Datasets:

Modalities:
Text
Formats:
parquet
Languages:
English
ArXiv:
Libraries:
Datasets
pandas
License:
Y-J-Ju commited on
Commit
ba35b82
·
verified ·
1 Parent(s): df9ca9d

Upload folder using huggingface_hub

Browse files
Files changed (39) hide show
  1. A-OKVQA/diverse_instruction-00000-of-00001.parquet +3 -0
  2. A-OKVQA/original-00000-of-00001.parquet +3 -0
  3. CIRR/diverse_instruction-00000-of-00001.parquet +3 -0
  4. CIRR/original-00000-of-00001.parquet +3 -0
  5. ChartQA/diverse_instruction-00000-of-00001.parquet +3 -0
  6. ChartQA/original-00000-of-00001.parquet +3 -0
  7. DocVQA/diverse_instruction-00000-of-00001.parquet +3 -0
  8. DocVQA/original-00000-of-00001.parquet +3 -0
  9. HatefulMemes/diverse_instruction-00000-of-00001.parquet +3 -0
  10. HatefulMemes/original-00000-of-00001.parquet +3 -0
  11. InfographicsVQA/diverse_instruction-00000-of-00001.parquet +3 -0
  12. InfographicsVQA/original-00000-of-00001.parquet +3 -0
  13. MSCOCO/diverse_instruction-00000-of-00001.parquet +3 -0
  14. MSCOCO/original-00000-of-00001.parquet +3 -0
  15. MSCOCO_i2t/diverse_instruction-00000-of-00001.parquet +3 -0
  16. MSCOCO_i2t/original-00000-of-00001.parquet +3 -0
  17. MSCOCO_t2i/diverse_instruction-00000-of-00001.parquet +3 -0
  18. MSCOCO_t2i/original-00000-of-00001.parquet +3 -0
  19. N24News/diverse_instruction-00000-of-00001.parquet +3 -0
  20. N24News/original-00000-of-00001.parquet +3 -0
  21. NIGHTS/diverse_instruction-00000-of-00001.parquet +3 -0
  22. NIGHTS/original-00000-of-00001.parquet +3 -0
  23. OK-VQA/diverse_instruction-00000-of-00001.parquet +3 -0
  24. OK-VQA/original-00000-of-00001.parquet +3 -0
  25. README.md +659 -3
  26. SUN397/diverse_instruction-00000-of-00001.parquet +3 -0
  27. SUN397/original-00000-of-00001.parquet +3 -0
  28. VOC2007/diverse_instruction-00000-of-00001.parquet +3 -0
  29. VOC2007/original-00000-of-00001.parquet +3 -0
  30. VisDial/diverse_instruction-00000-of-00001.parquet +3 -0
  31. VisDial/original-00000-of-00001.parquet +3 -0
  32. Visual7W/diverse_instruction-00000-of-00001.parquet +3 -0
  33. Visual7W/original-00000-of-00001.parquet +3 -0
  34. VisualNews_i2t/diverse_instruction-00000-of-00001.parquet +3 -0
  35. VisualNews_i2t/original-00000-of-00001.parquet +3 -0
  36. VisualNews_t2i/diverse_instruction-00000-of-00001.parquet +3 -0
  37. VisualNews_t2i/original-00000-of-00001.parquet +3 -0
  38. WebQA/diverse_instruction-00000-of-00001.parquet +3 -0
  39. WebQA/original-00000-of-00001.parquet +3 -0
A-OKVQA/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98608da491b0efe88cc58c3a57c07b1a2289983d2029557d6149f81fcd2e3b0f
3
+ size 916955
A-OKVQA/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98608da491b0efe88cc58c3a57c07b1a2289983d2029557d6149f81fcd2e3b0f
3
+ size 916955
CIRR/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d8b859cab938bf34c0eddba1c205bbd26e8b87ad1cc5e7ceb88b1ebf35f4cf09
3
+ size 1574513
CIRR/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d8b859cab938bf34c0eddba1c205bbd26e8b87ad1cc5e7ceb88b1ebf35f4cf09
3
+ size 1574513
ChartQA/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:594d5eeff835116b419911544a840eaedd4177fa9b8f8224022ccd0cafcdfd4d
