-
Notifications
You must be signed in to change notification settings - Fork 0
/
index.html
887 lines (867 loc) · 46.8 KB
/
index.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
<!DOCTYPE html>
<html>
<head>
<title>CODIS</title>
<style>
.hidden {
display: none;
}
</style>
<script src="https://cdn.jsdelivr.net/npm/chart.js"></script>
<script src="https://kit.fontawesome.com/f8ddf9854a.js" crossorigin="anonymous"></script>
<meta charset="utf-8">
<meta name="description" content="Benchmarking Context-Dependent Visual Comprehension for Multimodal Large Language Models">
<meta name="keywords" content="CODIS, MLLM, VLM, Multimodal Large Language Model, Vision-language Model, Evaluation, Benchmark, Context, AI">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title> CODIS: Benchmarking Context-Dependent Visual Comprehension for Multimodal Large Language Models </title>
<link rel="icon" href="./static/images/logo.png">
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro" rel="stylesheet">
<link rel="stylesheet" href="./static/css/bulma.min.css">
<link rel="stylesheet" href="./static/css/bulma-carousel.min.css">
<link rel="stylesheet" href="./static/css/bulma-slider.min.css">
<link rel="stylesheet" href="./static/css/fontawesome.all.min.css">
<link rel="stylesheet" href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
<link rel="stylesheet" href="./static/css/index.css">
<link rel="stylesheet" href="./static/css/leaderboard.css">
<script type="text/javascript" src="static/js/sort-table.js" defer></script>
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script type="text/x-mathjax-config">
MathJax.Hub.Config({
tex2jax: {inlineMath: [['$','$'], ['\\(','\\)']]}
});
</script>
<script type="text/javascript" src="https://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script>
<script defer src="./static/js/fontawesome.all.min.js"></script>
<script src="./static/js/bulma-carousel.min.js"></script>
<script src="./static/js/bulma-slider.min.js"></script>
<script src="./static/js/index.js"></script>
<script src="./static/js/question_card.js"></script>
</head>
<body>
<section class="hero">
<div class="hero-body">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column has-text-centered">
<h1 class="title is-1 publication-title is-bold">
<img src="static/images/logo.png" style="width:1em;vertical-align:middle" alt="logo">
<span class="codis" style="vertical-align:middle">CODIS</span>
</h1>
<h2 class="subtitle is-3 publication-subtitle">
Benchmarking Context-Dependent Visual Comprehension
<br>
for Multimodal Large Language Models
</h2>
<div class="is-size-5 publication-authors">
<span class="author-block">Fuwen Luo<sup style="color:#9b51e0;">1</sup>*,</span>
<span class="author-block">Chi Chen<sup style="color:#9b51e0;">1</sup>*,</span>
<span class="author-block">Zihao Wan<sup style="color:#9b51e0;">1</sup>,</span>
<span class="author-block">Zhaolu Kang<sup style="color:#007bff;">6</sup>,</span>
<span class="author-block">Qidong Yan<sup style="color:#a1a13e;">5</sup>,</span>
<span class="author-block">Yingjie Li<sup style="color:#a1a13e;">5</sup>,</span><br>
<span class="author-block">Xiaolong Wang<sup style="color:#9b51e0;">1</sup>,</span>
<span class="author-block">Siyu Wang<sup style="color:#ed4b82;">2</sup>,</span>
<span class="author-block">Ziyue Wang<sup style="color:#9b51e0;">1</sup>,</span>
<span class="author-block">Xiaoyue Mi<sup style="color:#ff6600;">7</sup>,</span><br>
<span class="author-block">Peng Li<sup style="color:#ed4b82;">2</sup><sup>,</sup><sup style="color:#33cac2;">3</sup><sup>†</sup>,</span>
<span class="author-block">Ning Ma<sup style="color:#a1a13e;">5</sup>,</span>
<span class="author-block">Maosong Sun<sup style="color:#9b51e0;">1</sup><sup>†</sup>,</span>
<span class="author-block">Yang Liu<sup style="color:#9b51e0;">1</sup><sup>,</sup><sup style="color:#ed4b82;">2</sup><sup>,</sup><sup style="color:#33cac2;">3</sup><sup>,</sup><sup style="color:#3f8f43;">4</sup></span><br>
</div>
<br>
<div class="is-size-5 publication-authors">
<span class="author-block"><sup style="color:#9b51e0;">1</sup>Dept. of Comp. Sci. & Tech., Institute for AI, Tsinghua University, Beijing, China</span><br>
<span class="author-block"><sup style="color:#ed4b82;">2</sup>Institute for AI Industry Research (AIR), Tsinghua University, Beijing, China</span><br>
<span class="author-block"><sup style="color:#33cac2;">3</sup>Shanghai Artificial Intelligence Laboratory, Shanghai, China</span><br>
<span class="author-block"><sup style="color:#3f8f43;">4</sup>Jiangsu Collaborative Innovation Center for Language Competence, Jiangsu, China</span><br>
