Overall": 30.22,
296
- "2D": 30.15,
297
- "3D": 30.30,
298
-
299
- "Logical Reasoning": 17.95,
300
-
301
- "Text": 1.15
302
- },
303
- {
304
- "model": "OmniGen2",
305
- "link": "https://arxiv.org/pdf/2506.18871",
306
- "hf": "https://huggingface.co/OmniGen2/OmniGen2",
307
- "open_source": true,
308
- "release_date": "2025-06",
309
-
310
- "Overall": 63.09,
311
- "Style": 91.90,
312
- "World Knowledge": 86.39,
313
-
314
- "Attribute-Overall": 72.12,
315
- "Quantity": 67.36,
316
- "Expression": 73.08,
317
- "Material": 66.04,
318
- "Size": 72.22,
319
- "Shape": 66.25,
320
- "Color": 95.00,
321
-
322
- "Action-Overall": 62.83,
323
- "Hand": 55.77,
324
- "Full body": 69.02,
325
- "Animal": 68.38,
326
- "Non Contact": 62.24,
327
- "Contact": 54.17,
328
- "State": 66.51,
329
-
330
- "Relationship-Overall": 68.27,
331
- "Composition": 68.24,
332
- "Similarity": 67.78,
333
- "Inclusion": 71.20,
334
- "Comparison": 64.84,
335
-
336
- "Compound-Overall": 56.31,
337
- "Imagination": 62.24,
338
- "Feature matching": 50.26,
339
-
340
- "Grammar-Overall": 59.89,
341
- "Pronoun Reference": 71.32,
342
- "Consistency": 60.65,
343
- "Negation": 47.31,
344
-
345
- "Layout-Overall": 71.64,
346
- "2D": 78.31,
347
- "3D": 64.77,
348
-
349
- "Logical Reasoning": 32.50,
350
-
351
- "Text": 29.02
352
- },
353
- {
354
- "model": "Infinity",
355
- "link": "https://arxiv.org/pdf/2412.04431",
356
- "hf": "https://huggingface.co/FoundationVision/Infinity/tree/main/infinity_8b_512x512_weights",
357
- "open_source": true,
358
- "release_date": "2024-12",
359
-
360
- "Overall": 59.81,
361
- "Style": 90.80,
362
- "World Knowledge": 87.97,
363
- "Attribute-Overall": 68.06,
364
- "Quantity": 66.67,
365
- "Expression": 53.21,
366
- "Material": 66.04,
367
- "Size": 77.78,
368
- "Shape": 58.75,
369
- "Color": 93.33,
370
-
371
- "Action-Overall": 60.17,
372
- "Hand": 55.13,
373
- "Full body": 65.22,
374
- "Animal": 72.06,
375
- "Non Contact": 58.16,
376
- "Contact": 49.40,
377
- "State": 62.26,
378
-
379
- "Relationship-Overall": 69.16,
380
- "Composition": 73.31,
381
- "Similarity": 65.00,
382
- "Inclusion": 67.39,
383
- "Comparison": 67.97,
384
-
385
- "Compound-Overall": 51.42,
386
- "Imagination": 55.87,
387
- "Feature matching": 46.88,
388
-
389
- "Grammar-Overall": 60.16,
390
- "Pronoun Reference": 73.16,
391
- "Consistency": 65.74,
392
- "Negation": 41.92,
393
-
394
- "Layout-Overall": 66.60,
395
- "2D": 71.69,
396
- "3D": 61.36,
397
-
398
- "Logical Reasoning": 31.36,
399
-
400
- "Text": 12.36
401
- },
402
- {
403
- "model": "OneCAT",
404
- "link": "https://arxiv.org/pdf/2509.03498",
405
- "hf": "https://huggingface.co/onecat-ai/OneCAT-3B",
406
- "open_source": true,
407
- "release_date": "2025-09",
408
-
409
- "Overall": 58.28,
410
- "Style": 93.30,
411
- "World Knowledge": 82.28,
412
-
413
- "Attribute-Overall": 63.46,
414
- "Quantity": 59.42,
415
- "Expression": 58.33,
416
- "Material": 67.45,
417
- "Size": 65.97,
418
- "Shape": 42.50,
419
- "Color": 92.50,
420
-
421
- "Action-Overall": 58.56,
422
- "Hand": 35.90,
423
- "Full body": 65.22,
424
- "Animal": 69.12,
425
- "Non Contact": 57.65,
426
- "Contact": 48.81,
427
- "State": 71.23,
428
-
429
- "Relationship-Overall": 68.15,
430
- "Composition": 78.04,
431
- "Similarity": 69.44,
432
- "Inclusion": 62.50,
433
- "Comparison": 51.56,
434
-
435
- "Compound-Overall": 56.96,
436
- "Imagination": 66.33,
437
- "Feature matching": 47.40,
438
-
439
- "Grammar-Overall": 60.83,
440
- "Pronoun Reference": 70.59,
441
- "Consistency": 59.72,
442
- "Negation": 51.54,
443
-
444
- "Layout-Overall": 64.74,
445
- "2D": 64.34,
446
- "3D": 65.15,
447
-
448
- "Logical Reasoning": 33.41,
449
-
450
- "Text": 1.15
451
- },
452
- {
453
- "model": "X-Omni",
454
- "link": "https://arxiv.org/pdf/2507.22058",
455
- "hf": "https://huggingface.co/X-Omni/X-Omni-En",
456
- "open_source": true,
457
- "release_date": "2025-08",
458
-
459
- "Overall": 53.77,
460
- "Style": 72.70,
461
- "World Knowledge": 76.27,
462
-
463
- "Attribute-Overall": 60.04,
464
- "Quantity": 63.19,
465
- "Expression": 53.21,
466
- "Material": 58.96,
467
- "Size": 55.56,
468
- "Shape": 53.75,
469
- "Color": 80.83,
470
-
471
- "Action-Overall": 54.47,
472
- "Hand": 46.79,
473
- "Full body": 56.52,
474
- "Animal": 62.50,
475
- "Non Contact": 56.63,
476
- "Contact": 42.26,
477
- "State": 60.85,
478
-
479
- "Relationship-Overall": 56.60,
480
- "Composition": 61.82,
481
- "Similarity": 56.11,
482
- "Inclusion": 51.09,
483
- "Comparison": 53.12,
484
-
485
- "Compound-Overall": 41.75,
486
- "Imagination": 47.45,
487
- "Feature matching": 35.94,
488
-
489
- "Grammar-Overall": 59.09,
490
- "Pronoun Reference": 66.91,
491
- "Consistency": 54.17,
492
- "Negation": 55.00,
493
-
494
- "Layout-Overall": 62.69,
495
- "2D": 69.49,
496
- "3D": 55.68,
497
-
498
- "Logical Reasoning": 29.09,
499
-
500
- "Text": 25.00
501
- },
502
- {
503
- "model": "FLUX.1-Krea-dev",
504
- "link": "https://www.krea.ai/blog/flux-krea-open-source-release",
505
- "hf": "https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev",
506
- "open_source": true,
507
- "release_date": "2025-08",
508
-
509
- "Overall": 69.88,
510
- "Style": 88.70,
511
- "World Knowledge": 92.56,
512
-
513
- "Attribute-Overall": 75.96,
514
- "Quantity": 70.83,
515
- "Expression": 60.90,
516
- "Material": 77.36,
517
- "Size": 79.17,
518
- "Shape": 73.12,
519
- "Color": 99.17,
520
-
521
- "Action-Overall": 71.01,
522
- "Hand": 64.74,
523
- "Full body": 70.11,
524
- "Animal": 77.94,
525
- "Non Contact": 72.96,
526
- "Contact": 67.26,
527
- "State": 73.11,
528
-
529
- "Relationship-Overall": 73.98,
530
- "Composition": 76.35,
531
- "Similarity": 66.11,
532
- "Inclusion": 77.17,
533
- "Comparison": 75.00,
534
-
535
- "Compound-Overall": 64.43,
536
- "Imagination": 67.35,
537
- "Feature matching": 61.46,
538
-
539
- "Grammar-Overall": 63.37,
540
- "Pronoun Reference": 77.21,
541
- "Consistency": 67.13,
542
- "Negation": 45.77,
543
-
544
- "Layout-Overall": 84.14,
545
- "2D": 86.76,
546
- "3D": 81.44,
547
-
548
- "Logical Reasoning": 39.77,
549
-
550
- "Text": 44.83
551
- },
552
- {
553
- "model": "Hunyuan-Image-2.1",
554
- "link": "https://github.com/Tencent-Hunyuan/HunyuanImage-2.1",
555
- "hf": "https://huggingface.co/spaces/tencent/HunyuanImage-2.1",
556
- "open_source": true,
557
- "release_date": "2025-09",
558
-
559
- "Overall": 74.64,
560
- "Style": 90.88,
561
- "World Knowledge": 92.06,
562
- "Attribute-Overall": 79.66,
563
- "Quantity": 86.62,
564
- "Expression": 72.44,
565
- "Material": 78.77,
566
- "Size": 78.47,
567
- "Shape": 68.12,
568
- "Color": 99.17,
569
-
570
- "Action-Overall": 77.81,
571
- "Hand": 75.00,
572
- "Full body": 80.98,
573
- "Animal": 82.35,
574
- "Non Contact": 73.71,
575
- "Contact": 72.02,
576
- "State": 82.55,
577
-
578
- "Relationship-Overall": 77.54,
579
- "Composition": 78.38,
580
- "Similarity": 70.56,
581
- "Inclusion": 84.78,
582
- "Comparison": 75.00,
583
-
584
- "Compound-Overall": 64.82,
585
- "Imagination": 64.54,
586
- "Feature matching": 65.10,
587
-
588
- "Grammar-Overall": 62.83,
589
- "Pronoun Reference": 77.94,
590
- "Consistency": 66.20,
591
- "Negation": 44.23,
592
-
593
- "Layout-Overall": 84.14,
594
- "2D": 86.76,
595
- "3D": 81.44,
596
-
597
- "Logical Reasoning": 46.59,
598
-
599
- "Text": 70.11
600
- },
601
- {
602
- "model": "BLIP3-o-Next",
603
- "link": "https://arxiv.org/pdf/2505.09568",
604
- "hf": "https://huggingface.co/BLIP3o/BLIP3o-NEXT-SFT-3B",
605
- "open_source": true,
606
- "release_date": "2025-08",
607
-
608
- "Overall": 65.15,
609
- "Style": 91.00,
610
- "World Knowledge": 86.71,
611
- "Attribute-Overall": 70.94,
612
- "Quantity": 67.36,
613
- "Expression": 73.72,
614
- "Material": 70.28,
615
- "Size": 76.39,
616
- "Shape": 60.62,
617
- "Color": 80.00,
618
-
619
- "Action-Overall": 66.83,
620
- "Hand": 57.69,
621
- "Full body": 75.00,
622
- "Animal": 73.53,
623
- "Non Contact": 67.35,
624
- "Contact": 57.74,
625
- "State": 68.87,
626
-
627
- "Relationship-Overall": 73.60,
628
- "Composition": 76.01,
629
- "Similarity": 65.00,
630
- "Inclusion": 77.17,
631
- "Comparison": 75.00,
632
-
633
- "Compound-Overall": 64.82,
634
- "Imagination": 73.72,
635
- "Feature matching": 55.73,
636
-
637
- "Grammar-Overall": 68.05,
638
- "Pronoun Reference": 76.47,
639
- "Consistency": 67.13,
640
- "Negation": 60.00,
641
-
642
- "Layout-Overall": 76.31,
643
- "2D": 80.15,
644
- "3D": 72.35,
645
-
646
- "Logical Reasoning": 48.64,
647
-
648
- "Text": 4.60
649
- },
650
- {
651
- "model": "Kolors",
652
- "link": "https://github.com/Kwai-Kolors/Kolors/blob/master/imgs/Kolors_paper.pdf",
653
- "hf": "https://huggingface.co/Kwai-Kolors/Kolors",
654
- "open_source": true,
655
- "release_date": "2024-7",
656
-
657
- "Overall": 45.47,
658
- "Style": 84.40,
659
- "World Knowledge": 77.22,
660
-
661
- "Attribute-Overall": 54.17,
662
- "Quantity": 62.50,
663
- "Expression": 33.33,
664
- "Material": 51.89,
665
- "Size": 62.50,
666
- "Shape": 40.62,
667
- "Color": 83.33,
668
-
669
- "Action-Overall": 48.00,
670
- "Hand": 42.95,
671
- "Full body": 42.39,
672
- "Animal": 56.62,
673
- "Non Contact": 45.92,
674
- "Contact": 39.88,
675
- "State": 59.43,
676
-
677
- "Relationship-Overall": 52.79,
678
- "Composition": 55.41,
679
- "Similarity": 53.89,
680
- "Inclusion": 51.63,
681
- "Comparison": 46.88,
682
-
683
- "Compound-Overall": 33.63,
684
- "Imagination": 41.33,
685
- "Feature matching": 25.78,
686
-
687
- "Grammar-Overall": 46.66,
688
- "Pronoun Reference": 56.62,
689
- "Consistency": 47.22,
690
- "Negation": 35.77,
691
-
692
- "Layout-Overall": 42.91,
693
- "2D": 43.01,
694
- "3D": 42.80,
695
-
696
- "Logical Reasoning": 19.77,
697
-
698
- "Text": 1.15
699
- },
700
- {
701
- "model": "Seedream-4.0",
702
- "link": "https://www.volcengine.com/docs/82379/1541523",
703
  "hf": "-",
704
  "open_source": false,
705
  "release_date": "2025-09",
706
-
707
- "Overall": 87.35,
708
-
709
- "Style": 98.80,
710
-
711
- "World Knowledge": 95.41,
712
-
713
- "Attribute-Overall": 88.57,
714
- "Quantity": 86.81,
715
- "Expression": 85.90,
716
- "Material": 97.17,
717
- "Size": 84.03,
718