3
+ size 1703393
ChartQA/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:594d5eeff835116b419911544a840eaedd4177fa9b8f8224022ccd0cafcdfd4d
3
+ size 1703393
DocVQA/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:229a34873d46dc863ae63c9c528051f1fd2a8ba948f5654ea714852946dae2b6
3
+ size 2255557
DocVQA/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:229a34873d46dc863ae63c9c528051f1fd2a8ba948f5654ea714852946dae2b6
3
+ size 2255557
HatefulMemes/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c06c8ebec5a54e49f96f96f2f346366fb4988a291955f964b552f046c6d64ed
3
+ size 77042
HatefulMemes/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c06c8ebec5a54e49f96f96f2f346366fb4988a291955f964b552f046c6d64ed
3
+ size 77042
InfographicsVQA/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e97d73adfc73a617d215efcac883f203af4b2246e6a2414b21e2940bfadc90d3
3
+ size 1517302
InfographicsVQA/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e97d73adfc73a617d215efcac883f203af4b2246e6a2414b21e2940bfadc90d3
3
+ size 1517302
MSCOCO/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1fe701c48ef4fd447fa86bf143537bb3ea0b66edafc864b3d45b13b0d8e9f1f0
3
+ size 1692757
MSCOCO/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:897f71d52bfb5a3e32ad84e086d5d19cc7e0899eec1fcc6befea7b3b097b7fb2
3
+ size 1528747
MSCOCO_i2t/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f19da51abb880e34071b5f20dd6c2de4ca1ad4684e4029457a3dfe59310178a3
3
+ size 4800883
MSCOCO_i2t/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f19da51abb880e34071b5f20dd6c2de4ca1ad4684e4029457a3dfe59310178a3
3
+ size 4800883
MSCOCO_t2i/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c5328da99a06bf1283f85bf2e1b410cc6c75aacf6678913249918a9e0d45726e
3
+ size 4387878
MSCOCO_t2i/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c5328da99a06bf1283f85bf2e1b410cc6c75aacf6678913249918a9e0d45726e
3
+ size 4387878
N24News/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c82bac33dc7973faa1780c3d59ceb463588b80e2313db26f4554c05f229f552
3
+ size 4923108
N24News/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c82bac33dc7973faa1780c3d59ceb463588b80e2313db26f4554c05f229f552
3
+ size 4923108
NIGHTS/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0107909b7feed6e0578aa380c22b1095980edfbad18e32d9dfa5100b0d0efbaf
3
+ size 244030
NIGHTS/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0107909b7feed6e0578aa380c22b1095980edfbad18e32d9dfa5100b0d0efbaf
3
+ size 244030
OK-VQA/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b1358410c4577682b980014a54119585869d3810d27c8747d6ab79a10830a41a
3
+ size 472275
OK-VQA/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b1358410c4577682b980014a54119585869d3810d27c8747d6ab79a10830a41a
3
+ size 472275
README.md CHANGED
@@ -1,3 +1,659 @@
1
- ---
2
- license: apache-2.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ license: apache-2.0
5
+ size_categories:
6
+ - 1M<n<10M
7
+ pretty_name: MMEB
8
+ dataset_info:
9
+ - config_name: A-OKVQA
10
+ features:
11
+ - name: qry
12
+ dtype: string
13
+ - name: qry_image_path
14
+ dtype: string
15
+ - name: pos_text
16
+ dtype: string
17
+ - name: pos_image_path
18