<span class="author-block"><sup style="color:#a1a13e;">5</sup>Key Laboratory of Linguistic and Cultural Computing Ministry of Education,<br>Northwest Minzu University, China</span><br>
<span class="author-block"><sup style="color:#007bff;">6</sup>College of Software, Jilin University, China</span><br>
<span class="author-block"><sup style="color:#ff6600;">7</sup>Institute of Computing Technology, Chinese Academy of Sciences</span><br>
</div>
<br>
<div class="is-size-5 publication-authors">
<span class="author-block">*Equal contribution</span><br>
<span class="author-block">†Corresponding authors</span><br>
<span class="author-block"><a href="mailto:[email protected]">[email protected]</a>, <a href="mailto:[email protected]">[email protected]</a></span><br>
<span class="author-block"><a href="mailto:[email protected]">[email protected]</a>, <a href="mailto:[email protected]">[email protected]</a></span><br>
</div>
<div class="column has-text-centered">
<div class="publication-links">
<span class="link-block">
<a href="https://arxiv.org/abs/2402.13607" class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fas fa-file-pdf"></i>
</span>
<span>arXiv</span>
</a>
</span>
<span class="link-block">
<a href="https://huggingface.co/datasets/CODIS/CODIS" class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<p style="font-size:18px">🤗</p>
</span>
<span>Dataset</span>
</a>
</span>
<span class="link-block">
<a href="https://github.com/THUNLP-MT/CODIS" class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Github</span>
</a>
</span>
<span class="link-block">
<a href="http://49.232.144.86:9000" class="external-link button is-normal is-rounded is-dark">
<img src="static/images/logo.png" style="width:1em;vertical-align:middle" alt="logo">
<span> Playground</span>
</a>
</span>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<style>
.center {
display: block;
margin-left: auto;
margin-right: auto;
width: 80%;
}
</style>
<!-- Introduction Image -->
<section class="hero teaser">
<div class="container is-max-desktop">
<div class="content has-text-centered">
<img src="static/images/introduction.jpg" alt="introduction" width="50%">
</div>
</div>
</section>
<!--/ Introduction Image -->
<!-- Introduction -->
<section class="section">
<div class="container" style="margin-bottom: 2vh;">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<div class="content has-text-justified">
<p>
Interpretation of images can be significantly influenced by contextual information. In this instance, the determination of whether the photographer was ascending or descending a staircase remains ambiguous without supplementary context (a). However, when additional information is provided, indicating the position of the greenery relative to the observer, the direction of movement of the observer becomes clear (b).
</p>
<p>
<strong>For more cases, please go to our <a href="http://49.232.144.86:9000">playground</a>.</strong>
</p>
</div>
</div>
</div>
</div>
</section>
<!--/ Introduction -->
<!-- CODIS Benchmark Banner -->
<section class="hero is-light is-small">
<div class="hero-body has-text-centered">
<h1 class="title is-1 codis">
<span class="codis" style="vertical-align: middle">CODIS Benchmark</span>
</h1>
</div>
</section>
<!--/ CODIS Benchmark Banner -->
<!-- CODIS Benchmark -->
<section class="section">
<div class="container">
<!-- Overview -->
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Overview</h2>
<div class="content has-text-justified">
<p>
In certain situations, images need to be interpreted within a broader context. We introduce a new benchmark, named as <b>CODIS</b> (<b>CO</b>ntext-<b>D</b>ependent <b>I</b>mage di<b>S</b>ambiguation), designed to assess the ability of models to use context provided in free-form text to enhance visual comprehension. It stands out from existing benchmarks in three main aspects:
<div>1. Each image in CODIS contains inherent ambiguity that can only be resolved with additional context.</div>
<div>2. The questions are deliberately designed to highlight these ambiguities, requiring external context for accurate interpretation.</div>
<div>3. For every image-question pair, we provide two contexts in a free-form text format.</div>
</p>
<p>
We have identified five representative types of context, including three types of global context that pertain to the overall scene—namely, the global background, which encompasses location and orientation, temporal information, and cultural background, and two types of local context related to objects within the scene, specifically the attributes of objects and the relationships between people.