- "Shape": 76.88,
719
- "Color": 100.0,
720
-
721
- "Action-Overall": 85.65,
722
- "Hand": 77.56,
723
- "Full body": 87.50,
724
- "Animal": 88.24,
725
- "Non Contact": 80.10,
726
- "Contact": 83.93,
727
- "State": 94.81,
728
-
729
- "Relationship-Overall": 87.69,
730
- "Composition": 88.18,
731
- "Similarity": 80.56,
732
- "Inclusion": 94.02,
733
- "Comparison": 87.50,
734
-
735
- "Compound-Overall": 86.08,
736
- "Imagination": 88.27,
737
- "Feature matching": 83.85,
738
-
739
- "Grammar-Overall": 78.88,
740
- "Pronoun Reference": 84.93,
741
- "Consistency": 79.17,
742
- "Negation": 72.31,
743
-
744
- "Layout-Overall": 90.67,
745
- "2D": 90.81,
746
- "3D": 90.53,
747
-
748
- "Logical Reasoning": 67.73,
749
-
750
- "Text": 93.97
751
- },
752
- {
753
- "model": "Imagen-4.0-generate-preview-06-06",
754
- "link": "https://deepmind.google/models/imagen/",
755
  "hf": "-",
756
  "open_source": false,
757
- "release_date": "2025-01",
758
-
759
- "Overall": 85.84,
760
-
761
- "Style": 97.80,
762
-
763
- "World Knowledge": 96.36,
764
-
765
- "Attribute-Overall": 84.94,
766
- "Quantity": 84.03,
767
- "Expression": 76.92,
768
- "Material": 90.57,
769
- "Size": 89.58,
770
- "Shape": 71.88,
771
- "Color": 98.33,
772
-
773
- "Action-Overall": 88.40,
774
- "Hand": 86.54,
775
- "Full body": 94.02,
776
- "Animal": 88.97,
777
- "Non Contact": 85.71,
778
- "Contact": 83.33,
779
- "State": 91.04,
780
-
781
- "Relationship-Overall": 89.34,
782
- "Composition": 93.58,
783
- "Similarity": 78.89,
784
- "Inclusion": 95.11,
785
- "Comparison": 85.94,
786
-
787
- "Compound-Overall": 85.31,
788
- "Imagination": 90.31,
789
- "Feature matching": 80.21,
790
-
791
- "Grammar-Overall": 79.68,
792
- "Pronoun Reference": 86.76,
793
- "Consistency": 77.31,
794
- "Negation": 74.23,
795
-
796
- "Layout-Overall": 88.81,
797
- "2D": 88.24,
798
- "3D": 89.39,
799
-
800
- "Logical Reasoning": 70.45,
801
-
802
- "Text": 77.30
803
- },
804
- {
805
- "model": "Runway-Gen4-Image",
806
- "link": "https://docs.dev.runwayml.com/api/#tag/Start-generating/paths/~1v1~1text_to_image/post",
807
- "hf": "-",
808
- "open_source": false,
809
- "release_date": "2024-11",
810
-
811
- "Overall": 69.75,
812
-
813
- "Style": 93.44,
814
-
815
- "World Knowledge": 90.36,
816
-
817
- "Attribute-Overall": 74.03,
818
- "Quantity": 72.86,
819
- "Expression": 51.97,
820
- "Material": 89.42,
821
- "Size": 68.06,
822
- "Shape": 65.62,
823
- "Color": 95.00,
824
-
825
- "Action-Overall": 70.21,
826
- "Hand": 62.18,
827
- "Full body": 79.35,
828
- "Animal": 82.35,
829
- "Non Contact": 66.15,
830
- "Contact": 60.37,
831
- "State": 71.70,
832
-
833
- "Relationship-Overall": 72.56,
834
- "Composition": 74.32,
835
- "Similarity": 62.22,
836
- "Inclusion": 77.84,
837
- "Comparison": 75.78,
838
-
839
- "Compound-Overall": 67.76,
840
- "Imagination": 71.65,
841
- "Feature matching": 63.71,
842
-
843
- "Grammar-Overall": 70.08,
844
- "Pronoun Reference": 71.21,
845
- "Consistency": 67.59,
846
- "Negation": 71.03,
847
-
848
- "Layout-Overall": 76.33,
849
- "2D": 77.61,
850
- "3D": 75.00,
851
-
852
- "Logical Reasoning": 49.31,
853
-
854
- "Text": 33.43
855
  },
856
- {
857
  "model": "Nano Banana",
858
- "link": "https://ainanobanana.io/",
859
  "hf": "-",
860
  "open_source": false,
861
  "release_date": "2025-08",
862
-
863
- "Overall": 87.45,
864
-
865
- "Style": 98.87,
866
-
867
- "World Knowledge": 96.32,
868
-
869
- "Attribute-Overall": 87.84,
870
- "Quantity": 85.00,
871
- "Expression": 83.33,
872
- "Material": 88.50,
873
- "Size": 95.74,
874
- "Shape": 78.21,
875
- "Color": 99.17,
876
-
877
- "Action-Overall": 86.83,
878
- "Hand": 82.05,
879
- "Full body": 93.41,
880
- "Animal": 86.03,
881
- "Non Contact": 82.47,
882
- "Contact": 83.33,
883
- "State": 91.98,
884
-
885
- "Relationship-Overall": 92.00,
886
- "Composition": 94.76,
887
- "Similarity": 86.52,
888
- "Inclusion": 91.26,
889
- "Comparison": 94.53,
890
-
891
- "Compound-Overall": 87.83,
892
- "Imagination": 89.66,
893
- "Feature matching": 86.02,
894
-
895
- "Grammar-Overall": 83.36,
896
- "Pronoun Reference": 90.71,
897
- "Consistency": 82.08,
898
- "Negation": 76.59,
899
-
900
- "Layout-Overall": 91.96,
901
- "2D": 92.65,
902
- "3D": 91.25,
903
-
904
- "Logical Reasoning": 74.26,
905
-
906
- "Text": 75.22
907
- },
908
- {
909
- "model": "Stability-AI-stable-image-ultra",
910
- "link": "https://platform.stability.ai/docs/api-reference#tag/Generate/paths/~1v2beta~1stable-image~1generate~1ultra/post",
911
- "hf": "-",
912
- "open_source": false,
913
- "release_date": "2024-06",
914
-
915
- "Overall": 61.96,
916
-
917
- "Style": 87.20,
918
-
919
- "World Knowledge": 87.18,
920
-
921
- "Attribute-Overall": 66.35,
922
- "Quantity": 67.36,
923
- "Expression": 48.08,
924
- "Material": 64.15,
925
- "Size": 69.44,
926
- "Shape": 64.38,
927
- "Color": 91.67,
928
-
929
- "Action-Overall": 59.22,
930
- "Hand": 55.77,
931
- "Full body": 58.15,
932
- "Animal": 63.24,
933
- "Non Contact": 61.22,
934
- "Contact": 51.79,
935
- "State": 64.15,
936
-
937
- "Relationship-Overall": 69.04,
938
- "Composition": 72.64,
939
- "Similarity": 66.67,
940
- "Inclusion": 70.11,
941
- "Comparison": 62.50,
942
-
943
- "Compound-Overall": 54.25,
944
- "Imagination": 60.97,
945
- "Feature matching": 47.40,
946
-
947
- "Grammar-Overall": 61.10,
948
- "Pronoun Reference": 78.68,
949
- "Consistency": 58.33,
950
- "Negation": 45.00,
951
-
952
- "Layout-Overall": 64.55,
953
- "2D": 67.28,
954
- "3D": 61.74,
955
-
956
- "Logical Reasoning": 31.59,
957
-
958
- "Text": 39.08
959
  },
960
  {
961
- "model": "HiDream_v2L",
962
- "link": "https://hidreamai.com/doc/txt2img/request",
963
- "hf": "-",
964
- "open_source": false,
965
- "release_date": "2025-07",
966
-
967
- "Overall": 61.64,
968
-
969
- "Style": 87.99,
970
-
971
- "World Knowledge": 89.62,
972
-
973
- "Attribute-Overall": 64.38,
974
- "Quantity": 65.71,
975
- "Expression": 44.87,
976
- "Material": 57.82,
977
- "Size": 74.26,
978
- "Shape": 59.87,
979
- "Color": 94.92,
980
-
981
- "Action-Overall": 59.50,
982
- "Hand": 51.28,
983
- "Full body": 58.56,
984
- "Animal": 67.65,
985
- "Non Contact": 61.98,
986
- "Contact": 51.52,
987
- "State": 65.09,
988
-
989
- "Relationship-Overall": 66.62,
990
- "Composition": 71.23,
991
- "Similarity": 64.20,
992
- "Inclusion": 65.93,
993
- "Comparison": 60.32,
994
-
995
- "Compound-Overall": 49.28,
996
- "Imagination": 53.75,
997
- "Feature matching": 44.76,
998
-
999
- "Grammar-Overall": 58.86,
1000
- "Pronoun Reference": 72.35,
1001
- "Consistency": 60.00,
1002
- "Negation": 44.23,
1003
-
1004
- "Layout-Overall": 69.06,
1005
- "2D": 70.41,
1006
- "3D": 67.68,
1007
-
1008
- "Logical Reasoning": 26.73,
1009
-
1010
- "Text": 44.31
1011
- },
1012
- {
1013
- "model": "Imagen-4.0-Fast-preview-06-06",
1014
- "link": "https://deepmind.google/models/imagen/",
1015
- "hf": "-",
1016
- "open_source": false,
1017
- "release_date": "2025-01",
1018
-
1019
- "Overall": 77.75,
1020
-
1021
- "Style": 92.00,
1022
-
1023
- "World Knowledge": 94.78,
1024
-
1025
- "Attribute-Overall": 83.65,
1026
- "Quantity": 77.08,
1027
- "Expression": 75.00,
1028
- "Material": 85.85,
1029
- "Size": 89.58,
1030
- "Shape": 78.75,
1031
- "Color": 98.33,
1032
-
1033
- "Action-Overall": 79.85,
1034
- "Hand": 73.72,
1035
- "Full body": 84.24,
1036
- "Animal": 81.62,
1037
- "Non Contact": 76.53,
1038
- "Contact": 76.79,
1039
- "State": 84.91,
1040
-
1041
- "Relationship-Overall": 82.36,
1042
- "Composition": 83.45,
1043
- "Similarity": 73.89,
1044
- "Inclusion": 89.13,
1045
- "Comparison": 82.03,
1046
-
1047
- "Compound-Overall": 74.10,
1048
- "Imagination": 80.10,
1049
- "Feature matching": 67.97,
1050
-
1051
- "Grammar-Overall": 76.74,
1052
- "Pronoun Reference": 86.03,
1053
- "Consistency": 75.00,
1054
- "Negation": 68.46,
1055
-
1056
- "Layout-Overall": 86.19,
1057
- "2D": 88.24,
1058
- "3D": 84.09,
1059
-
1060
- "Logical Reasoning": 56.36,
1061
-
1062
- "Text": 51.44
1063
- },
1064
- {
1065
- "model": "Recraft",
1066
- "link": "https://www.recraft.ai/docs#generate-image",
1067
- "hf": "-",
1068
- "open_source": false,
1069
- "release_date": "2024-12",
1070
-
1071
- "Overall": 62.63,
1072
- "Style": 87.20,
1073
- "World Knowledge": 90.19,
1074
-
1075
- "Attribute-Overall": 68.16,
1076
- "Quantity": 68.06,
1077
- "Expression": 56.41,
1078
- "Material": 70.75,
1079
- "Size": 65.97,
1080
- "Shape": 57.50,
1081
- "Color": 95.83,
1082
-
1083
- "Action-Overall": 60.55,
1084
- "Hand": 50.00,
1085
- "Full body": 70.65,
1086
- "Animal": 76.47,
1087
- "Non Contact": 55.61,
1088
- "Contact": 48.81,
1089
- "State": 63.21,
1090
-
1091
- "Relationship-Overall": 62.56,
1092
- "Composition": 64.53,
1093
- "Similarity": 59.44,
1094
- "Inclusion": 59.24,
1095
- "Comparison": 67.19,
1096
-
1097
- "Compound-Overall": 44.85,
1098
- "Imagination": 43.37,
1099
- "Feature matching": 46.35,
1100
-
1101
- "Grammar-Overall": 63.64,
1102
- "Pronoun Reference": 73.16,
1103
- "Consistency": 58.33,
1104
- "Negation": 58.08,
1105
-
1106
- "Layout-Overall": 57.84,
1107
- "2D": 58.82,
1108
- "3D": 56.82,
1109
-
1110
- "Logical Reasoning": 29.55,
1111
-
1112
- "Text": 61.78
1113
- },
1114
- {
1115
- "model": "FLUX-kontext-max",
1116
- "link": "https://bfl.ai/models/flux-kontext",
1117
- "hf": "-",
1118
- "open_source": false,
1119
- "release_date": "2025-05",
1120
-
1121
- "Overall": 80.00,
1122
- "Style": 96.59,
1123
- "World Knowledge": 94.19,
1124
-
1125
- "Attribute-Overall": 80.93,
1126
- "Quantity": 75.69,
1127
- "Expression": 74.32,
1128
- "Material": 82.55,
1129
- "Size": 86.81,
1130
- "Shape": 74.38,
1131
- "Color": 94.17,
1132
-
1133
- "Action-Overall": 77.38,
1134
- "Hand": 67.95,
1135
- "Full body": 83.15,
1136
- "Animal": 77.94,
1137
- "Non Contact": 77.04,
1138
- "Contact": 70.83,
1139
- "State": 84.43,
1140
-
1141
- "Relationship-Overall": 85.08,
1142
- "Composition": 87.50,
1143
- "Similarity": 78.89,
1144
- "Inclusion": 90.00,
1145
- "Comparison": 81.25,
1146
-
1147
- "Compound-Overall": 78.99,
1148
- "Imagination": 83.93,
1149
- "Feature matching": 73.96,
1150
-
1151
- "Grammar-Overall": 78.53,
1152
- "Pronoun Reference": 84.23,
1153