+ dtype: string
19
+ - name: neg_text
20
+ dtype: string
21
+ - name: neg_image_path
22
+ dtype: string
23
+ - name: task_instruction
24
+ dtype: string
25
+ - name: represent_prompt
26
+ dtype: string
27
+ splits:
28
+ - name: original
29
+ num_bytes: 3338954
30
+ num_examples: 17056
31
+ - name: diverse_instruction
32
+ num_bytes: 3133737
33
+ num_examples: 17056
34
+ download_size: 2047947
35
+ dataset_size: 6472691
36
+ - config_name: CIRR
37
+ features:
38
+ - name: qry
39
+ dtype: string
40
+ - name: qry_image_path
41
+ dtype: string
42
+ - name: pos_text
43
+ dtype: string
44
+ - name: pos_image_path
45
+ dtype: string
46
+ - name: task_instruction
47
+ dtype: string
48
+ - name: represent_prompt
49
+ dtype: string
50
+ splits:
51
+ - name: original
52
+ num_bytes: 7329334
53
+ num_examples: 26116
54
+ - name: diverse_instruction
55
+ num_bytes: 7372754
56
+ num_examples: 26116
57
+ download_size: 3565805
58
+ dataset_size: 14702088
59
+ - config_name: ChartQA
60
+ features:
61
+ - name: qry
62
+ dtype: string
63
+ - name: qry_image_path
64
+ dtype: string
65
+ - name: pos_text
66
+ dtype: string
67
+ - name: pos_image_path
68
+ dtype: string
69
+ - name: neg_text
70
+ dtype: string
71
+ - name: neg_image_path
72
+ dtype: string
73
+ - name: task_instruction
74
+ dtype: string
75
+ - name: represent_prompt
76
+ dtype: string
77
+ splits:
78
+ - name: original
79
+ num_bytes: 6038803
80
+ num_examples: 28299
81
+ - name: diverse_instruction
82
+ num_bytes: 5699001
83
+ num_examples: 28299
84
+ download_size: 3716222
85
+ dataset_size: 11737804
86
+ - config_name: DocVQA
87
+ features:
88
+ - name: qry
89
+ dtype: string
90
+ - name: qry_image_path
91
+ dtype: string
92
+ - name: pos_text
93
+ dtype: string
94
+ - name: pos_image_path
95
+ dtype: string
96
+ - name: neg_text
97
+ dtype: string
98
+ - name: neg_image_path
99
+ dtype: string
100
+ - name: task_instruction
101
+ dtype: string
102
+ - name: represent_prompt
103
+ dtype: string
104
+ splits:
105
+ - name: original
106
+ num_bytes: 8110162
107
+ num_examples: 39463
108
+ - name: diverse_instruction
109
+ num_bytes: 7639455
110
+ num_examples: 39463
111
+ download_size: 4945343
112
+ dataset_size: 15749617
113
+ - config_name: HatefulMemes
114
+ features:
115
+ - name: qry
116
+ dtype: string
117
+ - name: qry_image_path
118
+ dtype: string
119
+ - name: pos_text
120
+ dtype: string
121
+ - name: pos_image_path
122
+ dtype: string
123
+ - name: neg_text
124
+ dtype: string
125
+ - name: neg_image_path
126
+ dtype: string
127
+ - name: task_instruction
128
+ dtype: string
129
+ - name: represent_prompt
130
+ dtype: string
131
+ splits:
132
+ - name: original
133
+ num_bytes: 1732890
134
+ num_examples: 8500
135
+ - name: diverse_instruction
136
+ num_bytes: 1457996
137
+ num_examples: 8500
138
+ download_size: 164510
139
+ dataset_size: 3190886
140
+ - config_name: ImageNet_1K
141
+ features:
142
+ - name: qry
143
+ dtype: string
144
+ - name: qry_image_path
145
+ dtype: string
146
+ - name: pos_text
147
+ dtype: string
148
+ - name: pos_image_path
149
+ dtype: string
150
+ - name: task_instruction
151
+ dtype: string
152
+ - name: represent_prompt
153
+ dtype: string