</p>
<div class="content has-text-centered">
<img src="static/images/taxonomy.jpg" alt="taxonomy" width="90%">
</div>
<br>
<p>
To prevent models from guessing the correct answers without fully understanding context, we organize our dataset in pairs. Each pair contains two queries $ (I,Q,C_1) $ and $ (I,Q,C_2) $. The queries have identical image $ I $ and question $ Q $, but have two pieces of different context $ C_1 $ and $ C_2 $. We give MLLMs two queries separately, and get model outputs $ O_1 $ and $ O_2 $.
</p>
<p>
We leverage two metrics, pair-wise accuracy $ \mathrm{Acc}_p $ and query-wise accuracy $ \mathrm{Acc}_q $ for our evaluation metrics. For $ \mathrm{Acc}_p $, models score only if their answers to a pair of queries are both correct. For $ \mathrm{Acc}_q $, models score for each single query they answer correctly.
</p>
<p>
We manually collect images that contain ambiguities which can only be resolved with external contexts. The majority of these images are real-scene images from the publicly available dataset ShareGPT4V and the Internet, while the remainder are created manually. For each collected image, we manually write questions, context and answers for it.
</p>
</div>
</div>
</div>
<!--/ Overview -->
<!-- Comparisons with Existing Benchmarks -->
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Comparisons with Existing Benchmarks</h2>
<div class="content has-text-justified">
<p>
We summarize recent benchmarks for MLLMs in the following table. Most of these benchmarks do not pair images with additional context. Only two benchmarks, namely VisDial and MMDialog, include extra context to help in conversation with humans rather than to clarify the meaning of images. This limitation means these benchmarks are not fully capable of testing the ability of MLLMs to understand images in a context-dependent manner.
</p>
<div class="content has-text-centered">
<img src="static/images/other_benchmark.jpg" alt="other_benchmark" width="80%">
</div>
</div>
</div>
</div>
<!--/ Comparisons with Existing Benchmarks -->
<!-- Statistics -->
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Statistics</h2>
<div class="content has-text-justified">
<p>
We collect 216 images and successfully collected a total of 706 queries, spanning five categories and encompassing a wide range of scenarios. The distribution of categories and scenarios is illustrated in the following figure.
</p>
<div class="content has-text-centered">
<img src="static/images/categories.jpg" alt="categories" width="50%">
</div>
</div>
</div>
</div>
<!--/ Statistics -->
</div>
</section>
<!--/ CODIS Benchmark -->
<!-- Evaluation Banner -->
<section class="hero is-light is-small">
<div class="hero-body has-text-centered">
<h1 class="title is-1 codis">
<span class="codis" style="vertical-align: middle">Evaluation</span>
</h1>
</div>
</section>
<!--/ Evaluation Banner -->
<!-- Evaluation -->
<section class="section">
<div class="container">
<!-- Leaderboard -->
<div class="columns is-centered m-6">
<div class="column is-full has-text-centered content">
<h2 class="title is-3" id="leaderboard">Leaderboard</h2>
<div class="content">
<div class="content has-text-justified">
<p>
We perform evaluation on 14 popular MLLMs, which are divided into three groups: (1) API-based models; (2) Open-source ~7B models; (3) Open-source ~13B models. We evaluate various models including LLMs and LMMs. "Human" refers to the average performance of five independent people. We show $ \mathrm{Acc}_p $ and $ \mathrm{Acc}_q $ based on human and GPT-4 evaluation in the leaderboard.