- "Consistency": 78.70,
1154
- "Negation": 72.69,
1155
-
1156
- "Layout-Overall": 85.04,
1157
- "2D": 86.74,
1158
- "3D": 88.33,
1159
-
1160
- "Logical Reasoning": 61.36,
1161
-
1162
- "Text": 61.92
1163
- },
1164
- {
1165
- "model": "FLUX-kontext-pro",
1166
- "link": "https://bfl.ai/models/flux-kontext",
1167
- "hf": "-",
1168
- "open_source": false,
1169
- "release_date": "2025-05",
1170
-
1171
- "Overall": 75.84,
1172
- "Style": 94.78,
1173
- "World Knowledge": 91.61,
1174
-
1175
- "Attribute-Overall": 79.20,
1176
- "Quantity": 75.00,
1177
- "Expression": 71.62,
1178
- "Material": 76.89,
1179
- "Size": 84.72,
1180
- "Shape": 74.38,
1181
- "Color": 97.50,
1182
-
1183
- "Action-Overall": 77.66,
1184
- "Hand": 75.00,
1185
- "Full body": 79.35,
1186
- "Animal": 80.88,
1187
- "Non Contact": 71.94,
1188
- "Contact": 73.21,
1189
- "State": 84.91,
1190
-
1191
- "Relationship-Overall": 79.34,
1192
- "Composition": 81.42,
1193
- "Similarity": 75.56,
1194
- "Inclusion": 83.33,
1195
- "Comparison": 74.22,
1196
-
1197
- "Compound-Overall": 72.68,
1198
- "Imagination": 75.00,
1199
- "Feature matching": 70.31,
1200
-
1201
- "Grammar-Overall": 72.69,
1202
- "Pronoun Reference": 84.23,
1203
- "Consistency": 76.85,
1204
- "Negation": 57.69,
1205
-
1206
- "Layout-Overall": 84.47,
1207
- "2D": 85.98,
1208
- "3D": 82.95,
1209
-
1210
- "Logical Reasoning": 55.68,
1211
-
1212
- "Text": 50.29
1213
- },
1214
- {
1215
- "model": "wan2.2-t2i-plus",
1216
- "link": "https://help.aliyun.com/zh/model-studio/text-to-image-v2-api-reference",
1217
  "hf": "-",
1218
  "open_source": false,
1219
- "release_date": "2025-07",
1220
-
1221
- "Overall": 64.82,
1222
- "Style": 91.10,
1223
- "World Knowledge": 87.34,
1224
-
1225
- "Attribute-Overall": 70.19,
1226
- "Quantity": 76.39,
1227
- "Expression": 55.77,
1228
- "Material": 66.51,
1229
- "Size": 71.53,
1230
- "Shape": 64.38,
1231
- "Color": 94.17,
1232
-
1233
- "Action-Overall": 68.00,
1234
- "Hand": 58.33,
1235
- "Full body": 75.82,
1236
- "Animal": 69.12,
1237
- "Non Contact": 68.88,
1238
- "Contact": 57.74,
1239
- "State": 75.00,
1240
-
1241
- "Relationship-Overall": 73.03,
1242
- "Composition": 70.27,
1243
- "Similarity": 67.98,
1244
- "Inclusion": 77.72,
1245
- "Comparison": 76.69,
1246
-
1247
- "Compound-Overall": 61.37,
1248
- "Imagination": 66.92,
1249
- "Feature matching": 55.73,
1250
-
1251
- "Grammar-Overall": 66.53,
1252
- "Pronoun Reference": 73.90,
1253
- "Consistency": 56.74,
1254
- "Negation": 66.92,
1255
-
1256
- "Layout-Overall": 74.77,
1257
- "2D": 77.49,
1258
- "3D": 71.97,
1259
-
1260
- "Logical Reasoning": 42.05,
1261
-
1262
- "Text": 13.83
1263
- },
1264
- {
1265
- "model": "FLUX.1-dev",
1266
- "link": "https://bfl.ai/blog/24-08-01-bfl",
1267
- "hf": "https://huggingface.co/black-forest-labs/FLUX.1-dev",
1268
- "open_source": true,
1269
- "release_date": "2024-08",
1270
-
1271
- "Overall": 61.30,
1272
- "Style": 83.90,
1273
- "World Knowledge": 88.92,
1274
-
1275
- "Attribute-Overall": 67.84,
1276
- "Quantity": 72.22,
1277
- "Expression": 53.85,
1278
- "Material": 58.96,
1279
- "Size": 75.00,
1280
- "Shape": 65.00,
1281
- "Color": 91.67,
1282
-
1283
- "Action-Overall": 62.17,
1284
- "Hand": 51.28,
1285
- "Full body": 67.39,
1286
- "Animal": 69.85,
1287
- "Non Contact": 59.69,
1288
- "Contact": 58.93,
1289
- "State": 65.57,
1290
-
1291
- "Relationship-Overall": 67.26,
1292
- "Composition": 62.50,
1293
- "Similarity": 66.67,
1294
- "Inclusion": 72.83,
1295
- "Comparison": 62.50,
1296
-
1297
- "Compound-Overall": 47.04,
1298
- "Imagination": 47.96,
1299
- "Feature matching": 46.09,
1300
-
1301
- "Grammar-Overall": 60.96,
1302
- "Pronoun Reference": 73.16,
1303
- "Consistency": 63.43,
1304
- "Negation": 46.15,
1305
-
1306
- "Layout-Overall": 71.83,
1307
- "2D": 74.26,
1308
- "3D": 69.32,
1309
-
1310
- "Logical Reasoning": 30.91,
1311
-
1312
- "Text": 32.18
1313
- },
1314
- {
1315
- "model": "HiDream-I1-Full",
1316
- "link": "https://arxiv.org/pdf/2505.22705",
1317
- "hf": "https://huggingface.co/HiDream-ai/HiDream-I1-Full",
1318
- "open_source": true,
1319
- "release_date": "2025-05",
1320
-
1321
- "Overall": 71.81,
1322
- "Style": 92.50,
1323
- "World Knowledge": 94.15,
1324
-
1325
- "Attribute-Overall": 72.97,
1326
- "Quantity": 73.61,
1327
- "Expression": 59.62,
1328
- "Material": 72.17,
1329
- "Size": 79.17,
1330
- "Shape": 61.88,
1331
- "Color": 98.33,
1332
-
1333
- "Action-Overall": 73.00,
1334
- "Hand": 62.18,
1335
- "Full body": 76.09,
1336
- "Animal": 73.53,
1337
- "Non Contact": 74.49,
1338
- "Contact": 70.24,
1339
- "State": 78.77,
1340
-
1341
- "Relationship-Overall": 75.38,
1342
- "Composition": 79.05,
1343
- "Similarity": 68.33,
1344
- "Inclusion": 78.26,
1345
- "Comparison": 72.66,
1346
-
1347
- "Compound-Overall": 62.63,
1348
- "Imagination": 64.29,
1349
- "Feature matching": 60.94,
1350
-
1351
- "Grammar-Overall": 63.24,
1352
- "Pronoun Reference": 83.09,
1353
- "Consistency": 65.74,
1354
- "Negation": 40.38,
1355
-
1356
- "Layout-Overall": 78.17,
1357
- "2D": 82.72,
1358
- "3D": 73.48,
1359
-
1360
- "Logical Reasoning": 41.14,
1361
-
1362
- "Text": 64.94
1363
- },
1364
- {
1365
- "model": "Pref-GRPO",
1366
- "link": "https://github.com/CodeGoat24/UnifiedReward",
1367
- "hf": "https://huggingface.co/CodeGoat24/FLUX.1-dev-PrefGRPO",
1368
- "open_source": true,
1369
- "release_date": "2025-08",
1370
-
1371
- "Overall": 69.46,
1372
- "Style": 88.40,
1373
- "World Knowledge": 90.35,
1374
-
1375
- "Attribute-Overall": 75.00,
1376
- "Quantity": 71.53,
1377
- "Expression": 60.90,
1378
- "Material": 73.11,
1379
- "Size": 77.08,
1380
- "Shape": 74.38,
1381
- "Color": 99.17,
1382
-
1383
- "Action-Overall": 69.77,
1384
- "Hand": 60.90,
1385
- "Full body": 72.28,
1386
- "Animal": 77.21,
1387
- "Non Contact": 68.37,
1388
- "Contact": 64.88,
1389
- "State": 74.53,
1390
-
1391
- "Relationship-Overall": 76.52,
1392
- "Composition": 81.42,
1393
- "Similarity": 76.67,
1394
- "Inclusion": 76.09,
1395
- "Comparison": 65.62,
1396
-
1397
- "Compound-Overall": 63.27,
1398
- "Imagination": 65.56,
1399
- "Feature matching": 60.94,
1400
-
1401
- "Grammar-Overall": 62.43,
1402
- "Pronoun Reference": 79.04,
1403
- "Consistency": 66.20,
1404
- "Negation": 41.92,
1405
-
1406
- "Layout-Overall": 77.61,
1407
- "2D": 82.35,
1408
- "3D": 72.73,
1409
-
1410
- "Logical Reasoning": 47.13,
1411
-
1412
- "Text": 47.13
1413
- },
1414
- {
1415
- "model": "SD-3.5-Large",
1416
- "link": "https://stability.ai/news/introducing-stable-diffusion-3-5",
1417
- "hf": "https://huggingface.co/stabilityai/stable-diffusion-3.5-large",
1418
- "open_source": true,
1419
- "release_date": "2024-10",
1420
-
1421
- "Overall": 62.99,
1422
- "Style": 88.60,
1423
- "World Knowledge": 88.92,
1424
-
1425
- "Attribute-Overall": 68.59,
1426
- "Quantity": 71.53,
1427
- "Expression": 51.92,
1428
- "Material": 68.87,
1429
- "Size": 68.06,
1430
- "Shape": 65.62,
1431
- "Color": 90.83,
1432
-
1433
- "Action-Overall": 62.17,
1434
- "Hand": 57.05,
1435
- "Full body": 61.96,
1436
- "Animal": 63.24,
1437
- "Non Contact": 62.24,
1438
- "Contact": 59.52,
1439
- "State": 67.45,
1440
-
1441
- "Relationship-Overall": 69.80,
1442
- "Composition": 75.34,
1443
- "Similarity": 68.33,
1444
- "Inclusion": 68.48,
1445
- "Comparison": 60.94,
1446
-
1447
- "Compound-Overall": 58.76,
1448
- "Imagination": 64.80,
1449
- "Feature matching": 52.60,
1450
-
1451
- "Grammar-Overall": 58.96,
1452
- "Pronoun Reference": 74.63,
1453
- "Consistency": 61.11,
1454
- "Negation": 40.77,
1455
-
1456
- "Layout-Overall": 69.03,
1457
- "2D": 70.96,
1458
- "3D": 67.05,
1459
-
1460
- "Logical Reasoning": 32.27,
1461
-
1462
- "Text": 32.76
1463
- },
1464
- {
1465
- "model": "Janus-Pro",
1466
- "link": "https://arxiv.org/pdf/2501.17811",
1467
- "hf": "https://huggingface.co/deepseek-ai/Janus-Pro-7B",
1468
  "open_source": true,
1469
- "release_date": "2025-01",
1470
-
1471
- "Overall": 61.61,
1472
- "Style": 90.80,
1473
- "World Knowledge": 86.71,
1474
- "Attribute-Overall": 67.74,
1475
- "Quantity": 56.25,
1476
- "Expression": 55.77,
1477
- "Material": 71.70,
1478
- "Size": 73.61,
1479
- "Shape": 61.88,
1480
- "Color": 90.83,
1481
-
1482
- "Action-Overall": 64.26,
1483
- "Hand": 50.64,
1484
- "Full body": 63.04,
1485
- "Animal": 75.00,
1486
- "Non Contact": 62.24,
1487
- "Contact": 56.55,
1488
- "State": 76.42,
1489
-
1490
- "Relationship-Overall": 68.40,
1491
- "Composition": 76.01,
1492
- "Similarity": 56.11,
1493
- "Inclusion": 75.00,
1494
- "Comparison": 58.59,
1495
-
1496
- "Compound-Overall": 62.11,
1497
- "Imagination": 69.64,
1498
- "Feature matching": 54.43,
1499
-
1500
- "Grammar-Overall": 64.44,
1501
- "Pronoun Reference": 75.37,
1502
- "Consistency": 66.20,
1503
- "Negation": 51.54,
1504
-
1505
- "Layout-Overall": 72.01,
1506
- "2D": 74.63,
1507
- "3D": 69.32,
1508
-
1509
- "Logical Reasoning": 37.05,
1510
-
1511
- "Text": 2.59
1512
  },
1513
  {
1514
- "model": "Show-o2",
1515
- "link": "https://arxiv.org/pdf/2506.15564",
1516
- "hf": "https://huggingface.co/showlab/show-o2-7B",
1517
  "open_source": true,
1518
  "release_date": "2025-06",
1519
-
1520
- "Overall": 62.73,
1521
- "Style": 87.20,
1522
- "World Knowledge": 86.08,
1523
- "Attribute-Overall": 70.51,
1524
- "Quantity": 59.03,
1525
- "Expression": 63.46,
1526
- "Material": 73.58,
1527
- "Size": 72.92,
1528
- "Shape": 63.12,
1529
- "Color": 95.00,
1530
-
1531
- "Action-Overall": 69.58,
1532
- "Hand": 56.41,
1533
- "Full body": 77.72,
1534
- "Animal": 72.79,
1535
- "Non Contact": 70.41,
1536
- "Contact": 52.38,
1537
- "State": 83.02,
1538
-
1539
- "Relationship-Overall": 70.18,
1540
- "Composition": 79.05,
1541
- "Similarity": 61.11,
1542
- "Inclusion": 70.11,
1543
- "Comparison": 62.50,
1544
-
1545
- "Compound-Overall": 64.69,
1546
- "Imagination": 69.90,
1547
- "Feature matching": 59.38,
1548
-
1549
- "Grammar-Overall": 61.63,
1550
- "Pronoun Reference": 75.37,
1551
- "Consistency": 65.28,
1552
- "Negation": 44.23,
1553
-
1554
- "Layout-Overall": 75.37,
1555
- "2D": 77.94,
1556
- "3D": 72.73,
1557
-
1558
- "Logical Reasoning": 40.91,
1559
-
1560
- "Text": 1.15
1561
- },
1562
- {
1563
- "model": "Bagel",
 