154
+ splits:
155
+ - name: original
156
+ num_bytes: 13364615
157
+ num_examples: 100000
158
+ - name: diverse_instruction
159
+ num_bytes: 12420504
160
+ num_examples: 100000
161
+ download_size: 4229920
162
+ dataset_size: 25785119
163
+ - config_name: InfographicsVQA
164
+ features:
165
+ - name: qry
166
+ dtype: string
167
+ - name: qry_image_path
168
+ dtype: string
169
+ - name: pos_text
170
+ dtype: string
171
+ - name: pos_image_path
172
+ dtype: string
173
+ - name: neg_text
174
+ dtype: string
175
+ - name: neg_image_path
176
+ dtype: string
177
+ - name: task_instruction
178
+ dtype: string
179
+ - name: represent_prompt
180
+ dtype: string
181
+ splits:
182
+ - name: original
183
+ num_bytes: 5257807
184
+ num_examples: 23946
185
+ - name: diverse_instruction
186
+ num_bytes: 4971080
187
+ num_examples: 23946
188
+ download_size: 3285414
189
+ dataset_size: 10228887
190
+ - config_name: MSCOCO
191
+ features:
192
+ - name: qry
193
+ dtype: string
194
+ - name: qry_image_path
195
+ dtype: string
196
+ - name: pos_text
197
+ dtype: string
198
+ - name: pos_image_path
199
+ dtype: string
200
+ - name: task_instruction
201
+ dtype: string
202
+ - name: represent_prompt
203
+ dtype: string
204
+ splits:
205
+ - name: original
206
+ num_bytes: 24352395
207
+ num_examples: 100000
208
+ - name: diverse_instruction
209
+ num_bytes: 23402480
210
+ num_examples: 100000
211
+ download_size: 3245813
212
+ dataset_size: 47754875
213
+ - config_name: MSCOCO_i2t
214
+ features:
215
+ - name: qry
216
+ dtype: string
217
+ - name: qry_image_path
218
+ dtype: string
219
+ - name: pos_text
220
+ dtype: string
221
+ - name: pos_image_path
222
+ dtype: string
223
+ - name: task_instruction
224
+ dtype: string
225
+ - name: represent_prompt
226
+ dtype: string
227
+ splits:
228
+ - name: original
229
+ num_bytes: 21949351
230
+ num_examples: 113287
231
+ - name: diverse_instruction
232
+ num_bytes: 21905631
233
+ num_examples: 113287
234
+ download_size: 9735223
235
+ dataset_size: 43854982
236
+ - config_name: MSCOCO_t2i
237
+ features:
238
+ - name: qry
239
+ dtype: string
240
+ - name: qry_image_path
241
+ dtype: string
242
+ - name: pos_text
243
+ dtype: string
244
+ - name: pos_image_path
245
+ dtype: string
246
+ - name: task_instruction
247
+ dtype: string
248
+ - name: represent_prompt
249
+ dtype: string
250
+ splits:
251
+ - name: original
252
+ num_bytes: 22135986
253
+ num_examples: 100000
254
+ - name: diverse_instruction
255
+ num_bytes: 22855603
256
+ num_examples: 100000
257
+ download_size: 10137403
258
+ dataset_size: 44991589
259
+ - config_name: N24News
260
+ features:
261
+ - name: qry
262
+ dtype: string
263
+ - name: qry_image_path
264
+ dtype: string
265
+ - name: pos_text
266
+ dtype: string
267
+ - name: pos_image_path
268
+ dtype: string
269
+ - name: neg_text
270
+ dtype: string
271
+ - name: neg_image_path
272
+ dtype: string
273
+ - name: task_instruction
274
+ dtype: string
275
+ - name: represent_prompt
276
+ dtype: string
277
+ splits:
278
+ - name: original
279
+ num_bytes: 14534664
280
+ num_examples: 48988
281
+ - name: diverse_instruction
282
+ num_bytes: 13764900
283
+ num_examples: 48988
284
