Update leaderboard_data.json
Browse files- leaderboard_data.json +301 -4
leaderboard_data.json
CHANGED
@@ -1,11 +1,308 @@
|
|
1 |
{
|
2 |
"leaderboard": [
|
3 |
{
|
4 |
-
"model": "
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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-
|
9 |
|
10 |
"Overall": 59.87,
|
11 |
"Style": 92.80,
|
@@ -54,7 +351,7 @@
|
|
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-
|
58 |
|
59 |
"Overall": 56.30,
|
60 |
"Style": 82.00,
|
@@ -103,7 +400,7 @@
|
|
103 |
"link": "https://arxiv.org/pdf/2405.08748",
|
104 |
"hf": "https://huggingface.co/Tencent-Hunyuan/HunyuanDiT",
|
105 |
"open_source": true,
|
106 |
-
"release_date": "2024-
|
107 |
|
108 |
"Overall": 51.38,
|
109 |
|
|
|
1 |
{
|
2 |
"leaderboard": [
|
3 |
{
|
4 |
+
"model": "HiDream",
|
5 |
+
"link": "https://arxiv.org/pdf/2505.22705",
|
6 |
+
"hf": "https://huggingface.co/HiDream-ai/HiDream-I1-Dev",
|
7 |
+
"open_source": true,
|
8 |
+
"release_date": "2025-05",
|
9 |
+
|
10 |
+
"Overall": 71.81,
|
11 |
+
"Style": 92.50,
|
12 |
+
"World Knowledge": 94.15,
|
13 |
+
|
14 |
+
"Attribute-Overall": 72.97,
|
15 |
+
"Quantity": 73.61,
|
16 |
+
"Expression": 59.62,
|
17 |
+
"Material": 72.17,
|
18 |
+
"Size": 79.17,
|
19 |
+
"Shape": 61.88,
|
20 |
+
"Color": 98.33,
|
21 |
+
|
22 |
+
"Action-Overall": 73.00,
|
23 |
+
"Hand": 62.18,
|
24 |
+
"Full body": 76.09,
|
25 |
+
"Animal": 73.53,
|
26 |
+
"Non Contact": 74.49,
|
27 |
+
"Contact": 70.24,
|
28 |
+
"State": 78.77,
|
29 |
+
|
30 |
+
"Relationship-Overall": 75.38,
|
31 |
+
"Composition": 79.05,
|
32 |
+
"Similarity": 68.33,
|
33 |
+
"Inclusion": 78.26,
|
34 |
+
"Comparison": 72.66,
|
35 |
+
|
36 |
+
"Compound-Overall": 62.63,
|
37 |
+
"Imagination": 64.29,
|
38 |
+
"Feature matching": 60.94,
|
39 |
+
|
40 |
+
"Grammar-Overall": 63.24,
|
41 |
+
"Pronoun Reference": 83.09,
|
42 |
+
"Consistency": 65.74,
|
43 |
+
"Negation": 40.38,
|
44 |
+
|
45 |
+
"Layout-Overall": 78.17,
|
46 |
+
"2D": 82.72,
|
47 |
+
"3D": 73.48,
|
48 |
+
|
49 |
+
"Logical Reasoning": 41.14,
|
50 |
+
|
51 |
+
"Text": 64.94
|
52 |
+
},
|
53 |
+
{
|
54 |
+
"model": "Pref-GRPO",
|
55 |
+
"link": "https://github.com/CodeGoat24/UnifiedReward",
|
56 |
+
"hf": "https://huggingface.co/CodeGoat24/FLUX.1-dev-PrefGRPO",
|
57 |
+
"open_source": true,
|
58 |
+
"release_date": "2025-08",
|
59 |
+
|
60 |
+
"Overall": 69.46,
|
61 |
+
"Style": 88.40,
|
62 |
+
"World Knowledge": 90.35,
|
63 |