 
 
 
 
 
 
1564
  "link": "https://arxiv.org/pdf/2505.14683",
1565
  "hf": "https://huggingface.co/ByteDance-Seed/BAGEL-7B-MoT",
1566
  "open_source": true,
1567
  "release_date": "2025-05",
1568
-
1569
- "Overall": 61.53,
1570
- "Style": 90.20,
1571
- "World Knowledge": 85.60,
1572
- "Attribute-Overall": 67.74,
1573
- "Quantity": 59.03,
1574
- "Expression": 50.00,
1575
- "Material": 72.64,
1576
- "Size": 76.39,
1577
- "Shape": 59.38,
1578
- "Color": 93.33,
1579
-
1580
- "Action-Overall": 61.98,
1581
- "Hand": 52.56,
1582
- "Full body": 60.87,
1583
- "Animal": 69.12,
1584
- "Non Contact": 62.24,
1585
- "Contact": 58.93,
1586
- "State": 67.45,
1587
-
1588
- "Relationship-Overall": 70.69,
1589
- "Composition": 76.35,
1590
- "Similarity": 70.56,
1591
- "Inclusion": 69.57,
1592
- "Comparison": 59.38,
1593
-
1594
- "Compound-Overall": 58.12,
1595
- "Imagination": 67.35,
1596
- "Feature matching": 48.70,
1597
-
1598
- "Grammar-Overall": 66.44,
1599
- "Pronoun Reference": 71.69,
1600
- "Consistency": 68.52,
1601
- "Negation": 59.23,
1602
-
1603
- "Layout-Overall": 76.49,
1604
- "2D": 79.04,
1605
- "3D": 73.86,
1606
-
1607
- "Logical Reasoning": 30.23,
1608
-
1609
- "Text": 7.76
1610
- },
1611
- {
1612
- "model": "BLIP3-o",
1613
- "link": "https://arxiv.org/pdf/2505.09568",
1614
- "hf": "https://huggingface.co/BLIP3o/BLIP3o-Model-8B",
1615
- "open_source": true,
1616
- "release_date": "2025-05",
1617
-
1618
- "Overall": 59.87,
1619
- "Style": 92.80,
1620
- "World Knowledge": 80.22,
1621
- "Attribute-Overall": 63.89,
1622
- "Quantity": 51.39,
1623
- "Expression": 60.26,
1624
- "Material": 64.62,
1625
- "Size": 75.00,
1626
- "Shape": 54.37,
1627
- "Color": 81.67,
1628
-
1629
- "Action-Overall": 63.97,
1630
- "Hand": 58.33,
1631
- "Full body": 70.11,
1632
- "Animal": 70.59,
1633
- "Non Contact": 60.20,
1634
- "Contact": 51.79,
1635
- "State": 71.70,
1636
-
1637
- "Relationship-Overall": 66.50,
1638
- "Composition": 70.61,
1639
- "Similarity": 60.00,
1640
- "Inclusion": 67.39,
1641
- "Comparison": 64.84,
1642
-
1643
- "Compound-Overall": 53.74,
1644
- "Imagination": 61.73,
1645
- "Feature matching": 45.57,
1646
-
1647
- "Grammar-Overall": 68.58,
1648
- "Pronoun Reference": 79.04,
1649
- "Consistency": 61.11,
1650
- "Negation": 63.85,
1651
-
1652
- "Layout-Overall": 68.47,
1653
- "2D": 72.79,
1654
- "3D": 64.02,
1655
-
1656
- "Logical Reasoning": 39.55,
1657
-
1658
- "Text": 1.15
1659
- },
1660
- {
1661
- "model": "CogView4",
1662
- "link": "https://arxiv.org/pdf/2403.05121",
1663
- "hf": "https://huggingface.co/zai-org/CogView4-6B",
1664
- "open_source": true,
1665
- "release_date": "2024-03",
1666
-
1667
- "Overall": 56.30,
1668
- "Style": 82.00,
1669
- "World Knowledge": 83.07,
1670
- "Attribute-Overall": 63.25,
1671
- "Quantity": 71.53,
1672
- "Expression": 44.23,
1673
- "Material": 55.19,
1674
- "Size": 72.22,
1675
- "Shape": 57.50,
1676
- "Color": 89.17,
1677
-
1678
- "Action-Overall": 57.51,
1679
- "Hand": 53.85,
1680
- "Full body": 59.78,
1681
- "Animal": 68.38,
1682
- "Non Contact": 50.51,
1683
- "Contact": 51.19,
1684
- "State": 62.74,
1685
-
1686
- "Relationship-Overall": 62.44,
1687
- "Composition": 60.47,
1688
- "Similarity": 60.00,
1689
- "Inclusion": 69.57,
1690
- "Comparison": 60.16,
1691
-
1692
- "Compound-Overall": 44.72,
1693
- "Imagination": 47.19,
1694
- "Feature matching": 42.19,
1695
-
1696
- "Grammar-Overall": 54.81,
1697
- "Pronoun Reference": 69.49,
1698
- "Consistency": 56.02,
1699
- "Negation": 38.46,
1700
-
1701
- "Layout-Overall": 69.22,
1702
- "2D": 77.21,
1703
- "3D": 60.98,
1704
-
1705
- "Logical Reasoning": 28.18,
1706
-
1707
- "Text": 17.82
1708
- },
1709
- {
1710
- "model": "Hunyuan-DiT",
1711
- "link": "https://arxiv.org/pdf/2405.08748",
1712
- "hf": "https://huggingface.co/Tencent-Hunyuan/HunyuanDiT",
1713
- "open_source": true,
1714
- "release_date": "2024-05",
1715
-
1716
- "Overall": 51.38,
1717
-
1718
- "Style": 94.10,
1719
-
1720
- "World Knowledge": 80.70,
1721
-
1722
- "Attribute-Overall": 62.71,
1723
- "Quantity": 67.36,
1724
- "Expression": 44.23,
1725
- "Material": 71.70,
1726
- "Size": 61.81,
1727
- "Shape": 47.50,
1728
- "Color": 86.67,
1729
-
1730
- "Action-Overall": 49.05,
1731
- "Hand": 35.90,
1732
- "Full body": 54.89,
1733
- "Animal": 54.41,
1734
- "Non Contact": 46.94,
1735
- "Contact": 35.71,
1736
- "State": 62.74,
1737
-
1738
- "Relationship-Overall": 59.64,
1739
- "Composition": 60.14,
1740
- "Similarity": 64.44,
1741
- "Inclusion": 60.33,
1742
- "Comparison": 50.78,
1743
-
1744
- "Compound-Overall": 41.62,
1745
- "Imagination": 46.68,
1746
- "Feature matching": 36.46,
1747
-
1748
- "Grammar-Overall": 55.48,
1749
- "Pronoun Reference": 62.87,
1750
- "Consistency": 57.87,
1751
- "Negation": 45.77,
1752
-
1753
- "Layout-Overall": 44.78,
1754
- "2D": 39.34,
1755
- "3D": 50.38,
1756
-
1757
- "Logical Reasoning": 24.55,
1758
-
1759
- "Text": 1.15
1760
- },
1761
- {
1762
- "model": "Janus",
1763
- "link": "https://arxiv.org/pdf/2410.13848",
1764
- "hf": "https://huggingface.co/deepseek-ai/Janus-1.3B",
1765
- "open_source": true,
1766
- "release_date": "2024-10",
1767
-
1768
- "Overall": 51.23,
1769
-
1770
- "Style": 89.90,
1771
-
1772
- "World Knowledge": 73.58,
1773
-
1774
- "Attribute-Overall": 54.81,
1775
- "Quantity": 37.50,
1776
- "Expression": 37.82,
1777
- "Material": 58.96,
1778
- "Size": 65.97,
1779
- "Shape": 47.50,
1780
- "Color": 86.67,
1781
-
1782
- "Action-Overall": 50.38,
1783
- "Hand": 32.69,
1784
- "Full body": 51.63,
1785
- "Animal": 61.76,
1786
- "Non Contact": 48.47,
1787
- "Contact": 38.10,
1788
- "State": 66.51,
1789
-
1790
- "Relationship-Overall": 55.08,
1791
- "Composition": 56.76,
1792
- "Similarity": 53.89,
1793
- "Inclusion": 59.24,
1794
- "Comparison": 46.88,
1795
-
1796
- "Compound-Overall": 46.65,
1797
- "Imagination": 58.16,
1798
- "Feature matching": 34.90,
1799
-
1800
- "Grammar-Overall": 59.09,
1801
- "Pronoun Reference": 66.18,
1802
- "Consistency": 51.39,
1803
- "Negation": 58.08,
1804
-
1805
- "Layout-Overall": 54.85,
1806
- "2D": 57.72,
1807
- "3D": 51.89,
1808
-
1809
- "Logical Reasoning": 26.82,
1810
-
1811
- "Text": 1.15
1812
- },
1813
- {
1814
- "model": "Janus-flow",
1815
- "link": "https://arxiv.org/pdf/2411.07975",
1816
- "hf": "https://huggingface.co/deepseek-ai/JanusFlow-1.3B",
1817
- "open_source": true,
1818
- "release_date": "2024-11",
1819
-
1820
- "Overall": 46.39,
1821
-
1822
- "Style": 86.20,
1823
-
1824
- "World Knowledge": 62.50,
1825
-
1826
- "Attribute-Overall": 47.97,
1827
- "Quantity": 43.06,
1828
- "Expression": 30.77,
1829
- "Material": 55.19,
1830
- "Size": 55.56,
1831
- "Shape": 30.00,
1832
- "Color": 78.33,
1833
-
1834
- "Action-Overall": 43.35,
1835
- "Hand": 23.08,
1836
- "Full body": 48.37,
1837
- "Animal": 58.82,
1838
- "Non Contact": 36.73,
1839
- "Contact": 36.31,
1840
- "State": 55.66,
1841
-
1842
- "Relationship-Overall": 50.00,
1843
- "Composition": 59.80,
1844
- "Similarity": 38.89,
1845
- "Inclusion": 51.63,
1846
- "Comparison": 40.62,
1847
-
1848
- "Compound-Overall": 45.10,
1849
- "Imagination": 57.65,
1850
- "Feature matching": 32.29,
1851
-
1852
- "Grammar-Overall": 60.29,
1853
- "Pronoun Reference": 66.18,
1854
- "Consistency": 48.61,
1855
- "Negation": 63.85,
1856
-
1857
- "Layout-Overall": 46.46,
1858
- "2D": 49.26,
1859
- "3D": 43.56,
1860
-
1861
- "Logical Reasoning": 21.14,
1862
-
1863
- "Text": 0.86
1864
- },
1865
- {
1866
- "model": "Emu3",
1867
- "link": "https://arxiv.org/pdf/2409.18869",
1868
- "hf": "https://huggingface.co/BAAI/Emu3-Gen",
1869
- "open_source": true,
1870
- "release_date": "2024-09",
1871
-
1872
- "Overall": 46.02,
1873
-
1874
- "Style": 86.80,
1875
-
1876
- "World Knowledge": 77.06,
1877
-
1878
- "Attribute-Overall": 51.39,
1879
- "Quantity": 44.44,
1880
- "Expression": 45.51,
1881
- "Material": 53.77,
1882
- "Size": 43.06,
1883
- "Shape": 46.25,
1884
- "Color": 80.00,
1885
-
1886
- "Action-Overall": 40.11,
1887
- "Hand": 25.00,
1888
- "Full body": 47.28,
1889
- "Animal": 50.74,
1890
- "Non Contact": 35.20,
1891
- "Contact": 27.98,
1892
- "State": 52.36,
1893
-
1894
- "Relationship-Overall": 49.75,
1895
- "Composition": 56.76,
1896
- "Similarity": 46.67,
1897
- "Inclusion": 48.37,
1898
- "Comparison": 39.84,
1899
-
1900
- "Compound-Overall": 36.86,
1901
- "Imagination": 41.33,
1902
- "Feature matching": 32.29,
1903
-
1904
- "Grammar-Overall": 52.94,
1905
- "Pronoun Reference": 59.56,
1906
- "Consistency": 53.70,
1907
- "Negation": 45.38,
1908
-
1909
- "Layout-Overall": 44.78,
1910
- "2D": 45.22,
1911
- "3D": 44.32,
1912
-
1913
- "Logical Reasoning": 19.32,
1914
-
1915
- "Text": 1.15
1916
- },
1917
- {
1918
- "model": "Playground2.5",
1919
- "link": "https://arxiv.org/pdf/2402.17245",
1920
- "hf": "https://huggingface.co/playgroundai/playground-v2.5-1024px-aesthetic",
1921
- "open_source": true,
1922
- "release_date": "2024-02",
1923
-
1924
- "Overall": 45.61,
1925
-
1926
- "Style": 89.50,
1927
-
1928
- "World Knowledge": 76.11,
1929
-
1930
- "Attribute-Overall": 52.78,
1931
- "Quantity": 58.33,
1932
- "Expression": 43.59,
1933
- "Material": 57.08,
1934
- "Size": 44.44,
1935
- "Shape": 41.25,
1936
- "Color": 75.83,
1937
-
1938
- "Action-Overall": 42.68,
1939
- "Hand": 28.85,
1940
- "Full body": 50.00,
1941
- "Animal": 52.21,
1942
- "Non Contact": 35.20,
1943
- "Contact": 29.17,
1944
- "State": 58.02,
1945
-
1946
- "Relationship-Overall": 51.52,
1947
- "Composition": 60.14,
1948
- "Similarity": 49.44,
1949
- "Inclusion": 48.37,
1950
- "Comparison": 39.06,
1951
-
1952
- "Compound-Overall": 35.44,
1953
- "Imagination": 43.88,
1954
- "Feature matching": 26.82,
1955
-
1956
- "Grammar-Overall": 53.21,
1957
- "Pronoun Reference": 58.82,
1958
- "Consistency": 50.00,
1959
- "Negation": 50.00,
1960
-
1961
- "Layout-Overall": 37.13,
1962
- "2D": 34.56,
1963
- "3D": 39.77,
1964
-
1965
- "Logical Reasoning": 16.59,
1966
-
1967
- "Text": 1.15
1968
- },
1969
- {
1970
- "model": "SDXL",
1971
- "link": "https://arxiv.org/pdf/2307.01952",
1972
- "hf": "https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0",
1973
- "open_source": true,
1974
- "release_date": "2023-07",
1975
-
1976
- "Overall": 39.75,
1977
-
1978
- "Style": 87.40,
1979
-
1980
- "World Knowledge": 72.63,
1981
-
1982
- "Attribute-Overall": 44.34,
1983
- "Quantity": 44.44,
1984
- "Expression": 25.00,
1985
- "Material": 52.83,
1986
- "Size": 44.44,
1987
- "Shape": 33.75,
1988
- "Color": 68.33,
1989
-
1990
- "Action-Overall": 34.22,