+ download_size: 10773266
285
+ dataset_size: 28299564
286
+ - config_name: NIGHTS
287
+ features:
288
+ - name: qry
289
+ dtype: string
290
+ - name: qry_image_path
291
+ dtype: string
292
+ - name: pos_text
293
+ dtype: string
294
+ - name: pos_image_path
295
+ dtype: string
296
+ - name: task_instruction
297
+ dtype: string
298
+ - name: represent_prompt
299
+ dtype: string
300
+ splits:
301
+ - name: original
302
+ num_bytes: 3331669
303
+ num_examples: 15941
304
+ - name: diverse_instruction
305
+ num_bytes: 3042355
306
+ num_examples: 15941
307
+ download_size: 507810
308
+ dataset_size: 6374024
309
+ - config_name: OK-VQA
310
+ features:
311
+ - name: qry
312
+ dtype: string
313
+ - name: qry_image_path
314
+ dtype: string
315
+ - name: pos_text
316
+ dtype: string
317
+ - name: pos_image_path
318
+ dtype: string
319
+ - name: neg_text
320
+ dtype: string
321
+ - name: neg_image_path
322
+ dtype: string
323
+ - name: task_instruction
324
+ dtype: string
325
+ - name: represent_prompt
326
+ dtype: string
327
+ splits:
328
+ - name: original
329
+ num_bytes: 1703825
330
+ num_examples: 9009
331
+ - name: diverse_instruction
332
+ num_bytes: 1596220
333
+ num_examples: 9009
334
+ download_size: 1054499
335
+ dataset_size: 3300045
336
+ - config_name: SUN397
337
+ features:
338
+ - name: qry
339
+ dtype: string
340
+ - name: qry_image_path
341
+ dtype: string
342
+ - name: pos_text
343
+ dtype: string
344
+ - name: pos_image_path
345
+ dtype: string
346
+ - name: task_instruction
347
+ dtype: string
348
+ - name: represent_prompt
349
+ dtype: string
350
+ splits:
351
+ - name: original
352
+ num_bytes: 2229890
353
+ num_examples: 19850
354
+ - name: diverse_instruction
355
+ num_bytes: 2243231
356
+ num_examples: 19850
357
+ download_size: 336494
358
+ dataset_size: 4473121
359
+ - config_name: VOC2007
360
+ features:
361
+ - name: qry
362
+ dtype: string
363
+ - name: qry_image_path
364
+ dtype: string
365
+ - name: pos_text
366
+ dtype: string
367
+ - name: pos_image_path
368
+ dtype: string
369
+ - name: task_instruction
370
+ dtype: string
371
+ - name: represent_prompt
372
+ dtype: string
373
+ splits:
374
+ - name: original
375
+ num_bytes: 875081
376
+ num_examples: 7844
377
+ - name: diverse_instruction
378
+ num_bytes: 854695
379
+ num_examples: 7844
380
+ download_size: 142534
381
+ dataset_size: 1729776
382
+ - config_name: VisDial
383
+ features:
384
+ - name: qry
385
+ dtype: string
386
+ - name: qry_image_path
387
+ dtype: string
388
+ - name: pos_text
389
+ dtype: string
390
+ - name: pos_image_path
391
+ dtype: string
392
+ - name: neg_text
393
+ dtype: string
394
+ - name: neg_image_path
395
+ dtype: string
396
+ - name: task_instruction
397
+ dtype: string
398
+ - name: represent_prompt
399
+ dtype: string
400
+ splits:
401
+ - name: original
402
+ num_bytes: 80780541
403
+ num_examples: 123287
404
+ - name: diverse_instruction
405
+ num_bytes: 80709365
406
+ num_examples: 123287
407
+ download_size: 58273400
408
+ dataset_size: 161489906
409
+ - config_name: Visual7W
410
+ features:
411
+ - name: qry
412
+ dtype: string
413
+ - name: qry_image_path
414
+ dtype: string
415
+ - name: pos_text
416
+ dtype: string