</p>
</div>
<!-- Human Evaluation -->
<button id="toggleButton"><b style='font-size: larger;'>Human Evaluation</b></button>
<div class="model-labels-container">
<span class="leaderboard-label" style="background-color: rgba(99, 178, 238, 0.25);">Human</span>
<span class="leaderboard-label" style="background-color: rgba(118, 218, 145, 0.25);">API-based Models</span>
<span class="leaderboard-label" style="background-color: rgba(248, 203, 127, 0.25);">Open-source ~13B Models</span>
<span class="leaderboard-label" style="background-color: rgba(248, 149, 136, 0.25);">Open-source ~7B Models</span>
</div>
<table id="table" class="js-sort-table">
<tr style="background-color: rgba(0, 0, 0, 0);" hidden="hidden"></tr>
<tr style="background-color: rgba(0, 0, 0, 0.1);">
<td class="js-sort-number" rowspan="2" style="vertical-align: middle;"><strong>Model</strong></td>
<td class="js-sort-number" colspan="2" width="14%"><strong>Loc & Ori</strong></td>
<td class="js-sort-number" colspan="2" width="14%"><strong>Temporal</strong></td>
<td class="js-sort-number" colspan="2" width="14%"><strong>Cultural</strong></td>
<td class="js-sort-number" colspan="2" width="14%"><strong>Attributes</strong></td>
<td class="js-sort-number" colspan="2" width="14%"><strong>Relationships</strong></td>
<td class="js-sort-number" colspan="2" width="14%"><strong>Overall</strong></td>
</tr>
<tr style="background-color: rgba(0, 0, 0, 0.1);">
<td class="js-sort-number"><strong>$ \mathrm{Acc}_p $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_q $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_p $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_q $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_p $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_q $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_p $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_q $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_p $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_q $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_p $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_q $</strong></td>
</tr>
<tr style="background-color: rgba(99, 178, 238, 0.25);">
<td style="text-align: left;"><b>Human</b></td>
<td><b>85.2</b></td>
<td><b>86.1</b></td>
<td><b>90.9</b></td>
<td><b>92.8</b></td>
<td><b>72.8</b></td>
<td><b>76.4</b></td>
<td><b>87.2</b></td>
<td><b>88.4</b></td>
<td><b>89.6</b></td>
<td><b>90.0</b></td>
<td><b>86.2</b></td>
<td><b>87.7</b></td>
</tr>
<tr style="background-color: rgba(118, 218, 145, 0.25);">
<td style="text-align: left;"><a href="https://openai.com/gpt-4"><b>GPT-4V</b></a></td>
<td><b>33.3</b></td>
<td><b>54.2</b></td>
<td style="text-decoration: underline;">28.4</td>
<td><b>52.1</b></td>
<td><b>25.5</b></td>
<td><b>60.6</b></td>
<td><b>26.7</b></td>
<td><b>54.7</b></td>
<td><b>51.9</b></td>
<td><b>70.2</b></td>
<td><b>32.3</b></td>
<td><b>56.9</b></td>
</tr>
<tr style="background-color: rgba(118, 218, 145, 0.25);">
<td style="text-align: left;"><a href="https://deepmind.google/technologies/gemini"><b>Gemini</b></a></td>
<td style="text-decoration: underline;">21.4</td>
<td style="text-decoration: underline;">49.4</td>
<td><b>29.5</b></td>
<td style="text-decoration: underline;">51.1</td>
<td style="text-decoration: underline;">21.3</td>
<td style="text-decoration: underline;">56.4</td>
<td style="text-decoration: underline;">24.0</td>
<td style="text-decoration: underline;">52.0</td>
<td style="text-decoration: underline;">34.6</td>
<td style="text-decoration: underline;">58.7</td>
<td style="text-decoration: underline;">26.1</td>
<td style="text-decoration: underline;">52.7</td>
</tr>
<tr style="background-color: rgba(248, 203, 127, 0.25);">
<td style="text-align: left;"><a href="https://github.com/haotian-liu/LLaVA"><b>LLaVA-1.5-13B</b></a></td>
<td><b><span style="visibility: hidden">0</span>6.0</b></td>
<td><b>41.1</b></td>
<td style="text-decoration: underline;"><span style="visibility: hidden">0</span>4.2</td>
<td style="text-decoration: underline;">44.7</td>
<td><b>10.6</b></td>
<td><b>50.0</b></td>
<td><b>14.7</b></td>
<td><b>51.3</b></td>
<td><b>13.5</b></td>
<td><b>54.8</b></td>
<td><b><span style="visibility: hidden">0</span>9.1</b></td>
<td><b>47.5</b></td>
</tr>