1991
- "Hand": 19.23,
1992
- "Full body": 35.33,
1993
- "Animal": 43.38,
1994
- "Non Contact": 26.53,
1995
- "Contact": 24.40,
1996
- "State": 53.30,
1997
-
1998
- "Relationship-Overall": 44.92,
1999
- "Composition": 53.72,
2000
- "Similarity": 38.33,
2001
- "Inclusion": 39.67,
2002
- "Comparison": 41.41,
2003
-
2004
- "Compound-Overall": 26.68,
2005
- "Imagination": 33.93,
2006
- "Feature matching": 19.27,
2007
-
2008
- "Grammar-Overall": 47.33,
2009
- "Pronoun Reference": 50.37,
2010
- "Consistency": 42.59,
2011
- "Negation": 48.08,
2012
-
2013
- "Layout-Overall": 29.85,
2014
- "2D": 26.47,
2015
- "3D": 33.33,
2016
-
2017
- "Logical Reasoning": 9.55,
2018
-
2019
- "Text": 1.15
2020
- },
2021
- {
2022
- "model": "GPT-4o",
2023
- "link": "https://platform.openai.com/docs/guides/image-generation",
2024
- "hf": "-",
2025
- "open_source": false,
2026
- "release_date": "2025-03",
2027
-
2028
- "Overall": 92.77,
2029
-
2030
- "Style": 98.57,
2031
-
2032
- "World Knowledge": 98.87,
2033
-
2034
- "Attribute-Overall": 93.59,
2035
- "Quantity": 90.00,
2036
- "Expression": 94.70,
2037
- "Material": 94.20,
2038
- "Size": 91.61,
2039
- "Shape": 92.50,
2040
- "Color": 99.17,
2041
-
2042
- "Action-Overall": 90.79,
2043
- "Hand": 89.74,
2044
- "Full body": 92.22,
2045
- "Animal": 87.12,
2046
- "Non Contact": 90.43,
2047
- "Contact": 89.82,
2048
- "State": 93.75,
2049
-
2050
- "Relationship-Overall": 94.97,
2051
- "Composition": 96.23,
2052
- "Similarity": 95.00,
2053
- "Inclusion": 94.89,
2054
- "Comparison": 92.19,
2055
-
2056
- "Compound-Overall": 93.55,
2057
- "Imagination": 95.64,
2058
- "Feature matching": 91.40,
2059
-
2060
- "Grammar-Overall": 91.76,
2061
- "Pronoun Reference": 92.91,
2062
- "Consistency": 91.67,
2063
- "Negation": 90.57,
2064
-
2065
- "Layout-Overall": 91.35,
2066
- "2D": 91.04,
2067
- "3D": 91.67,
2068
-
2069
- "Logical Reasoning": 84.97,
2070
-
2071
- "Text": 89.24
2072
- },
2073
- {
2074
- "model": "Imagen-4.0-Ultra-preview-06-06",
2075
- "link": "https://deepmind.google/models/imagen/",
2076
- "hf": "-",
2077
- "open_source": false,
2078
- "release_date": "2025-06",
2079
-
2080
- "Overall": 91.54,
2081
-
2082
- "Style": 99.20,
2083
-
2084
- "World Knowledge": 97.47,
2085
-
2086
- "Attribute-Overall": 92.52,
2087
- "Quantity": 93.06,
2088
- "Expression": 81.41,
2089
- "Material": 94.34,
2090
- "Size": 95.83,
2091
- "Shape": 91.88,
2092
- "Color": 100.0,
2093
-
2094
- "Action-Overall": 92.20,
2095
- "Hand": 90.38,
2096
- "Full body": 93.44,
2097
- "Animal": 91.91,
2098
- "Non Contact": 90.31,
2099
- "Contact": 89.29,
2100
- "State": 96.70,
2101
-
2102
- "Relationship-Overall": 93.02,
2103
- "Composition": 95.27,
2104
- "Similarity": 84.44,
2105
- "Inclusion": 98.37,
2106
- "Comparison": 92.19,
2107
-
2108
- "Compound-Overall": 91.37,
2109
- "Imagination": 92.86,
2110
- "Feature matching": 89.84,
2111
-
2112
- "Grammar-Overall": 87.97,
2113
- "Pronoun Reference": 94.12,
2114
- "Consistency": 87.04,
2115
- "Negation": 82.31,
2116
-
2117
- "Layout-Overall": 93.10,
2118
- "2D": 92.65,
2119
- "3D": 93.56,
2120
-
2121
- "Logical Reasoning": 79.55,
2122
-
2123
- "Text": 89.08
2124
- },
2125
- {
2126
- "model": "Seedream-3.0",
2127
- "link": "https://www.byteplus.com/en/product/Seedream",
2128
- "hf": "-",
2129
- "open_source": false,
2130
- "release_date": "2025-06",
2131
-
2132
- "Overall": 78.95,
2133
-
2134
- "Style": 98.10,
2135
-
2136
- "World Knowledge": 95.25,
2137
-
2138
- "Attribute-Overall": 85.58,
2139
- "Quantity": 80.56,
2140
- "Expression": 82.05,
2141
- "Material": 90.57,
2142
- "Size": 85.42,
2143
- "Shape": 78.12,
2144
- "Color": 97.50,
2145
-
2146
- "Action-Overall": 82.98,
2147
- "Hand": 75.00,
2148
- "Full body": 89.67,
2149
- "Animal": 85.29,
2150
- "Non Contact": 75.51,
2151
- "Contact": 80.95,
2152
- "State": 90.09,
2153
-
2154
- "Relationship-Overall": 80.84,
2155
- "Composition": 82.77,
2156
- "Similarity": 73.89,
2157
- "Inclusion": 84.24,
2158
- "Comparison": 81.25,
2159
-
2160
- "Compound-Overall": 73.84,
2161
- "Imagination": 78.57,
2162
- "Feature matching": 69.01,
2163
-
2164
- "Grammar-Overall": 61.36,
2165
- "Pronoun Reference": 79.78,
2166
- "Consistency": 69.91,
2167
- "Negation": 35.00,
2168
-
2169
- "Layout-Overall": 87.31,
2170
- "2D": 86.76,
2171
- "3D": 87.88,
2172
-
2173
- "Logical Reasoning": 52.73,
2174
-
2175
- "Text": 71.55
2176
- },
2177
- {
2178
- "model": "Imagen-3.0-generate-002",
2179
- "link": "https://arxiv.org/pdf/2408.07009",
2180
- "hf": "-",
2181
- "open_source": false,
2182
- "release_date": "2025-02",
2183
-
2184
- "Overall": 71.85,
2185
-
2186
- "Style": 89.25,
2187
-
2188
- "World Knowledge": 94.75,
2189
-
2190
- "Attribute-Overall": 77.33,
2191
- "Quantity": 75.78,
2192
- "Expression": 64.67,
2193
- "Material": 80.66,
2194
- "Size": 82.84,
2195
- "Shape": 70.00,
2196
- "Color": 93.10,
2197
-
2198
- "Action-Overall": 81.46,
2199
- "Hand": 80.00,
2200
- "Full body": 83.89,
2201
- "Animal": 85.29,
2202
- "Non Contact": 77.37,
2203
- "Contact": 74.40,
2204
- "State": 87.38,
2205
-
2206
- "Relationship-Overall": 82.86,
2207
- "Composition": 83.90,
2208
- "Similarity": 73.33,
2209
- "Inclusion": 88.64,
2210
- "Comparison": 83.90,
2211
-
2212
- "Compound-Overall": 71.71,
2213
- "Imagination": 79.23,
2214
- "Feature matching": 64.06,
2215
-
2216
- "Grammar-Overall": 69.84,
2217
- "Pronoun Reference": 79.04,
2218
- "Consistency": 70.75,
2219
- "Negation": 59.13,
2220
-
2221
- "Layout-Overall": 81.34,
2222
- "2D": 82.72,
2223
- "3D": 79.92,
2224
-
2225
- "Logical Reasoning": 48.36,
2226
-
2227
- "Text": 21.55
2228
- },
2229
- {
2230
- "model": "DALL-E-3",
2231
- "link": "https://openai.com/zh-Hans-CN/index/dall-e-3/",
2232
- "hf": "-",
2233
- "open_source": false,
2234
- "release_date": "2023-09",
2235
-
2236
- "Overall": 69.18,
2237
-
2238
- "Style": 95.06,
2239
-
2240
- "World Knowledge": 93.51,
2241
-
2242
- "Attribute-Overall": 75.97,
2243
- "Quantity": 62.14,
2244
- "Expression": 59.87,
2245
- "Material": 87.74,
2246
- "Size": 87.50,
2247
- "Shape": 65.00,
2248
- "Color": 92.50,
2249
-
2250
- "Action-Overall": 69.83,
2251
- "Hand": 60.90,
2252
- "Full body": 75.00,
2253
- "Animal": 76.47,
2254
- "Non Contact": 66.84,
2255
- "Contact": 63.41,
2256
- "State": 75.47,
2257
-
2258
- "Relationship-Overall": 78.06,
2259
- "Composition": 82.43,
2260
- "Similarity": 69.44,
2261
- "Inclusion": 87.78,
2262
- "Comparison": 66.41,
2263
-
2264
- "Compound-Overall": 70.60,
2265
- "Imagination": 76.79,
2266
- "Feature matching": 64.21,
2267
-
2268
- "Grammar-Overall": 68.07,
2269
- "Pronoun Reference": 74.24,
2270
- "Consistency": 74.07,
2271
- "Negation": 56.64,
2272
-
2273
- "Layout-Overall": 66.67,
2274
- "2D": 57.72,
2275
- "3D": 76.17,
2276
-
2277
- "Logical Reasoning": 48.18,
2278
-
2279
- "Text": 25.86
2280
- },
2281
- {
2282
- "model": "FLUX-pro-1.1-Ultra",
2283
- "link": "https://bfl.ai/",
2284
- "hf": "-",
2285
- "open_source": false,
2286
- "release_date": "2024-11",
2287
-
2288
- "Overall": 70.67,
2289
-
2290
- "Style": 90.60,
2291
-
2292
- "World Knowledge": 91.61,
2293
-
2294
- "Attribute-Overall": 76.50,
2295
- "Quantity": 75.69,
2296
- "Expression": 59.62,
2297
- "Material": 78.77,
2298
- "Size": 77.78,
2299
- "Shape": 74.38,
2300
- "Color": 96.67,
2301
-
2302
- "Action-Overall": 76.50,
2303
- "Hand": 57.69,
2304
- "Full body": 68.48,
2305
- "Animal": 77.21,
2306
- "Non Contact": 76.53,
2307
- "Contact": 64.29,
2308
- "State": 76.89,
2309
-
2310
- "Relationship-Overall": 77.54,
2311
- "Composition": 80.41,
2312
- "Similarity": 72.78,
2313
- "Inclusion": 82.07,
2314
- "Comparison": 71.09,
2315
-
2316
- "Compound-Overall": 67.78,
2317
- "Imagination": 74.74,
2318
- "Feature matching": 60.68,
2319
-
2320
- "Grammar-Overall": 70.05,
2321
- "Pronoun Reference": 84.56,
2322
- "Consistency": 68.98,
2323
- "Negation": 55.77,
2324
-
2325
- "Layout-Overall": 81.53,
2326
- "2D": 80.15,
2327
- "3D": 82.95,
2328
-
2329
- "Logical Reasoning": 43.18,
2330
-
2331
- "Text": 37.36
2332
- },
2333
- {
2334
- "model": "Keling-Ketu",
2335
- "link": "https://kolors.kuaishou.com/",
2336
- "hf": "-",
2337
- "open_source": false,
2338
- "release_date": "2025-04",
2339
-
2340
- "Overall": 65.93,
2341
-
2342
- "Style": 92.27,
2343
-
2344
- "World Knowledge": 86.62,
2345
-
2346
- "Attribute-Overall": 71.66,
2347
- "Quantity": 75.00,
2348
- "Expression": 56.41,
2349
- "Material": 78.77,
2350
- "Size": 79.17,
2351
- "Shape": 53.12,
2352
- "Color": 91.38,
2353
-
2354
- "Action-Overall": 68.73,
2355
- "Hand": 54.49,
2356
- "Full body": 76.09,
2357
- "Animal": 72.79,
2358
- "Non Contact": 69.90,
2359
- "Contact": 58.93,
2360
- "State": 76.89,
2361
-
2362
- "Relationship-Overall": 70.94,
2363
- "Composition": 68.92,
2364
- "Similarity": 70.56,
2365
- "Inclusion": 74.46,
2366
- "Comparison": 71.09,
2367
-
2368
- "Compound-Overall": 60.81,
2369
- "Imagination": 66.24,
2370
- "Feature matching": 55.26,
2371
-
2372
- "Grammar-Overall": 71.26,
2373
- "Pronoun Reference": 77.21,
2374
- "Consistency": 67.59,
2375
- "Negation": 68.08,
2376
-
2377
- "Layout-Overall": 77.23,
2378
- "2D": 80.97,
2379
- "3D": 73.36,
2380
-
2381
- "Logical Reasoning": 43.75,
2382
-
2383
- "Text": 16.03
2384
- },
2385
- {
2386
- "model": "Qwen-Image",
2387
  "link": "https://arxiv.org/pdf/2508.02324",
2388
- "hf": "https://huggingface.co/Qwen/Qwen-Image",
2389
  "open_source": true,
2390
  "release_date": "2025-08",
2391
-
2392
- "Overall": 78.81,
2393
-
2394
- "Style": 95.10,
2395
-
2396
- "World Knowledge": 94.30,
2397
-
2398
- "Attribute-Overall": 87.61,
2399
- "Quantity": 81.94,
2400
- "Expression": 84.62,
2401
- "Material": 91.98,
2402
- "Size": 84.03,
2403
- "Shape": 84.38,
2404
- "Color": 99.17,
2405
-
2406
- "Action-Overall": 84.13,
2407
- "Hand": 82.05,
2408
- "Full body": 88.59,
2409
- "Animal": 88.24,
2410
- "Non Contact": 80.61,
2411
- "Contact": 77.38,
2412
- "State": 87.74,
2413
-
2414
- "Relationship-Overall": 79.70,
2415
- "Composition": 81.76,
2416
- "Similarity": 67.78,
2417
- "Inclusion": 86.96,
2418
- "Comparison": 81.25,
2419
-
2420
- "Compound-Overall": 73.32,
2421
- "Imagination": 73.21,
2422
- "Feature matching": 73.44,
2423
-
2424
- "Grammar-Overall": 60.29,
2425
- "Pronoun Reference": 83.82,
2426
- "Consistency": 70.37,
2427
- "Negation": 27.31,
2428
-
2429
- "Layout-Overall": 85.52,
2430
- "2D": 86.40,
2431
- "3D": 85.23,
2432
-
2433
- "Logical Reasoning": 53.64,
2434
-
2435
- "Text": 76.14
 