417
+ - name: pos_image_path
418
+ dtype: string
419
+ - name: neg_text
420
+ dtype: string
421
+ - name: neg_image_path
422
+ dtype: string
423
+ - name: task_instruction
424
+ dtype: string
425
+ - name: represent_prompt
426
+ dtype: string
427
+ splits:
428
+ - name: original
429
+ num_bytes: 12380002
430
+ num_examples: 69817
431
+ - name: diverse_instruction
432
+ num_bytes: 11542967
433
+ num_examples: 69817
434
+ download_size: 5649037
435
+ dataset_size: 23922969
436
+ - config_name: VisualNews_i2t
437
+ features:
438
+ - name: qry
439
+ dtype: string
440
+ - name: qry_image_path
441
+ dtype: string
442
+ - name: pos_text
443
+ dtype: string
444
+ - name: pos_image_path
445
+ dtype: string
446
+ - name: task_instruction
447
+ dtype: string
448
+ - name: represent_prompt
449
+ dtype: string
450
+ splits:
451
+ - name: original
452
+ num_bytes: 24364789
453
+ num_examples: 100000
454
+ - name: diverse_instruction
455
+ num_bytes: 24692754
456
+ num_examples: 100000
457
+ download_size: 18948257
458
+ dataset_size: 49057543
459
+ - config_name: VisualNews_t2i
460
+ features:
461
+ - name: qry
462
+ dtype: string
463
+ - name: qry_image_path
464
+ dtype: string
465
+ - name: pos_text
466
+ dtype: string
467
+ - name: pos_image_path
468
+ dtype: string
469
+ - name: task_instruction
470
+ dtype: string
471
+ - name: represent_prompt
472
+ dtype: string
473
+ splits:
474
+ - name: original
475
+ num_bytes: 26342434
476
+ num_examples: 99903
477
+ - name: diverse_instruction
478
+ num_bytes: 27553500
479
+ num_examples: 99903
480
+ download_size: 19990061
481
+ dataset_size: 53895934
482
+ - config_name: WebQA
483
+ features:
484
+ - name: qry
485
+ dtype: string
486
+ - name: qry_image_path
487
+ dtype: string
488
+ - name: pos_text
489
+ dtype: string
490
+ - name: pos_image_path
491
+ dtype: string
492
+ - name: task_instruction
493
+ dtype: string
494
+ - name: represent_prompt
495
+ dtype: string
496
+ splits:
497
+ - name: original
498
+ num_bytes: 6439378
499
+ num_examples: 17166
500
+ - name: diverse_instruction
501
+ num_bytes: 6675874
502
+ num_examples: 17166
503
+ download_size: 5202887
504
+ dataset_size: 13115252
505
+ configs:
506
+ - config_name: A-OKVQA
507
+ data_files:
508
+ - split: original
509
+ path: A-OKVQA/original-*
510
+ - split: diverse_instruction
511
+ path: A-OKVQA/diverse_instruction-*
512
+ - config_name: CIRR
513
+ data_files:
514
+ - split: original
515
+ path: CIRR/original-*
516
+ - split: diverse_instruction
517
+ path: CIRR/diverse_instruction-*
518
+ - config_name: ChartQA
519
+ data_files:
520
+ - split: original
521
+ path: ChartQA/original-*
522
+ - split: diverse_instruction
523
+ path: ChartQA/diverse_instruction-*
524
+ - config_name: DocVQA
525
+ data_files:
526
+ - split: original
527
+ path: DocVQA/original-*
528
+ - split: diverse_instruction
529
+ path: DocVQA/diverse_instruction-*
530
+ - config_name: HatefulMemes
531
+ data_files:
532
+ - split: original
533
+ path: HatefulMemes/original-*
534
+ - split: diverse_instruction
535
+ path: HatefulMemes/diverse_instruction-*
536
+ - config_name: ImageNet_1K
537
+ data_files:
538
+ - split: original
539
+ path: ImageNet_1K/original-*
