Update leaderboard_data.json
Browse files- leaderboard_data.json +410 -0
leaderboard_data.json
CHANGED
@@ -1,5 +1,415 @@
|
|
1 |
{
|
2 |
"leaderboard": [
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
{
|
4 |
"model": "GPT-4o",
|
5 |
"link": "https://platform.openai.com/docs/guides/image-generation",
|
|
|
1 |
{
|
2 |
"leaderboard": [
|
3 |
+
{
|
4 |
+
"model": "BLIP-3o",
|
5 |
+
"link": "https://arxiv.org/pdf/2505.09568",
|
6 |
+
"hf": "https://huggingface.co/BLIP3o/BLIP3o-Model-8B",
|
7 |
+
"open_source": true,
|
8 |
+
"release_date": "2025-5",
|
9 |
+
|
10 |
+
"Overall": 59.87,
|
11 |
+
"Style": 92.80,
|
12 |
+
"World Knowledge": 80.22,
|
13 |
+
"Attribute-Overall": 63.89,
|
14 |
+
"Quantity": 51.39,
|
15 |
+
"Expression": 60.26,
|
16 |
+
"Material": 64.62,
|
17 |
+
"Size": 75.00,
|
18 |
+
"Shape": 54.37,
|
19 |
+
"Color": 81.67,
|
20 |
+
|
21 |
+
"Action-Overall": 63.97,
|
22 |
+
"Hand": 58.33,
|
23 |
+
"Full body": 70.11,
|
24 |
+
"Animal": 70.59,
|
25 |
+
"Non Contact": 60.20,
|
26 |
+
"Contact": 51.79,
|
27 |
+
"State": 71.70,
|
28 |
+
|
29 |
+
"Relationship-Overall": 66.50,
|
30 |
+
"Composition": 70.61,
|
31 |
+
"Similarity": 60.00,
|
32 |
+
"Inclusion": 67.39,
|
33 |
+
"Comparison": 64.84,
|
34 |
+
|
35 |
+
"Compound-Overall": 53.74,
|
36 |
+
"Imagination": 61.73,
|
37 |
+
"Feature matching": 45.57,
|
38 |
+
|
39 |
+
"Grammar-Overall": 68.58,
|
40 |
+
"Pronoun Reference": 79.04,
|
41 |
+
"Consistency": 61.11,
|
42 |
+
"Negation": 63.85,
|
43 |
+
|
44 |
+
"Layout-Overall": 68.47,
|
45 |
+
"2D": 72.79,
|
46 |
+
"3D": 64.02,
|
47 |
+
|
48 |
+
"Logical Reasoning": 39.55,
|
49 |
+
|
50 |
+
"Text": 1.15
|
51 |
+
},
|
52 |
+
{
|
53 |
+
"model": "CogVew4",
|
54 |
+
"link": "https://arxiv.org/pdf/2403.05121",
|
55 |
+
"hf": "https://huggingface.co/zai-org/CogView4-6B",
|
56 |
+
"open_source": true,
|
57 |
+
"release_date": "2024-3",
|
58 |
+
|
59 |
+
"Overall": 56.30,
|
60 |
+
"Style": 82.00,
|
61 |
+
"World Knowledge": 83.07,
|
62 |
+
"Attribute-Overall": 63.25,
|
63 |
+
"Quantity": 71.53,
|
64 |
+
"Expression": 44.23,
|
65 |
+
"Material": 55.19,
|
66 |
+
"Size": 72.22,
|
67 |
+
"Shape": 57.50,
|
68 |
+
"Color": 89.17,
|
69 |
+
|
70 |
+
"Action-Overall": 57.51,
|
71 |
+
"Hand": 53.85,
|
72 |
+
"Full body": 59.78,
|
73 |
+
"Animal": 68.38,
|
74 |
+
"Non Contact": 50.51,
|
75 |
+
"Contact": 51.19,
|
76 |
+
"State": 62.74,
|
77 |
+
|
78 |
+
"Relationship-Overall": 62.44,
|
79 |
+
"Composition": 60.47,
|
80 |
+
"Similarity": 60.00,
|
81 |
+
"Inclusion": 69.57,
|
82 |
+
"Comparison": 60.16,
|
83 |
+
|
84 |
+
"Compound-Overall": 44.72,
|
85 |
+
"Imagination": 47.19,
|
86 |
+
"Feature matching": 42.19,
|
87 |
+
|
88 |
+
"Grammar-Overall": 54.81,
|
89 |
+
"Pronoun Reference": 69.49,
|
90 |
+
"Consistency": 56.02,
|
91 |
+
"Negation": 38.46,
|
92 |
+
|
93 |
+
"Layout-Overall": 69.22,
|
94 |
+
"2D": 77.21,
|
95 |
+
"3D": 60.98,