 
 
 
2436
  }
2437
  ]
2438
- }
 
1
  {
2
  "leaderboard": [
3
  {
4
+ "model": "FLUX-Kontext-Pro",
5
+ "link": "https://fal.ai/models/fal-ai/flux-pro/kontext",
6
  "hf": "-",
7
  "open_source": false,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8
  "release_date": "2025-05",
9
+ "Overall": 43.72,
10
+ "Real-World-Overall": 43.59,
11
+ "Viewpoint Transformation": 34.75,
12
+ "Material Modification": 56.62,
13
+ "Pose Adjustment": 54.95,
14
+ "Temporal Evolution": 36.36,
15
+ "Structural Integrity Change": 42.15,
16
+ "Motion State Change": 42.52,
17
+ "Spatial Arrangement": 43.89,
18
+ "Mechanical Reaction": 38.01,
19
+ "Medium Interaction": 42.33,
20
+ "Game-World-Overall": 44.36,
21
+ "3D Reconstruction": 43.54,
22
+ "Space Invader": 47.45,
23
+ "Jewel2": 34.09,
24
+ "Pacman": 50.55,
25
+ "Word Search": 34.88,
26
+ "Tictactoe": 58.44,
27
+ "Sudoku": 26.88,
28
+ "Maze": 46.7,
29
+ "Sokoban": 52.9
30
+ },
31
+ {
32
+ "model": "Seedream4.0",
33
+ "link": "https://seed.bytedance.com/en/seedream4_0",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
34
  "hf": "-",
35
  "open_source": false,
36
  "release_date": "2025-09",
37
+ "Overall": 57.05,
38
+ "Real-World-Overall": 68.24,
39
+ "Viewpoint Transformation": 64.77,
40
+ "Material Modification": 74.32,
41
+ "Pose Adjustment": 80.19,
42
+ "Temporal Evolution": 66.07,
43
+ "Structural Integrity Change": 59.48,
44
+ "Motion State Change": 64.68,
45
+ "Spatial Arrangement": 79.48,
46
+ "Mechanical Reaction": 61.58,
47
+ "Medium Interaction": 63.21,
48
+ "Game-World-Overall": 44.63,
49
+ "3D Reconstruction": 39.27,
50
+ "Space Invader": 46.28,
51
+ "Jewel2": 43.0,
52
+ "Pacman": 41.34,
53
+ "Word Search": 48.55,
54
+ "Tictactoe": 38.43,
55
+ "Sudoku": 68.75,
56
+ "Maze": 54.15,
57
+ "Sokoban": 33.43
58
+ },
59
+ {
60
+ "model": "Wan2.5",
61
+ "link": "https://wan.video/",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
62
  "hf": "-",
63
  "open_source": false,
64
+ "release_date": "2025-09",
65
+ "Overall": 60.59,
66
+ "Real-World-Overall": 68.36,
67
+ "Viewpoint Transformation": 72.97,
68
+ "Material Modification": 75.3,
69
+ "Pose Adjustment": 79.97,
70
+ "Temporal Evolution": 59.86,
71
+ "Structural Integrity Change": 64.51,
72
+ "Motion State Change": 66.56,
73
+ "Spatial Arrangement": 61.6,
74
+ "Mechanical Reaction": 63.67,
75
+ "Medium Interaction": 66.66,
76
+ "Game-World-Overall": 54.52,
77
+ "3D Reconstruction": 63.39,
78
+ "Space Invader": 44.67,
79
+ "Jewel2": 53.73,
80
+ "Pacman": 40.0,
81
+ "Word Search": 58.54,
82
+ "Tictactoe": 60.59,
83
+ "Sudoku": 45.94,
84
+ "Maze": 65.17,
85
+ "Sokoban": 47.47
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
86
  },
87
+ {
88
  "model": "Nano Banana",
89
+ "link": "https://aistudio.google.com/models/gemini-2-5-flash-image",
90
  "hf": "-",
91
  "open_source": false,
92
  "release_date": "2025-08",
93
+ "Overall": 68.38,
94
+ "Real-World-Overall": 76.45,
95
+ "Viewpoint Transformation": 75.37,
96
+ "Material Modification": 79.72,
97
+ "Pose Adjustment": 85.55,
98
+ "Temporal Evolution": 71.35,
99
+ "Structural Integrity Change": 70.65,
100
+ "Motion State Change": 73.36,
101
+ "Spatial Arrangement": 73.22,
102
+ "Mechanical Reaction": 79.44,
103
+ "Medium Interaction": 75.34,
104
+ "Game-World-Overall": 61.32,
105
+ "3D Reconstruction": 66.26,
106
+ "Space Invader": 61.43,
107
+ "Jewel2": 54.65,
108
+ "Pacman": 54.4,
109
+ "Word Search": 64.65,
110
+ "Tictactoe": 40.92,
111
+ "Sudoku": 91.99,
112
+ "Maze": 62.65,
113
+ "Sokoban": 50.05
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
114
  },
115
  {
116
+ "model": "GPT-4o",
117
+ "link": "https://platform.openai.com/docs/guides/image-generation",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
118
  "hf": "-",
119
  "open_source": false,
120
+ "release_date": "2025-03",
121
+ "Overall": 71.64,
122
+ "Real-World-Overall": 81.31,
123
+ "Viewpoint Transformation": 83.55,
124
+ "Material Modification": 82.98,
125
+ "Pose Adjustment": 92.16,
126
+ "Temporal Evolution": 75.7,
127
+ "Structural Integrity Change": 76.1,
128
+ "Motion State Change": 76.32,
129
+ "Spatial Arrangement": 88.88,
130
+ "Mechanical Reaction": 78.97,
131
+ "Medium Interaction": 76.28,
132
+ "Game-World-Overall": 65.19,
133
+ "3D Reconstruction": 78.82,
134
+ "Space Invader": 58.88,
135
+ "Jewel2": 44.95,
136
+ "Pacman": 47.02,
137
+ "Word Search": 64.55,
138
+ "Tictactoe": 63.8,
139
+ "Sudoku": 68.62,
140
+ "Maze": 82.27,
141
+ "Sokoban": 49.77
142
+ },
143
+ {
144
+ "model": "MagicBrush",
145
+ "link": "https://arxiv.org/pdf/2306.10012",
146
+ "hf": "https://huggingface.co/MagicBrush/MagicBrush",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
147
  "open_source": true,
148
+ "release_date": "2023-06",
149
+ "Overall": 40.98,
150
+ "Real-World-Overall": 45.23,
151
+ "Viewpoint Transformation": 35.65,
152
+ "Material Modification": 50.55,
153
+ "Pose Adjustment": 48.34,
154
+ "Temporal Evolution": 44.13,
155
+ "Structural Integrity Change": 45.07,
156
+ "Motion State Change": 47.83,
157
+ "Spatial Arrangement": 38.73,
158
+ "Mechanical Reaction": 47.88,
159
+ "Medium Interaction": 47.52,
160
+ "Game-World-Overall": 40.8,
161
+ "3D Reconstruction": 63.43,
162
+ "Space Invader": 44.74,
163
+ "Jewel2": 26.26,
164
+ "Pacman": 30.3,
165
+ "Word Search": 34.12,
166
+ "Tictactoe": 31.05,
167
+ "Sudoku": 40.33,
168
+ "Maze": 32.65,
169
+ "Sokoban": 29.0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
170
  },
171
  {
172
+ "model": "OmniGen2",
173
+ "link": "https://arxiv.org/pdf/2506.18871",
174
+ "hf": "https://huggingface.co/osunlp/InstructPix2Pix-MagicBrush",
175
  "open_source": true,
176
  "release_date": "2025-06",
177
+ "Overall": 43.96,
178
+ "Real-World-Overall": 54.58,
179
+ "Viewpoint Transformation": 50.32,
180
+ "Material Modification": 58.03,
181
+ "Pose Adjustment": 67.73,
182
+ "Temporal Evolution": 47.77,
183
+ "Structural Integrity Change": 49.9,
184
+ "Motion State Change": 57.63,
185
+ "Spatial Arrangement": 51.09,
186
+ "Mechanical Reaction": 55.05,
187
+ "Medium Interaction": 51.98,
188
+ "Game-World-Overall": 40.05,
189
+ "3D Reconstruction": 70.78,
190
+ "Space Invader": 51.17,
191
+ "Jewel2": 1.5,
192
+ "Pacman": 30.17,
193
+ "Word Search": 23.07,
194
+ "Tictactoe": 46.17,
195
+ "Sudoku": 4.85,
196
+ "Maze": 39.63,
197
+ "Sokoban": 34.58
198
+ },
199
+ {
200
+ "model": "Step1X-Edit",
201
+ "link": "https://arxiv.org/pdf/2504.17761",
202
+ "hf": "https://huggingface.co/stepfun-ai/Step1X-Edit",
203
+ "open_source": true,
204
+ "release_date": "2025-04",
205
+ "Overall": 50.12,
206
+ "Real-World-Overall": 57.42,
207
+ "Viewpoint Transformation": 51.82,
208
+ "Material Modification": 67.18,
209
+ "Pose Adjustment": 63.29,
210
+ "Temporal Evolution": 51.45,
211
+ "Structural Integrity Change": 54.98,
212
+ "Motion State Change": 62.18,
213
+ "Spatial Arrangement": 51.03,
214
+ "Mechanical Reaction": 55.72,
215
+ "Medium Interaction": 57.78,
216
+ "Game-World-Overall": 46.43,
217
+ "3D Reconstruction": 62.9,
218
+ "Space Invader": 33.62,
219
+ "Jewel2": 35.4,
220
+ "Pacman": 33.48,
221
+ "Word Search": 43.02,
222
+ "Tictactoe": 49.92,
223
+ "Sudoku": 39.17,
224
+ "Maze": 54.53,
225
+ "Sokoban": 34.73
226
+ },
227
+ {
228
+ "model": "Bagel-Think",
229
  "link": "https://arxiv.org/pdf/2505.14683",
230
  "hf": "https://huggingface.co/ByteDance-Seed/BAGEL-7B-MoT",
231
  "open_source": true,
232
  "release_date": "2025-05",
233
+ "Overall": 51.25,
234
+ "Real-World-Overall": 58.04,
235
+ "Viewpoint Transformation": 58.38,
236
+ "Material Modification": 64.15,