540
+ - split: diverse_instruction
541
+ path: ImageNet_1K/diverse_instruction-*
542
+ - config_name: InfographicsVQA
543
+ data_files:
544
+ - split: original
545
+ path: InfographicsVQA/original-*
546
+ - split: diverse_instruction
547
+ path: InfographicsVQA/diverse_instruction-*
548
+ - config_name: MSCOCO
549
+ data_files:
550
+ - split: original
551
+ path: MSCOCO/original-*
552
+ - split: diverse_instruction
553
+ path: MSCOCO/diverse_instruction-*
554
+ - config_name: MSCOCO_i2t
555
+ data_files:
556
+ - split: original
557
+ path: MSCOCO_i2t/original-*
558
+ - split: diverse_instruction
559
+ path: MSCOCO_i2t/diverse_instruction-*
560
+ - config_name: MSCOCO_t2i
561
+ data_files:
562
+ - split: original
563
+ path: MSCOCO_t2i/original-*
564
+ - split: diverse_instruction
565
+ path: MSCOCO_t2i/diverse_instruction-*
566
+ - config_name: N24News
567
+ data_files:
568
+ - split: original
569
+ path: N24News/original-*
570
+ - split: diverse_instruction
571
+ path: N24News/diverse_instruction-*
572
+ - config_name: NIGHTS
573
+ data_files:
574
+ - split: original
575
+ path: NIGHTS/original-*
576
+ - split: diverse_instruction
577
+ path: NIGHTS/diverse_instruction-*
578
+ - config_name: OK-VQA
579
+ data_files:
580
+ - split: original
581
+ path: OK-VQA/original-*
582
+ - split: diverse_instruction
583
+ path: OK-VQA/diverse_instruction-*
584
+ - config_name: SUN397
585
+ data_files:
586
+ - split: original
587
+ path: SUN397/original-*
588
+ - split: diverse_instruction
589
+ path: SUN397/diverse_instruction-*
590
+ - config_name: VOC2007
591
+ data_files:
592
+ - split: original
593
+ path: VOC2007/original-*
594
+ - split: diverse_instruction
595
+ path: VOC2007/diverse_instruction-*
596
+ - config_name: VisDial
597
+ data_files:
598
+ - split: original
599
+ path: VisDial/original-*
600
+ - split: diverse_instruction
601
+ path: VisDial/diverse_instruction-*
602
+ - config_name: Visual7W
603
+ data_files:
604
+ - split: original
605
+ path: Visual7W/original-*
606
+ - split: diverse_instruction
607
+ path: Visual7W/diverse_instruction-*
608
+ - config_name: VisualNews_i2t
609
+ data_files:
610
+ - split: original
611
+ path: VisualNews_i2t/original-*
612
+ - split: diverse_instruction
613
+ path: VisualNews_i2t/diverse_instruction-*
614
+ - config_name: VisualNews_t2i
615
+ data_files:
616
+ - split: original
617
+ path: VisualNews_t2i/original-*
618
+ - split: diverse_instruction
619
+ path: VisualNews_t2i/diverse_instruction-*
620
+ - config_name: WebQA
621
+ data_files:
622
+ - split: original
623
+ path: WebQA/original-*
624
+ - split: diverse_instruction
625
+ path: WebQA/diverse_instruction-*
626
+ tags:
627
+ - embedding
628
+ ---
629
+
630
+
631
+ # Massive Multimodal Embedding Benchmark
632
+
633
+ The training data split used for training VLM2Vec models in the paper [VLM2Vec: Training Vision-Language Models for Massive Multimodal Embedding Tasks](https://arxiv.org/abs/2410.05160) (ICLR 2025).
634
+
635
+ MMEB benchmark covers 4 meta tasks and 36 datasets meticulously selected for evaluating capabilities of multimodal embedding models.
636
+
637
+ During training, we utilize 20 out of the 36 datasets.