<tr style="background-color: rgba(248, 203, 127, 0.25);">
<td style="text-align: left;"><a href="https://github.com/salesforce/LAVIS/tree/main/projects/blip2"><b>BLIP-2-11B</b></a></td>
<td><b><span style="visibility: hidden">0</span>6.0</b></td>
<td>32.7</td>
<td><b><span style="visibility: hidden">0</span>8.4</b></td>
<td><b>45.8</b></td>
<td style="text-decoration: underline;"><span style="visibility: hidden">0</span>4.3</td>
<td>35.1</td>
<td style="text-decoration: underline;"><span style="visibility: hidden">0</span>6.7</td>
<td>42.0</td>
<td style="text-decoration: underline;">11.5</td>
<td style="text-decoration: underline;">51.9</td>
<td style="text-decoration: underline;"><span style="visibility: hidden">0</span>7.4</td>
<td>41.4</td>
</tr>
<tr style="background-color: rgba(248, 203, 127, 0.25);">
<td style="text-align: left;"><a href="https://github.com/salesforce/LAVIS/tree/main/projects/instructblip"><b>InstructBLIP-13B</b></a></td>
<td><b><span style="visibility: hidden">0</span>6.0</b></td>
<td style="text-decoration: underline;">39.3</td>
<td><span style="visibility: hidden">0</span>2.1</td>
<td>41.6</td>
<td style="text-decoration: underline;"><span style="visibility: hidden">0</span>4.3</td>
<td><b>50.0</b></td>
<td><span style="visibility: hidden">0</span>4.0</td>
<td style="text-decoration: underline;">44.7</td>
<td><span style="visibility: hidden">0</span>7.7</td>
<td>51.0</td>
<td><span style="visibility: hidden">0</span>4.5</td>
<td style="text-decoration: underline;">44.2</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/X-PLUG/mPLUG-Owl"><b>mPLUG-Owl-2-7B</b></a></td>
<td><b>13.1</b></td>
<td style="text-decoration: underline;">42.3</td>
<td><b><span style="visibility: hidden">0</span>9.5</b></td>
<td>41.6</td>
<td><b><span style="visibility: hidden">0</span>6.4</b></td>
<td>42.6</td>
<td><b>12.0</b></td>
<td style="text-decoration: underline;">44.7</td>
<td><b>19.2</b></td>
<td style="text-decoration: underline;">51.9</td>
<td><b>11.9</b></td>
<td style="text-decoration: underline;">44.1</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/Vision-CAIR/MiniGPT-4"><b>MiniGPT4-7B</b></a></td>
<td>10.7</td>
<td>36.3</td>
<td><span style="visibility: hidden">0</span>3.2</td>
<td>34.2</td>
<td><span style="visibility: hidden">0</span>0.0</td>
<td>27.7</td>
<td><b>12.0</b></td>
<td>35.3</td>
<td style="text-decoration: underline;">13.5</td>
<td>47.1</td>
<td style="text-decoration: underline;"><span style="visibility: hidden">0</span>7.9</td>
<td>36.0</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/haotian-liu/LLaVA"><b>LLaVA-1.5-7B</b></a></td>
<td style="text-decoration: underline;">11.9</td>
<td><b>42.9</b></td>
<td><span style="visibility: hidden">0</span>5.3</td>
<td style="text-decoration: underline;">44.7</td>
<td style="text-decoration: underline;"><span style="visibility: hidden">0</span>4.3</td>
<td>43.6</td>
<td><span style="visibility: hidden">0</span>9.3</td>
<td>39.3</td>
<td><span style="visibility: hidden">0</span>7.7</td>
<td>47.1</td>
<td style="text-decoration: underline;"><span style="visibility: hidden">0</span>7.9</td>
<td>43.3</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/salesforce/LAVIS/tree/main/projects/instructblip"><b>InstructBLIP-7B</b></a></td>
<td><span style="visibility: hidden">0</span>1.2</td>
<td>33.3</td>
<td style="text-decoration: underline;"><span style="visibility: hidden">0</span>7.4</td>
<td><b>45.8</b></td>
<td><span style="visibility: hidden">0</span>0.0</td>
<td style="text-decoration: underline;">46.8</td>
<td><span style="visibility: hidden">0</span>4.0</td>
<td>43.3</td>
<td>11.5</td>
<td>48.1</td>
<td><span style="visibility: hidden">0</span>4.8</td>
<td>42.8</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/Luodian/Otter"><b>Otter-7B</b></a></td>
<td><span style="visibility: hidden">0</span>2.4</td>
<td>32.7</td>
<td><span style="visibility: hidden">0</span>5.3</td>
<td>41.1</td>
<td style="text-decoration: underline;"><span style="visibility: hidden">0</span>4.3</td>
<td>28.7</td>
<td><span style="visibility: hidden">0</span>0.0</td>
<td>26.0</td>
<td><span style="visibility: hidden">0</span>5.8</td>
<td>40.4</td>
<td><span style="visibility: hidden">0</span>3.4</td>