237
+ "Pose Adjustment": 63.38,
238
+ "Temporal Evolution": 56.27,
239
+ "Structural Integrity Change": 54.4,
240
+ "Motion State Change": 58.63,
241
+ "Spatial Arrangement": 48.82,
242
+ "Mechanical Reaction": 56.47,
243
+ "Medium Interaction": 57.75,
244
+ "Game-World-Overall": 48.28,
245
+ "3D Reconstruction": 65.65,
246
+ "Space Invader": 47.3,
247
+ "Jewel2": 42.12,
248
+ "Pacman": 40.97,
249
+ "Word Search": 47.8,
250
+ "Tictactoe": 40.35,
251
+ "Sudoku": 32.4,
252
+ "Maze": 48.83,
253
+ "Sokoban": 38.8
254
+ },
255
+ {
256
+ "model": "Qwen-Image-Edit",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
257
  "link": "https://arxiv.org/pdf/2508.02324",
258
+ "hf": "https://huggingface.co/Qwen/Qwen-Image-Edit",
259
  "open_source": true,
260
  "release_date": "2025-08",
261
+ "Overall": 56.46,
262
+ "Real-World-Overall": 71.92,
263
+ "Viewpoint Transformation": 72.08,
264
+ "Material Modification": 78.31,
265
+ "Pose Adjustment": 81.75,
266
+ "Temporal Evolution": 64.57,
267
+ "Structural Integrity Change": 68.78,
268
+ "Motion State Change": 69.51,
269
+ "Spatial Arrangement": 77.12,
270
+ "Mechanical Reaction": 67.83,
271
+ "Medium Interaction": 67.25,
272
+ "Game-World-Overall": 44.22,
273
+ "3D Reconstruction": 57.03,
274
+ "Space Invader": 36.58,
275
+ "Jewel2": 37.79,
276
+ "Pacman": 30.02,
277
+ "Word Search": 48.47,
278
+ "Tictactoe": 33.0,
279
+ "Sudoku": 31.63,
280
+ "Maze": 60.02,
281
+ "Sokoban": 34.71
282
+ },
283
+ {
284
+ "model": "UniREdit-Bagel(Ours)",
285
+ "link": "-",
286
+ "hf": "https://huggingface.co/maplebb/UniREdit-Bagel",
287
+ "open_source": true,
288
+ "release_date": "2025-11",
289
+ "Overall": 78.87,
290
+ "Real-World-Overall": 76.52,
291
+ "Viewpoint Transformation": 84.43,
292
+ "Material Modification": 74.06,
293
+ "Pose Adjustment": 85.13,
294
+ "Temporal Evolution": 72.08,
295
+ "Structural Integrity Change": 74.31,
296
+ "Motion State Change": 71.95,
297
+ "Spatial Arrangement": 83.05,
298
+ "Mechanical Reaction": 71.17,
299
+ "Medium Interaction": 72.45,
300
+ "Game-World-Overall": 82.27,
301
+ "3D Reconstruction": 84.93,
302
+ "Space Invader": 86.93,
303
+ "Jewel2": 61.25,
304
+ "Pacman": 73.76,
305
+ "Word Search": 87.48,
306
+ "Tictactoe": 70.65,
307
+ "Sudoku": 93.9,
308
+ "Maze": 97.73,
309
+ "Sokoban": 74.55
310
  }
311
  ]
312
+ }
src/__pycache__/about.cpython-310.pyc ADDED
Binary file (2.08 kB). View file
 
src/__pycache__/json_leaderboard.cpython-310.pyc ADDED
Binary file (3.2 kB). View file
 
src/about.py CHANGED
@@ -21,23 +21,35 @@ NUM_FEWSHOT = 0 # Change with your few shot
21
 
22
 
23
  # Your leaderboard name
24
- TITLE = """<h1 align="center" id="space-title">🥇 <a href="https://github.com/CodeGoat24/UniGenBench" target="_blank">UniGenBench</a> Leaderboard (English)</h1> """
25
 
26
  # Links and conference info
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
27
  LINKS_AND_INFO = """
28
  <div align="center">
29
- <p><a href="https://hunyuan.tencent.com/" target="_blank">Hunyuan</a>, Tencent</p> <br>
30
-
31
- <a href="https://codegoat24.github.io/UniGenBench" target="_blank">🏠 Homepage</a> |
32
- <a href="https://arxiv.org/pdf/2510.18701" target="_blank">📄 arXiv Paper</a> |
33
- <a href="https://huggingface.co/datasets/CodeGoat24/UniGenBench/tree/main">😊 Huggingface</a>
34
 
35
- <a href="https://github.com/CodeGoat24/UniGenBench" target="_blank" rel="noopener noreferrer"><img alt="Code" src="https://img.shields.io/github/stars/CodeGoat24/UniGenBench.svg?style=social&amp;label=Official"></a>
 
36
 
37
- 🏅 <a href="https://huggingface.co/spaces/CodeGoat24/UniGenBench_Leaderboard"><b>Leaderboard</b>(English)</a> |
38
- <a href="https://huggingface.co/spaces/CodeGoat24/UniGenBench_Leaderboard_Chinese"><b>Leaderboard</b>(Chinese)</a> |
39
- <a href="https://huggingface.co/spaces/CodeGoat24/UniGenBench_Leaderboard_English_Long"><b>Leaderboard</b>(English Long)</a> |
40
- <a href="https://huggingface.co/spaces/CodeGoat24/UniGenBench_Leaderboard_Chinese_Long"><b>Leaderboard</b>(Chinese Long)</a> 🏅
41
 
42
 
43
 
@@ -46,13 +58,11 @@ LINKS_AND_INFO = """
46
 
47
  # What does your leaderboard evaluate?
48
  INTRODUCTION_TEXT = """
49
- 📚 [UniGenBench](https://github.com/CodeGoat24/UniGenBench) is a unified benchmark for T2I generation that integrates diverse prompt themes with a comprehensive suite of fine-grained evaluation criteria.
50
 
51
- 🔧 You can use the official [GitHub repo](https://github.com/CodeGoat24/UniGenBench) to evaluate your model on [UniGenBench](https://github.com/CodeGoat24/UniGenBench).
52
 
53
- 😊 We release **all generated images from the T2I models** evaluated in our UniGenBench on [UniGenBench-Eval-Images](https://huggingface.co/datasets/CodeGoat24/UniGenBench-Eval-Images). Feel free to use any evaluation model that is convenient and suitable for you to assess and compare the performance of your models.
54
-
55
- 📝 To add your own model to the leaderboard, please send an Email to [Yibin Wang](https://codegoat24.github.io/), then we will help with the evaluation and updating the leaderboard.
56
 
57
  """
58
 
@@ -60,19 +70,20 @@ INTRODUCTION_TEXT = """
60
 
61
 
62
  CITATION_BUTTON_LABEL = "Copy the following snippet to cite these results"
63
- CITATION_BUTTON_TEXT = r"""
64
- @article{UniGenBench++,
65
- title={UniGenBench++: A Unified Semantic Evaluation Benchmark for Text-to-Image Generation},
66
- author={Wang, Yibin and Li, Zhimin and Zang, Yuhang and Bu, Jiazi and Zhou, Yujie and Xin, Yi and He, Junjun and Wang, Chunyu and Lu, Qinglin and Jin, Cheng and others},
67
- journal={arXiv preprint arXiv:2510.18701},
68
- year={2025}
69
- }
70
-
71
-
72
- @article{UniGenBench&Pref-GRPO,
73
- title={Pref-GRPO: Pairwise Preference Reward-based GRPO for Stable Text-to-Image Reinforcement Learning},
74
- author={Wang, Yibin and Li, Zhimin and Zang, Yuhang and Zhou, Yujie and Bu, Jiazi and Wang, Chunyu and Lu, Qinglin and Jin, Cheng and Wang, Jiaqi},
75
- journal={arXiv preprint arXiv:2508.20751},
76
- year={2025}
77
- }
78
- """
 
 
21
 
22
 
23
  # Your leaderboard name
24
+ TITLE = """<h1 align="center" id="space-title">🥇 <a href="https://github.com/Maplebb/UniREditBench" target="_blank">UniREditBench</a> Leaderboard</h1> """
25
 
26
  # Links and conference info
27
+ # LINKS_AND_INFO = """
28
+ # <div align="center">
29
+ # <p><a href="https://hunyuan.tencent.com/" target="_blank">Hunyuan</a>, Tencent</p> <br>
30
+
31
+ # <a href="https://maplebb.github.io/UniREditBench" target="_blank">🏠 Homepage</a> |
32
+ # <a href="https://arxiv.org/pdf/2510.18701" target="_blank">📄 arXiv Paper</a> |
33
+ # <a href="https://huggingface.co/datasets/CodeGoat24/UniREditBench/tree/main">😊 Huggingface</a>
34
+
35
+ # <a href="https://github.com/Maplebb/UniREditBench" target="_blank" rel="noopener noreferrer"><img alt="Code" src="https://img.shields.io/github/stars/CodeGoat24/UniREditBench.svg?style=social&amp;label=Official"></a>
36
+
37
+ # 🏅 <a href="https://huggingface.co/spaces/CodeGoat24/UniREditBench_Leaderboard"><b>Leaderboard</b>(English)</a> |
38
+ # <a href="https://huggingface.co/spaces/CodeGoat24/UniREditBench_Leaderboard_Chinese"><b>Leaderboard</b>(Chinese)</a> |
39
+ # <a href="https://huggingface.co/spaces/CodeGoat24/UniREditBench_Leaderboard_English_Long"><b>Leaderboard</b>(English Long)</a> |
40
+ # <a href="https://huggingface.co/spaces/CodeGoat24/UniREditBench_Leaderboard_Chinese_Long"><b>Leaderboard</b>(Chinese Long)</a> 🏅
41
+
42
+
43
+
44
+ # </div>
45
+ # """
46
  LINKS_AND_INFO = """
47
  <div align="center">
 