|
96 |
+
|
97 |
+
"Logical Reasoning": 28.18,
|
98 |
+
|
99 |
+
"Text": 17.82
|
100 |
+
},
|
101 |
+
{
|
102 |
+
"model": "Hunyuan-DiT",
|
103 |
+
"link": "https://arxiv.org/pdf/2405.08748",
|
104 |
+
"hf": "https://huggingface.co/Tencent-Hunyuan/HunyuanDiT",
|
105 |
+
"open_source": true,
|
106 |
+
"release_date": "2024-5",
|
107 |
+
|
108 |
+
"Overall": 51.38,
|
109 |
+
|
110 |
+
"Style": 94.10,
|
111 |
+
|
112 |
+
"World Knowledge": 80.70,
|
113 |
+
|
114 |
+
"Attribute-Overall": 62.71,
|
115 |
+
"Quantity": 67.36,
|
116 |
+
"Expression": 44.23,
|
117 |
+
"Material": 71.70,
|
118 |
+
"Size": 61.81,
|
119 |
+
"Shape": 47.50,
|
120 |
+
"Color": 86.67,
|
121 |
+
|
122 |
+
"Action-Overall": 49.05,
|
123 |
+
"Hand": 35.90,
|
124 |
+
"Full body": 54.89,
|
125 |
+
"Animal": 54.41,
|
126 |
+
"Non Contact": 46.94,
|
127 |
+
"Contact": 35.71,
|
128 |
+
"State": 62.74,
|
129 |
+
|
130 |
+
"Relationship-Overall": 59.64,
|
131 |
+
"Composition": 60.14,
|
132 |
+
"Similarity": 64.44,
|
133 |
+
"Inclusion": 60.33,
|
134 |
+
"Comparison": 50.78,
|
135 |
+
|
136 |
+
"Compound-Overall": 41.62,
|
137 |
+
"Imagination": 46.68,
|
138 |
+
"Feature matching": 36.46,
|
139 |
+
|
140 |
+
"Grammar-Overall": 55.48,
|
141 |
+
"Pronoun Reference": 62.87,
|
142 |
+
"Consistency": 57.87,
|
143 |
+
"Negation": 45.77,
|
144 |
+
|
145 |
+
"Layout-Overall": 44.78,
|
146 |
+
"2D": 39.34,
|
147 |
+
"3D": 50.38,
|
148 |
+
|
149 |
+
"Logical Reasoning": 24.55,
|
150 |
+
|
151 |
+
"Text": 1.15
|
152 |
+
},
|
153 |
+
{
|
154 |
+
"model": "Janus",
|
155 |
+
"link": "https://arxiv.org/pdf/2410.13848",
|
156 |
+
"hf": "https://huggingface.co/deepseek-ai/Janus-1.3B",
|
157 |
+
"open_source": true,
|
158 |
+
"release_date": "2024-10",
|
159 |
+
|
160 |
+
"Overall": 51.23,
|
161 |
+
|
162 |
+
"Style": 89.90,
|
163 |
+
|
164 |
+
"World Knowledge": 73.58,
|
165 |
+
|
166 |
+
"Attribute-Overall": 54.81,
|
167 |
+
"Quantity": 37.50,
|
168 |
+
"Expression": 37.82,
|
169 |
+
"Material": 58.96,
|
170 |
+
"Size": 65.97,
|
171 |
+
"Shape": 47.50,
|
172 |
+
"Color": 86.67,
|
173 |
+
|
174 |
+
"Action-Overall": 50.38,
|
175 |
+
"Hand": 32.69,
|
176 |
+
"Full body": 51.63,
|
177 |
+
"Animal": 61.76,
|
178 |
+
"Non Contact": 48.47,
|
179 |
+
"Contact": 38.10,
|
180 |
+
"State": 66.51,
|
181 |
+
|
182 |
+
"Relationship-Overall": 55.08,
|
183 |
+
"Composition": 56.76,
|
184 |
+
"Similarity": 53.89,
|
185 |
+
"Inclusion": 59.24,
|
186 |
+
"Comparison": 46.88,
|
187 |
+
|
188 |
+
"Compound-Overall": 46.65,
|
189 |
+
"Imagination": 58.16,
|
190 |
+
"Feature matching": 34.90,
|
191 |
+
|
192 |
+
"Grammar-Overall": 59.09,
|
193 |
+
"Pronoun Reference": 66.18,
|
194 |
+
"Consistency": 51.39,
|
195 |
+
"Negation": 58.08,
|
196 |
+
|
197 |
+
"Layout-Overall": 54.85,
|
198 |
+
"2D": 57.72,
|
199 |
+
"3D": 51.89,
|
200 |
+
|
201 |
+
"Logical Reasoning": 26.82,
|
202 |
+
|
203 |
+