<td>34.1</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/haotian-liu/LLaVA"><b>LLaVA-7B</b></a></td>
<td><span style="visibility: hidden">0</span>2.4</td>
<td>30.4</td>
<td><span style="visibility: hidden">0</span>6.3</td>
<td>34.2</td>
<td><span style="visibility: hidden">0</span>0.0</td>
<td>25.5</td>
<td><span style="visibility: hidden">0</span>1.3</td>
<td>34.0</td>
<td><span style="visibility: hidden">0</span>5.8</td>
<td>41.3</td>
<td><span style="visibility: hidden">0</span>3.4</td>
<td>33.1</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/QwenLM/Qwen-VL"><b>Qwen-VL-Chat</b></a></td>
<td><span style="visibility: hidden">0</span>3.6</td>
<td>23.8</td>
<td><span style="visibility: hidden">0</span>3.2</td>
<td>24.7</td>
<td><span style="visibility: hidden">0</span>0.0</td>
<td>24.5</td>
<td><span style="visibility: hidden">0</span>1.3</td>
<td>32.0</td>
<td><span style="visibility: hidden">0</span>9.6</td>
<td>34.6</td>
<td><span style="visibility: hidden">0</span>3.4</td>
<td>27.5</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/mlfoundations/open_flamingo"><b>OpenFlamingo-7B</b></a></td>
<td><span style="visibility: hidden">0</span>2.4</td>
<td>40.5</td>
<td><span style="visibility: hidden">0</span>2.1</td>
<td>38.9</td>
<td><span style="visibility: hidden">0</span>0.0</td>
<td>27.7</td>
<td><span style="visibility: hidden">0</span>5.3</td>
<td>36.0</td>
<td><span style="visibility: hidden">0</span>5.8</td>
<td>47.1</td>
<td><span style="visibility: hidden">0</span>3.1</td>
<td>38.4</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/salesforce/LAVIS/tree/main/projects/blip2"><b>BLIP-2-6.7B</b></a></td>
<td><span style="visibility: hidden">0</span>0.0</td>
<td>41.1</td>
<td><span style="visibility: hidden">0</span>1.1</td>
<td style="text-decoration: underline;">44.7</td>
<td><span style="visibility: hidden">0</span>2.1</td>
<td><b>48.9</b></td>
<td><span style="visibility: hidden">0</span>2.7</td>
<td><b>46.0</b></td>
<td><span style="visibility: hidden">0</span>7.7</td>
<td><b>53.8</b></td>
<td><span style="visibility: hidden">0</span>2.3</td>
<td><b>46.0</b></td>
</tr>
</table>
<!--/ Human Evaluation -->
<!-- GPT-4 Evaluation -->
<button id="toggleButton"><b style='font-size: larger;'>GPT-4 Evaluation</b></button>
<div class="model-labels-container">
<span class="leaderboard-label" style="background-color: rgba(118, 218, 145, 0.25);">API-based Models</span>
<span class="leaderboard-label" style="background-color: rgba(248, 203, 127, 0.25);">Open-source ~13B Models</span>
<span class="leaderboard-label" style="background-color: rgba(248, 149, 136, 0.25);">Open-source ~7B Models</span>
</div>
<table id="table" class="js-sort-table">
<tr style="background-color: rgba(0, 0, 0, 0);" hidden="hidden"></tr>
<tr style="background-color: rgba(0, 0, 0, 0.1);">
<td class="js-sort-number" rowspan="2" style="vertical-align: middle;"><strong>Model</strong></td>
<td class="js-sort-number" colspan="2" width="14%"><strong>Loc & Ori</strong></td>
<td class="js-sort-number" colspan="2" width="14%"><strong>Temporal</strong></td>
<td class="js-sort-number" colspan="2" width="14%"><strong>Cultural</strong></td>
<td class="js-sort-number" colspan="2" width="14%"><strong>Attributes</strong></td>
<td class="js-sort-number" colspan="2" width="14%"><strong>Relationships</strong></td>
<td class="js-sort-number" colspan="2" width="14%"><strong>Overall</strong></td>
</tr>
<tr style="background-color: rgba(0, 0, 0, 0.1);">
<td class="js-sort-number"><strong>$ \mathrm{Acc}_p $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_q $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_p $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_q $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_p $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_q $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_p $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_q $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_p $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_q $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_p $</strong></td>
<td class="js-sort-number"><strong>$ \mathrm{Acc}_q $</strong></td>
</tr>
<tr style="background-color: rgba(118, 218, 145, 0.25);">