638
+
639
+ For evaluation, we assess performance on the 20 in-domain (IND) datasets and the remaining 16 out-of-domain (OOD) datasets. Please refer to [TIGER-Lab/MMEB-eval](https://huggingface.co/datasets/TIGER-Lab/MMEB-eval) for the test split of MMEB.
640
+
641
+ # News
642
+ [2025-01]: We have updated our training data. Each subset now contains two splits: ```original``` and ```diverse_instruction```. The ```original``` split is provided to support the reproduction of our paper results. The ```diverse_instruction``` split includes paraphrased instructions for each task, designed to enhance instruction diversity and improve the model's robustness to unseen instructions and tasks. Moving forward, our future releases will primarily use the ```diverse_instruction``` split.
643
+
644
+ ## Dataset Usage
645
+ For each dataset, we have 1000 examples for evaluation. Each example contains a query and a set of targets. Both the query and target could be any combination of image and text. The first one in the candidate list is the groundtruth target.
646
+
647
+ ## Statistics
648
+ We show the statistics of all the datasets as follows:
649
+ <img width="900" alt="abs" src="https://huggingface.co/datasets/TIGER-Lab/MMEB-eval/resolve/main/statistics.png?download=true">
650
+
651
+ ## Cite Us
652
+ ```
653
+ @article{jiang2024vlm2vec,
654
+ title={VLM2Vec: Training Vision-Language Models for Massive Multimodal Embedding Tasks},
655
+ author={Jiang, Ziyan and Meng, Rui and Yang, Xinyi and Yavuz, Semih and Zhou, Yingbo and Chen, Wenhu},
656
+ journal={arXiv preprint arXiv:2410.05160},
657
+ year={2024}
658
+ }
659
+ ```
SUN397/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7bb81a8abbc44c35521f7d2e5d98874b21a4c3fc6dfa1bf4bc6b64a52878fefa
3
+ size 161143
SUN397/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7bb81a8abbc44c35521f7d2e5d98874b21a4c3fc6dfa1bf4bc6b64a52878fefa
3
+ size 161143
VOC2007/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c65bdbfc5b97ce20fb15c7b61ffe5b45350c8e806b35cfce5c0c825fb20fab99
3
+ size 68874
VOC2007/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c65bdbfc5b97ce20fb15c7b61ffe5b45350c8e806b35cfce5c0c825fb20fab99
3
+ size 68874
VisDial/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb423474b0a087a26b75b7293bfb11a7c46a573c12eefef57168f1881bd12755
3
+ size 27599491
VisDial/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb423474b0a087a26b75b7293bfb11a7c46a573c12eefef57168f1881bd12755
3
+ size 27599491
Visual7W/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:58bc93a3987da5871979d917ce03fef889807c023349892bf0b704ef75d3321e
3
+ size 2466336
Visual7W/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:58bc93a3987da5871979d917ce03fef889807c023349892bf0b704ef75d3321e
3
+ size 2466336
VisualNews_i2t/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8de75c5eb792ac83075626ba499d25a1b193c9160a995a583ebff91ee8eafff1
3
+ size 9402055
VisualNews_i2t/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8de75c5eb792ac83075626ba499d25a1b193c9160a995a583ebff91ee8eafff1
3
+ size 9402055
VisualNews_t2i/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f43390caff02ea6fbd0f9eae99952c79625a66e92ec9ea6695d4419a62d33572
3
+ size 10086770
VisualNews_t2i/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4764449d6ce3d83b9e2f5632c764b7b77af15425caa7324ab68aafbb7117659e
3
+ size 9458311
WebQA/diverse_instruction-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7fd82ee3cc2e3d84d672dc62905063f9d9bfb49afcdffc27b5915af8e80e033b
3
+ size 2373438
WebQA/original-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7fd82ee3cc2e3d84d672dc62905063f9d9bfb49afcdffc27b5915af8e80e033b
3
+ size 2373438