+
|
64 |
+
"Attribute-Overall": 75.00,
|
65 |
+
"Quantity": 71.53,
|
66 |
+
"Expression": 60.90,
|
67 |
+
"Material": 73.11,
|
68 |
+
"Size": 77.08,
|
69 |
+
"Shape": 74.38,
|
70 |
+
"Color": 99.17,
|
71 |
+
|
72 |
+
"Action-Overall": 69.77,
|
73 |
+
"Hand": 60.90,
|
74 |
+
"Full body": 72.28,
|
75 |
+
"Animal": 77.21,
|
76 |
+
"Non Contact": 68.37,
|
77 |
+
"Contact": 64.88,
|
78 |
+
"State": 74.53,
|
79 |
+
|
80 |
+
"Relationship-Overall": 76.52,
|
81 |
+
"Composition": 81.42,
|
82 |
+
"Similarity": 76.67,
|
83 |
+
"Inclusion": 76.09,
|
84 |
+
"Comparison": 65.62,
|
85 |
+
|
86 |
+
"Compound-Overall": 63.27,
|
87 |
+
"Imagination": 65.56,
|
88 |
+
"Feature matching": 60.94,
|
89 |
+
|
90 |
+
"Grammar-Overall": 62.43,
|
91 |
+
"Pronoun Reference": 79.04,
|
92 |
+
"Consistency": 66.20,
|
93 |
+
"Negation": 41.92,
|
94 |
+
|
95 |
+
"Layout-Overall": 77.61,
|
96 |
+
"2D": 82.35,
|
97 |
+
"3D": 72.73,
|
98 |
+
|
99 |
+
"Logical Reasoning": 47.13,
|
100 |
+
|
101 |
+
"Text": 47.13
|
102 |
+
},
|
103 |
+
{
|
104 |
+
"model": "SD-3.5-Large",
|
105 |
+
"link": "https://stability.ai/news/introducing-stable-diffusion-3-5",
|
106 |
+
"hf": "https://huggingface.co/stabilityai/stable-diffusion-3.5-large",
|
107 |
+
"open_source": true,
|
108 |
+
"release_date": "2024-10",
|
109 |
+
|
110 |
+
"Overall": 62.99,
|
111 |
+
"Style": 88.60,
|
112 |
+
"World Knowledge": 88.92,
|
113 |
+
|
114 |
+
"Attribute-Overall": 68.59,
|
115 |
+
"Quantity": 71.53,
|
116 |
+
"Expression": 51.92,
|
117 |
+
"Material": 68.87,
|
118 |
+
"Size": 68.06,
|
119 |
+
"Shape": 65.62,
|
120 |
+
"Color": 90.83,
|
121 |
+
|
122 |
+
"Action-Overall": 62.17,
|
123 |
+
"Hand": 57.05,
|
124 |
+
"Full body": 61.96,
|
125 |
+
"Animal": 63.24,
|
126 |
+
"Non Contact": 62.24,
|
127 |
+
"Contact": 59.52,
|
128 |
+
"State": 67.45,
|
129 |
+
|
130 |
+
"Relationship-Overall": 69.80,
|
131 |
+
"Composition": 75.34,
|
132 |
+
"Similarity": 68.33,
|
133 |
+
"Inclusion": 68.48,
|
134 |
+
"Comparison": 60.94,
|
135 |
+
|
136 |
+
"Compound-Overall": 58.76,
|
137 |
+
"Imagination": 64.80,
|
138 |
+
"Feature matching": 52.60,
|
139 |
+
|
140 |
+
"Grammar-Overall": 58.96,
|
141 |
+
"Pronoun Reference": 74.63,
|
142 |
+
"Consistency": 61.11,
|
143 |
+
"Negation": 40.77,
|
144 |
+
|
145 |
+
"Layout-Overall": 69.03,
|
146 |
+
"2D": 70.96,
|
147 |
+
"3D": 67.05,
|
148 |
+
|
149 |
+
"Logical Reasoning": 32.27,
|
150 |
+
|
151 |
+
"Text": 32.76
|
152 |
+