<td style="text-align: left;"><a href="https://openai.com/gpt-4"><b>GPT-4V</b></a></td>
<td><b>33.3</b></td>
<td><b>53.6</b></td>
<td><b>28.4</b></td>
<td><b>50.5</b></td>
<td><b>21.3</b></td>
<td style="text-decoration: underline;">53.2</td>
<td><b>25.3</b></td>
<td><b>54.0</b></td>
<td><b>50.0</b></td>
<td><b>69.2</b></td>
<td><b>31.2</b></td>
<td><b>55.1</b></td>
</tr>
<tr style="background-color: rgba(118, 218, 145, 0.25);">
<td style="text-align: left;"><a href="https://deepmind.google/technologies/gemini"><b>Gemini</b></a></td>
<td style="text-decoration: underline;">20.2</td>
<td style="text-decoration: underline;">48.8</td>
<td style="text-decoration: underline;">27.4</td>
<td style="text-decoration: underline;">50.0</td>
<td><b>21.3</b></td>
<td><b>54.3</b></td>
<td style="text-decoration: underline;">22.7</td>
<td style="text-decoration: underline;">51.3</td>
<td style="text-decoration: underline;">30.8</td>
<td style="text-decoration: underline;">54.8</td>
<td style="text-decoration: underline;">24.4</td>
<td style="text-decoration: underline;">51.3</td>
</tr>
<tr style="background-color: rgba(248, 203, 127, 0.25);">
<td style="text-align: left;"><a href="https://github.com/haotian-liu/LLaVA"><b>LLaVA-1.5-13B</b></a></td>
<td><b><span style="visibility: hidden">0</span>6.0</b></td>
<td><b>41.1</b></td>
<td style="text-decoration: underline;"><span style="visibility: hidden">0</span>3.2</td>
<td style="text-decoration: underline;">43.2</td>
<td><b>12.8</b></td>
<td><b>46.8</b></td>
<td><b>13.3</b></td>
<td><b>50.0</b></td>
<td><b>11.5</b></td>
<td><b>53.8</b></td>
<td><b><span style="visibility: hidden">0</span>8.5</b></td>
<td><b>46.2</b></td>
</tr>
<tr style="background-color: rgba(248, 203, 127, 0.25);">
<td style="text-align: left;"><a href="https://github.com/salesforce/LAVIS/tree/main/projects/blip2"><b>BLIP-2-11B</b></a></td>
<td><b><span style="visibility: hidden">0</span>6.0</b></td>
<td>34.5</td>
<td><b>10.5</b></td>
<td><b>44.2</b></td>
<td><span style="visibility: hidden">0</span>4.3</td>
<td>30.9</td>
<td style="text-decoration: underline;"><span style="visibility: hidden">0</span>6.7</td>
<td>40.7</td>
<td><b>11.5</b></td>
<td>47.1</td>
<td style="text-decoration: underline;"><span style="visibility: hidden">0</span>8.0</td>
<td>39.8</td>
</tr>
<tr style="background-color: rgba(248, 203, 127, 0.25);">
<td style="text-align: left;"><a href="https://github.com/salesforce/LAVIS/tree/main/projects/instructblip"><b>InstructBLIP-13B</b></a></td>
<td><b><span style="visibility: hidden">0</span>6.0</b></td>
<td style="text-decoration: underline;">39.9</td>
<td><span style="visibility: hidden">0</span>2.1</td>
<td>41.1</td>
<td style="text-decoration: underline;"><span style="visibility: hidden">0</span>6.4</td>
<td><b>46.8</b></td>
<td><span style="visibility: hidden">0</span>4.0</td>
<td style="text-decoration: underline;">44.7</td>
<td><span style="visibility: hidden">0</span>5.8</td>
<td style="text-decoration: underline;">48.1</td>
<td><span style="visibility: hidden">0</span>4.5</td>
<td style="text-decoration: underline;">43.3</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/X-PLUG/mPLUG-Owl"><b>mPLUG-Owl-2-7B</b></a></td>
<td><b>13.1</b></td>
<td>39.9</td>
<td><b><span style="visibility: hidden">0</span>9.5</b></td>
<td>40.0</td>
<td><b><span style="visibility: hidden">0</span>4.3</b></td>
<td>41.5</td>
<td><b><span style="visibility: hidden">0</span>9.3</b></td>
<td>42.7</td>
<td><b>11.5</b></td>
<td style="text-decoration: underline;">48.1</td>
<td><b><span style="visibility: hidden">0</span>9.9</b></td>
<td>41.9</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/Vision-CAIR/MiniGPT-4"><b>MiniGPT4-7B</b></a></td>
<td style="text-decoration: underline;">10.7</td>
<td>34.5</td>
<td><span style="visibility: hidden">0</span>4.2</td>
<td>32.1</td>
<td><span style="visibility: hidden">0</span>0.0</td>
<td>27.7</td>
<td><span style="visibility: hidden">0</span>8.0</td>
<td>32.7</td>
<td><span style="visibility: hidden">0</span>9.6</td>
<td>43.3</td>
<td style="text-decoration: underline;"><span style="visibility: hidden">0</span>6.8</td>
<td>33.9</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/haotian-liu/LLaVA"><b>LLaVA-1.5-7B</b></a></td>