"Text": 1.15
|
204 |
+
},
|
205 |
+
{
|
206 |
+
"model": "Janus-flow",
|
207 |
+
"link": "https://arxiv.org/pdf/2411.07975",
|
208 |
+
"hf": "https://huggingface.co/deepseek-ai/JanusFlow-1.3B",
|
209 |
+
"open_source": true,
|
210 |
+
"release_date": "2024-11",
|
211 |
+
|
212 |
+
"Overall": 46.39,
|
213 |
+
|
214 |
+
"Style": 86.20,
|
215 |
+
|
216 |
+
"World Knowledge": 62.50,
|
217 |
+
|
218 |
+
"Attribute-Overall": 47.97,
|
219 |
+
"Quantity": 43.06,
|
220 |
+
"Expression": 30.77,
|
221 |
+
"Material": 55.19,
|
222 |
+
"Size": 55.56,
|
223 |
+
"Shape": 30.00,
|
224 |
+
"Color": 78.33,
|
225 |
+
|
226 |
+
"Action-Overall": 43.35,
|
227 |
+
"Hand": 23.08,
|
228 |
+
"Full body": 48.37,
|
229 |
+
"Animal": 58.82,
|
230 |
+
"Non Contact": 36.73,
|
231 |
+
"Contact": 36.31,
|
232 |
+
"State": 55.66,
|
233 |
+
|
234 |
+
"Relationship-Overall": 50.00,
|
235 |
+
"Composition": 59.80,
|
236 |
+
"Similarity": 38.89,
|
237 |
+
"Inclusion": 51.63,
|
238 |
+
"Comparison": 40.62,
|
239 |
+
|
240 |
+
"Compound-Overall": 45.10,
|
241 |
+
"Imagination": 57.65,
|
242 |
+
"Feature matching": 32.29,
|
243 |
+
|
244 |
+
"Grammar-Overall": 60.29,
|
245 |
+
"Pronoun Reference": 66.18,
|
246 |
+
"Consistency": 48.61,
|
247 |
+
"Negation": 63.85,
|
248 |
+
|
249 |
+
"Layout-Overall": 46.46,
|
250 |
+
"2D": 49.26,
|
251 |
+
"3D": 43.56,
|
252 |
+
|
253 |
+
"Logical Reasoning": 21.14,
|
254 |
+
|
255 |
+
"Text": 0.86
|
256 |
+
},
|
257 |
+
{
|
258 |
+
"model": "Emu-3",
|
259 |
+
"link": "https://arxiv.org/pdf/2409.18869",
|
260 |
+
"hf": "https://huggingface.co/BAAI/Emu3-Gen",
|
261 |
+
"open_source": true,
|
262 |
+
"release_date": "2024-09",
|
263 |
+
|
264 |
+
"Overall": 46.02,
|
265 |
+
|
266 |
+
"Style": 86.80,
|
267 |
+
|
268 |
+
"World Knowledge": 77.06,
|
269 |
+
|
270 |
+
"Attribute-Overall": 51.39,
|
271 |
+
"Quantity": 44.44,
|
272 |
+
"Expression": 45.51,
|
273 |
+
"Material": 53.77,
|
274 |
+
"Size": 43.06,
|
275 |
+
"Shape": 46.25,
|
276 |
+
"Color": 80.00,
|
277 |
+
|
278 |
+
"Action-Overall": 40.11,
|
279 |
+
"Hand": 25.00,
|
280 |
+
"Full body": 47.28,
|
281 |
+
"Animal": 50.74,
|
282 |
+
"Non Contact": 35.20,
|
283 |
+
"Contact": 27.98,
|
284 |
+
"State": 52.36,
|
285 |
+
|
286 |
+
"Relationship-Overall": 49.75,
|
287 |
+
"Composition": 56.76,
|
288 |
+
"Similarity": 46.67,
|
289 |
+
"Inclusion": 48.37,
|
290 |
+
"Comparison": 39.84,
|
291 |
+
|
292 |
+
"Compound-Overall": 36.86,
|
293 |
+
"Imagination": 41.33,
|
294 |
+
"Feature matching": 32.29,
|
295 |
+
|
296 |
+
"Grammar-Overall": 52.94,
|
297 |
+
"Pronoun Reference": 59.56,
|
298 |
+
"Consistency": 53.70,
|
299 |
+
"Negation": 45.38,
|
300 |
+
|
301 |
+
"Layout-Overall": 44.78,
|
302 |
+
"2D": 45.22,
|
303 |
+
"3D": 44.32,
|
304 |
+
|
305 |
+
"Logical Reasoning": 19.32,
|
306 |
+
|
307 |
+
"Text": 1.15
|
308 |
+
},
|
309 |
+
{
|
310 |
+
"model": "Playground2.5",
|
311 |
+