},
|
153 |
+
{
|
154 |
+
"model": "Janus-Pro",
|
155 |
+
"link": "https://arxiv.org/pdf/2501.17811",
|
156 |
+
"hf": "https://huggingface.co/deepseek-ai/Janus-Pro-7B",
|
157 |
+
"open_source": true,
|
158 |
+
"release_date": "2025-01",
|
159 |
+
|
160 |
+
"Overall": 61.61,
|
161 |
+
"Style": 90.80,
|
162 |
+
"World Knowledge": 86.71,
|
163 |
+
"Attribute-Overall": 67.74,
|
164 |
+
"Quantity": 56.25,
|
165 |
+
"Expression": 55.77,
|
166 |
+
"Material": 71.70,
|
167 |
+
"Size": 73.61,
|
168 |
+
"Shape": 61.88,
|
169 |
+
"Color": 90.83,
|
170 |
+
|
171 |
+
"Action-Overall": 64.26,
|
172 |
+
"Hand": 50.64,
|
173 |
+
"Full body": 63.04,
|
174 |
+
"Animal": 75.00,
|
175 |
+
"Non Contact": 62.24,
|
176 |
+
"Contact": 56.55,
|
177 |
+
"State": 76.42,
|
178 |
+
|
179 |
+
"Relationship-Overall": 68.40,
|
180 |
+
"Composition": 76.01,
|
181 |
+
"Similarity": 56.11,
|
182 |
+
"Inclusion": 75.00,
|
183 |
+
"Comparison": 58.59,
|
184 |
+
|
185 |
+
"Compound-Overall": 62.11,
|
186 |
+
"Imagination": 69.64,
|
187 |
+
"Feature matching": 54.43,
|
188 |
+
|
189 |
+
"Grammar-Overall": 64.44,
|
190 |
+
"Pronoun Reference": 75.37,
|
191 |
+
"Consistency": 66.20,
|
192 |
+
"Negation": 51.54,
|
193 |
+
|
194 |
+
"Layout-Overall": 72.01,
|
195 |
+
"2D": 74.63,
|
196 |
+
"3D": 69.32,
|
197 |
+
|
198 |
+
"Logical Reasoning": 37.05,
|
199 |
+
|
200 |
+
"Text": 2.59
|
201 |
+
},
|
202 |
+
{
|
203 |
+
"model": "Show-o2",
|
204 |
+
"link": "https://arxiv.org/pdf/2506.15564",
|
205 |
+
"hf": "https://huggingface.co/showlab/show-o2-7B",
|
206 |
+
"open_source": true,
|
207 |
+
"release_date": "2025-06",
|
208 |
+
|
209 |
+
"Overall": 62.73,
|
210 |
+
"Style": 87.20,
|
211 |
+
"World Knowledge": 86.08,
|
212 |
+
"Attribute-Overall": 70.51,
|
213 |
+
"Quantity": 59.03,
|
214 |
+
"Expression": 63.46,
|
215 |
+
"Material": 73.58,
|
216 |
+
"Size": 72.92,
|
217 |
+
"Shape": 63.12,
|
218 |
+
"Color": 95.00,
|
219 |
+
|
220 |
+
"Action-Overall": 69.58,
|
221 |
+
"Hand": 56.41,
|
222 |
+
"Full body": 77.72,
|
223 |
+
"Animal": 72.79,
|
224 |
+
"Non Contact": 70.41,
|
225 |
+
"Contact": 52.38,
|
226 |
+
"State": 83.02,
|
227 |
+
|
228 |
+
"Relationship-Overall": 70.18,
|
229 |
+
"Composition": 79.05,
|
230 |
+
"Similarity": 61.11,
|
231 |
+
"Inclusion": 70.11,
|
232 |
+
"Comparison": 62.50,
|
233 |
+
|
234 |
+
"Compound-Overall": 64.69,
|
235 |
+
"Imagination": 69.90,
|
236 |
+
"Feature matching": 59.38,
|
237 |
+
|
238 |
+