 
 
 
 
48
 
49
+ <a href="https://maplebb.github.io/UniREditBench" target="_blank">🏠 Homepage</a> |
50
+ <a href="https://huggingface.co/datasets/maplebb/UniREditBench/tree/main">😊 Huggingface</a>
51
 
52
+ <a href="https://github.com/Maplebb/UniREditBench" target="_blank" rel="noopener noreferrer"><img alt="Code" src="https://img.shields.io/github/stars/Maplebb/UniREditBench.svg?style=social&amp;label=Official"></a>
 
 
 
53
 
54
 
55
 
 
58
 
59
  # What does your leaderboard evaluate?
60
  INTRODUCTION_TEXT = """
61
+ 📚 [UniREditBench](https://github.com/Maplebb/UniREditBench) is a unified benchmark for reasoning-base image editing that integrates diverse prompt themes with a comprehensive suite of dual-reference evaluation pipeline.
62
 
63
+ 🔧 You can use the official [GitHub repo](https://github.com/Maplebb/UniREditBench) to evaluate your model on [UniREditBench](https://github.com/Maplebb/UniREditBench).
64
 
65
+ 📝 To add your own model to the leaderboard, please send an Email to [Feng Han]([email protected]) and [Yibin Wang](https://codegoat24.github.io/), then we will help with the evaluation and updating the leaderboard.
 
 
66
 
67
  """
68
 
 
70
 
71
 
72
  CITATION_BUTTON_LABEL = "Copy the following snippet to cite these results"
73
+ # CITATION_BUTTON_TEXT = r"""
74
+ # @article{UniREditBench++,
75
+ # title={UniREditBench++: A Unified Semantic Evaluation Benchmark for Text-to-Image Generation},
76
+ # author={Wang, Yibin and Li, Zhimin and Zang, Yuhang and Bu, Jiazi and Zhou, Yujie and Xin, Yi and He, Junjun and Wang, Chunyu and Lu, Qinglin and Jin, Cheng and others},
77
+ # journal={arXiv preprint arXiv:2510.18701},
78
+ # year={2025}
79
+ # }
80
+
81
+
82
+ # @article{UniREditBench&Pref-GRPO,
83
+ # title={Pref-GRPO: Pairwise Preference Reward-based GRPO for Stable Text-to-Image Reinforcement Learning},
84
+ # author={Wang, Yibin and Li, Zhimin and Zang, Yuhang and Zhou, Yujie and Bu, Jiazi and Wang, Chunyu and Lu, Qinglin and Jin, Cheng and Wang, Jiaqi},
85
+ # journal={arXiv preprint arXiv:2508.20751},
86
+ # year={2025}
87
+ # }
88
+ # """
89
+ CITATION_BUTTON_TEXT = r""""""
src/display/__pycache__/css_html_js.cpython-310.pyc ADDED
Binary file (6.89 kB). View file
 
src/json_leaderboard.py CHANGED
@@ -1,7 +1,6 @@
 
1
  import json
2
  import pandas as pd
3
- from pathlib import Path
4
-
5
 
6
  def load_leaderboard_from_json(json_path="leaderboard_data.json"):
7
  """Load leaderboard data from JSON file"""
@@ -16,99 +15,65 @@ def load_leaderboard_from_json(json_path="leaderboard_data.json"):
16
  print(f"Error decoding JSON file {json_path}")
17
  return []
18
 
19
-
20
  def create_leaderboard_df(json_path="leaderboard_data.json"):
21
- """Create a pandas DataFrame from JSON leaderboard data"""
22
  leaderboard_data = load_leaderboard_from_json(json_path)
23
-
24
  if not leaderboard_data:
25
  return pd.DataFrame()
26
-
27
- # Convert to DataFrame
28
- df = pd.DataFrame(leaderboard_data)
29
-
30
- # Sort by ACC score (descending)
31
- df = df.sort_values('Overall', ascending=False).reset_index(drop=True)
32
-
33
- # Add ranking icons and make model names clickable links to papers
34
- def add_ranking_icon_and_link(index, model_name, paper_link):
35
- if index == 0:
36
- return f'🥇 <a href="{paper_link}" target="_blank">{model_name}</a>'
37
- elif index == 1:
38
- return f'🥈 <a href="{paper_link}" target="_blank">{model_name}</a>'
39
- elif index == 2:
40
- return f'🥉 <a href="{paper_link}" target="_blank">{model_name}</a>'
41
- else:
42
- return f'<a href="{paper_link}" target="_blank">{model_name}</a>'
43
-
44
- # Format the DataFrame for display
45
- display_df = pd.DataFrame({
46
- 'Model Name (clickable)': [add_ranking_icon_and_link(i, model, link) for i, (model, link) in enumerate(zip(df['model'], df['link']))],
47
- 'Release Date': df['release_date'],
48
- 'HF Model': df['hf'].apply(lambda x: f'<a href="{x}" target="_blank">🤗</a>' if x != "-" else "-"),
49
- 'Open Source': df['open_source'].apply(lambda x: '✓' if x else '✗'),
50
 
51
- 'Overall': df['Overall'].apply(lambda x: f"{x:.2f}"),
52
-
53
- 'Style': df['Style'].apply(lambda x: f"{x:.2f}"),
54
 
55
- 'World Knowledge': df['World Knowledge'].apply(lambda x: f"{x:.2f}"),
56
-
57
- 'Logical Reasoning': df['Logical Reasoning'].apply(lambda x: f"{x:.2f}"),
 
 
58
 
59
- 'Text': df['Text'].apply(lambda x: f"{x:.2f}"),
 
60
 
61
- 'Attribute-Overall': df['Attribute-Overall'].apply(lambda x: f"{x:.2f}"),
62
- 'Quantity': df['Quantity'].apply(lambda x: f"{x:.2f}"),
63
- 'Expression': df['Expression'].apply(lambda x: f"{x:.2f}"),
64
- 'Material': df['Material'].apply(lambda x: f"{x:.2f}"),
65
- 'Size': df['Size'].apply(lambda x: f"{x:.2f}"),
66
- 'Shape': df['Shape'].apply(lambda x: f"{x:.2f}"),
67
- 'Color': df['Color'].apply(lambda x: f"{x:.2f}"),
68
 
69
- 'Action-Overall': df['Action-Overall'].apply(lambda x: f"{x:.2f}"),
70
- 'Hand': df['Hand'].apply(lambda x: f"{x:.2f}"),
71
- 'Full body': df['Full body'].apply(lambda x: f"{x:.2f}"),
72
- 'Animal': df['Animal'].apply(lambda x: f"{x:.2f}"),
73
- 'Non Contact': df['Non Contact'].apply(lambda x: f"{x:.2f}"),
74
- 'Contact': df['Contact'].apply(lambda x: f"{x:.2f}"),
75
- 'State': df['State'].apply(lambda x: f"{x:.2f}"),
76
-
77
- 'Relationship-Overall': df['Relationship-Overall'].apply(lambda x: f"{x:.2f}"),
78
- 'Composition': df['Composition'].apply(lambda x: f"{x:.2f}"),
79
- 'Similarity': df['Similarity'].apply(lambda x: f"{x:.2f}"),
80
- 'Inclusion': df['Inclusion'].apply(lambda x: f"{x:.2f}"),
81
- 'Comparison': df['Comparison'].apply(lambda x: f"{x:.2f}"),
82
 
83
- 'Compound-Overall': df['Compound-Overall'].apply(lambda x: f"{x:.2f}"),
84
- 'Imagination': df['Imagination'].apply(lambda x: f"{x:.2f}"),
85
- 'Feature matching': df['Feature matching'].apply(lambda x: f"{x:.2f}"),
 
 
 
 
86
 
87
- 'Grammar-Overall': df['Grammar-Overall'].apply(lambda x: f"{x:.2f}"),
88
- 'Pronoun Reference': df['Pronoun Reference'].apply(lambda x: f"{x:.2f}"),
89
- 'Consistency': df['Consistency'].apply(lambda x: f"{x:.2f}"),
90
- 'Negation': df['Negation'].apply(lambda x: f"{x:.2f}"),
91
 
92
- 'Layout-Overall': df['Layout-Overall'].apply(lambda x: f"{x:.2f}"),
93
- 'Two-dimensional': df['2D'].apply(lambda x: f"{x:.2f}"),
94
- 'Three-dimensional': df['3D'].apply(lambda x: f"{x:.2f}"),
95
- })
96
-
97
  return display_df
98
 
99
-
100
  def get_leaderboard_stats(json_path="leaderboard_data.json"):
101
  """Get statistics about the leaderboard"""
102
  leaderboard_data = load_leaderboard_from_json(json_path)
103
-
104
  if not leaderboard_data:
105
  return {}
106
-
107
  df = pd.DataFrame(leaderboard_data)
108
-
109
  stats = {
110
  'total_models': len(df),
111
- 'open_source_models': df['open_source'].sum(),
112
  }
113
-
114
  return stats
 
1
+
2
  import json
3
  import pandas as pd
 
 
4
 
5
  def load_leaderboard_from_json(json_path="leaderboard_data.json"):
6
  """Load leaderboard data from JSON file"""
 
15
  print(f"Error decoding JSON file {json_path}")
16
  return []
17
 
 
18
  def create_leaderboard_df(json_path="leaderboard_data.json"):
19
+ """Create a pandas DataFrame from JSON leaderboard data (dynamic columns)."""
20
  leaderboard_data = load_leaderboard_from_json(json_path)
 
21
  if not leaderboard_data:
22
  return pd.DataFrame()
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
23
 
24
+ df = pd.DataFrame(leaderboard_data)
 
 
25
 
26
+ # Sort by Overall if present
27
+ if 'Overall' in df.columns:
28
+ df = df.sort_values('Overall', ascending=False).reset_index(drop=True)
29
+ else:
30
+ df = df.reset_index(drop=True)
31
 
32
+ meta_cols = ['model', 'link', 'hf', 'open_source', 'release_date']
33
+ score_cols = [c for c in df.columns if c not in meta_cols]
34
 
35
+ # Keep "Overall" first if present, then the rest in file order
36
+ ordered_score_cols = []
37
+ if 'Overall' in score_cols:
38
+ ordered_score_cols.append('Overall')
39
+ ordered_score_cols += [c for c in score_cols if c != 'Overall']
 
 
40
 
41
+ def add_ranking_icon_and_link(idx, row):
42
+ model_name = row.get('model', 'N/A')
43
+ paper_link = row.get('link', '-') or '-'
44
+ anchor = f'<a href="{paper_link}" target="_blank">{model_name}</a>' if paper_link and paper_link != '-' else model_name
45
+ if idx == 0:
46
+ return f'🥇 {anchor}'
47
+ elif idx == 1:
48
+ return f'🥈 {anchor}'
49
+ elif idx == 2:
50
+ return f'🥉 {anchor}'
51
+ else:
52
+ return anchor
 
53
 
54
+ # Base display columns
55
+ display_df = pd.DataFrame({
56
+ 'Model Name (clickable)': [add_ranking_icon_and_link(i, r) for i, r in df.iterrows()],
57
+ 'Release Date': df.get('release_date', '-'),
58
+ 'HF Model': df.get('hf', '-').apply(lambda x: f'<a href="{x}" target="_blank">🤗</a>' if isinstance(x, str) and x != "-" else "-"),
59
+ 'Open Source': df.get('open_source', False).apply(lambda x: '✓' if bool(x) else '✗'),
60
+ })
61
 
62
+ # Append each score column with 2-decimal formatting
63
+ for col in ordered_score_cols:
64
+ col_vals = pd.to_numeric(df[col], errors='coerce')
65
+ display_df[col] = col_vals.map(lambda x: f"{x:.2f}" if pd.notnull(x) else "-")
66
 
 
 
 
 
 
67
  return display_df
68
 
 
69
  def get_leaderboard_stats(json_path="leaderboard_data.json"):
70
  """Get statistics about the leaderboard"""
71
  leaderboard_data = load_leaderboard_from_json(json_path)
 
72
  if not leaderboard_data:
73
  return {}
 
74
  df = pd.DataFrame(leaderboard_data)
 
75
  stats = {
76
  'total_models': len(df),
77
+ 'open_source_models': int(df['open_source'].sum()) if 'open_source' in df.columns else 0,
78
  }
 
79
  return stats