"link": "https://arxiv.org/pdf/2402.17245",
|
312 |
+
"hf": "https://huggingface.co/playgroundai/playground-v2-512px-base",
|
313 |
+
"open_source": true,
|
314 |
+
"release_date": "2024-02",
|
315 |
+
|
316 |
+
"Overall": 45.61,
|
317 |
+
|
318 |
+
"Style": 89.50,
|
319 |
+
|
320 |
+
"World Knowledge": 76.11,
|
321 |
+
|
322 |
+
"Attribute-Overall": 52.78,
|
323 |
+
"Quantity": 58.33,
|
324 |
+
"Expression": 43.59,
|
325 |
+
"Material": 57.08,
|
326 |
+
"Size": 44.44,
|
327 |
+
"Shape": 41.25,
|
328 |
+
"Color": 75.83,
|
329 |
+
|
330 |
+
"Action-Overall": 42.68,
|
331 |
+
"Hand": 28.85,
|
332 |
+
"Full body": 50.00,
|
333 |
+
"Animal": 52.21,
|
334 |
+
"Non Contact": 35.20,
|
335 |
+
"Contact": 29.17,
|
336 |
+
"State": 58.02,
|
337 |
+
|
338 |
+
"Relationship-Overall": 51.52,
|
339 |
+
"Composition": 60.14,
|
340 |
+
"Similarity": 49.44,
|
341 |
+
"Inclusion": 48.37,
|
342 |
+
"Comparison": 39.06,
|
343 |
+
|
344 |
+
"Compound-Overall": 35.44,
|
345 |
+
"Imagination": 43.88,
|
346 |
+
"Feature matching": 26.82,
|
347 |
+
|
348 |
+
"Grammar-Overall": 53.21,
|
349 |
+
"Pronoun Reference": 58.82,
|
350 |
+
"Consistency": 50.00,
|
351 |
+
"Negation": 50.00,
|
352 |
+
|
353 |
+
"Layout-Overall": 37.13,
|
354 |
+
"2D": 34.56,
|
355 |
+
"3D": 39.77,
|
356 |
+
|
357 |
+
"Logical Reasoning": 16.59,
|
358 |
+
|
359 |
+
"Text": 1.15
|
360 |
+
},
|
361 |
+
{
|
362 |
+
"model": "SDXL",
|
363 |
+
"link": "https://arxiv.org/pdf/2307.01952",
|
364 |
+
"hf": "https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0",
|
365 |
+
"open_source": true,
|
366 |
+
"release_date": "2023-07",
|
367 |
+
|
368 |
+
"Overall": 39.75,
|
369 |
+
|
370 |
+
"Style": 87.40,
|
371 |
+
|
372 |
+
"World Knowledge": 72.63,
|
373 |
+
|
374 |
+
"Attribute-Overall": 44.34,
|
375 |
+
"Quantity": 44.44,
|
376 |
+
"Expression": 25.00,
|
377 |
+
"Material": 52.83,
|
378 |
+
"Size": 44.44,
|
379 |
+
"Shape": 33.75,
|
380 |
+
"Color": 68.33,
|
381 |
+
|
382 |
+
"Action-Overall": 34.22,
|
383 |
+
"Hand": 19.23,
|
384 |
+
"Full body": 35.33,
|
385 |
+
"Animal": 43.38,
|
386 |
+
"Non Contact": 26.53,
|
387 |
+
"Contact": 24.40,
|
388 |
+
"State": 53.30,
|
389 |
+
|
390 |
+
"Relationship-Overall": 44.92,
|
391 |
+
"Composition": 53.72,
|
392 |
+
"Similarity": 38.33,
|
393 |
+
"Inclusion": 39.67,
|
394 |
+
"Comparison": 41.41,
|
395 |
+
|
396 |
+
"Compound-Overall": 26.68,
|
397 |
+
"Imagination": 33.93,
|
398 |
+
"Feature matching": 19.27,
|
399 |
+
|
400 |
+
"Grammar-Overall": 47.33,
|
401 |
+
"Pronoun Reference": 50.37,
|
402 |
+
"Consistency": 42.59,
|
403 |
+
"Negation": 48.08,
|
404 |
+
|
405 |
+
"Layout-Overall": 29.85,
|
406 |
+
"2D": 26.47,
|
407 |
+
"3D": 33.33,
|
408 |
+
|
409 |
+
"Logical Reasoning": 9.55,
|
410 |
+
|
411 |
+
"Text": 1.15
|
412 |
+
},
|
413 |
{
|
414 |
"model": "GPT-4o",
|
415 |
"link": "https://platform.openai.com/docs/guides/image-generation",
|