"Grammar-Overall": 61.63,
|
239 |
+
"Pronoun Reference": 75.37,
|
240 |
+
"Consistency": 65.28,
|
241 |
+
"Negation": 44.23,
|
242 |
+
|
243 |
+
"Layout-Overall": 75.37,
|
244 |
+
"2D": 77.94,
|
245 |
+
"3D": 72.73,
|
246 |
+
|
247 |
+
"Logical Reasoning": 40.91,
|
248 |
+
|
249 |
+
"Text": 1.15
|
250 |
+
},
|
251 |
+
{
|
252 |
+
"model": "Bagel",
|
253 |
+
"link": "https://arxiv.org/pdf/2505.14683",
|
254 |
+
"hf": "https://huggingface.co/ByteDance-Seed/BAGEL-7B-MoT",
|
255 |
+
"open_source": true,
|
256 |
+
"release_date": "2025-05",
|
257 |
+
|
258 |
+
"Overall": 61.53,
|
259 |
+
"Style": 90.20,
|
260 |
+
"World Knowledge": 85.60,
|
261 |
+
"Attribute-Overall": 67.74,
|
262 |
+
"Quantity": 59.03,
|
263 |
+
"Expression": 50.00,
|
264 |
+
"Material": 72.64,
|
265 |
+
"Size": 76.39,
|
266 |
+
"Shape": 59.38,
|
267 |
+
"Color": 93.33,
|
268 |
+
|
269 |
+
"Action-Overall": 61.98,
|
270 |
+
"Hand": 52.56,
|
271 |
+
"Full body": 60.87,
|
272 |
+
"Animal": 69.12,
|
273 |
+
"Non Contact": 62.24,
|
274 |
+
"Contact": 58.93,
|
275 |
+
"State": 67.45,
|
276 |
+
|
277 |
+
"Relationship-Overall": 70.69,
|
278 |
+
"Composition": 76.35,
|
279 |
+
"Similarity": 70.56,
|
280 |
+
"Inclusion": 69.57,
|
281 |
+
"Comparison": 59.38,
|
282 |
+
|
283 |
+
"Compound-Overall": 58.12,
|
284 |
+
"Imagination": 67.35,
|
285 |
+
"Feature matching": 48.70,
|
286 |
+
|
287 |
+
"Grammar-Overall": 66.44,
|
288 |
+
"Pronoun Reference": 71.69,
|
289 |
+
"Consistency": 68.52,
|
290 |
+
"Negation": 59.23,
|
291 |
+
|
292 |
+
"Layout-Overall": 76.49,
|
293 |
+
"2D": 79.04,
|
294 |
+
"3D": 73.86,
|
295 |
+
|
296 |
+
"Logical Reasoning": 30.23,
|
297 |
+
|
298 |
+
"Text": 7.76
|
299 |
+
},
|
300 |
+
{
|
301 |
+
"model": "BLIP3-o",
|
302 |
"link": "https://arxiv.org/pdf/2505.09568",
|
303 |
"hf": "https://huggingface.co/BLIP3o/BLIP3o-Model-8B",
|
304 |
"open_source": true,
|
305 |
+
"release_date": "2025-05",
|
306 |
|
307 |
"Overall": 59.87,
|
308 |
"Style": 92.80,
|
|
|
351 |
"link": "https://arxiv.org/pdf/2403.05121",
|
352 |
"hf": "https://huggingface.co/zai-org/CogView4-6B",
|
353 |
"open_source": true,
|
354 |
+
"release_date": "2024-03",
|
355 |
|
356 |
"Overall": 56.30,
|
357 |
"Style": 82.00,
|
|
|
400 |
"link": "https://arxiv.org/pdf/2405.08748",
|
401 |
"hf": "https://huggingface.co/Tencent-Hunyuan/HunyuanDiT",
|
402 |
"open_source": true,
|
403 |
+
"release_date": "2024-05",
|
404 |
|
405 |
"Overall": 51.38,
|
406 |
|