<td><span style="visibility: hidden">0</span>8.3</td>
<td>37.5</td>
<td><span style="visibility: hidden">0</span>1.1</td>
<td>36.3</td>
<td><span style="visibility: hidden">0</span>2.1</td>
<td>40.4</td>
<td><b><span style="visibility: hidden">0</span>9.3</b></td>
<td>37.3</td>
<td><span style="visibility: hidden">0</span>7.7</td>
<td style="text-decoration: underline;">48.1</td>
<td><span style="visibility: hidden">0</span>5.7</td>
<td>39.1</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/salesforce/LAVIS/tree/main/projects/instructblip"><b>InstructBLIP-7B</b></a></td>
<td><span style="visibility: hidden">0</span>1.2</td>
<td>34.5</td>
<td style="text-decoration: underline;"><span style="visibility: hidden">0</span>5.3</td>
<td><b>43.7</b></td>
<td><span style="visibility: hidden">0</span>0.0</td>
<td style="text-decoration: underline;">45.7</td>
<td><span style="visibility: hidden">0</span>4.0</td>
<td style="text-decoration: underline;">44.0</td>
<td><b>11.5</b></td>
<td>47.1</td>
<td><span style="visibility: hidden">0</span>4.2</td>
<td style="text-decoration: underline;"''>42.4</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/Luodian/Otter"><b>Otter-7B</b></a></td>
<td><span style="visibility: hidden">0</span>2.4</td>
<td>31.5</td>
<td><span style="visibility: hidden">0</span>3.2</td>
<td>35.3</td>
<td><span style="visibility: hidden">0</span>0.0</td>
<td>23.4</td>
<td><span style="visibility: hidden">0</span>1.3</td>
<td>27.3</td>
<td><span style="visibility: hidden">0</span>5.8</td>
<td>34.6</td>
<td><span style="visibility: hidden">0</span>2.5</td>
<td>31.0</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/haotian-liu/LLaVA"><b>LLaVA-7B</b></a></td>
<td><span style="visibility: hidden">0</span>2.4</td>
<td>29.8</td>
<td><span style="visibility: hidden">0</span>4.2</td>
<td>33.7</td>
<td><span style="visibility: hidden">0</span>0.0</td>
<td>17.0</td>
<td><span style="visibility: hidden">0</span>2.7</td>
<td>33.3</td>
<td><span style="visibility: hidden">0</span>1.9</td>
<td>37.5</td>
<td><span style="visibility: hidden">0</span>2.5</td>
<td>31.0</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/QwenLM/Qwen-VL"><b>Qwen-VL-Chat</b></a></td>
<td><span style="visibility: hidden">0</span>4.8</td>
<td>23.8</td>
<td><span style="visibility: hidden">0</span>3.2</td>
<td>23.7</td>
<td><span style="visibility: hidden">0</span>0.0</td>
<td>23.4</td>
<td><span style="visibility: hidden">0</span>1.3</td>
<td>32.0</td>
<td><span style="visibility: hidden">0</span>7.7</td>
<td>33.7</td>
<td><span style="visibility: hidden">0</span>3.4</td>
<td>26.9</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/mlfoundations/open_flamingo"><b>OpenFlamingo-7B</b></a></td>
<td><span style="visibility: hidden">0</span>2.4</td>
<td style="text-decoration: underline;">40.5</td>
<td><span style="visibility: hidden">0</span>2.1</td>
<td>38.9</td>
<td><span style="visibility: hidden">0</span>0.0</td>
<td>27.7</td>
<td><span style="visibility: hidden">0</span>5.3</td>
<td>36.0</td>
<td><span style="visibility: hidden">0</span>5.8</td>
<td>47.1</td>
<td><span style="visibility: hidden">0</span>3.1</td>
<td>38.4</td>
</tr>
<tr style="background-color: rgba(248, 149, 136, 0.25);">
<td style="text-align: left;"><a href="https://github.com/salesforce/LAVIS/tree/main/projects/blip2"><b>BLIP-2-6.7B</b></a></td>
<td><span style="visibility: hidden">0</span>0.0</td>
<td><b>42.3</b></td>
<td><span style="visibility: hidden">0</span>1.1</td>
<td style="text-decoration: underline;">43.2</td>
<td><b><span style="visibility: hidden">0</span>4.3</b></td>
<td><b>48.9</b></td>
<td><span style="visibility: hidden">0</span>4.0</td>
<td><b>46.7</b></td>
<td><span style="visibility: hidden">0</span>5.8</td>
<td><b>51.0</b></td>
<td><span style="visibility: hidden">0</span>2.5</td>
<td><b>45.6</b></td>
</tr>
</table>
<!--/ GPT-4 Evaluation -->
<p> Overall results of different models on CODIS. The best-performing model in each category is <b>in-bold</b>, and the second best is <u>underlined</u>.</p>
</div>
</div>
</div>
<!-- Leaderboard -->
<!--/ Examples -->
<div class="columns is-centered m-6">
<div class="column is-full has-text-centered content">
<h2 class="title is-3" id="examples">Examples</h2>
<div id="results-carousel" class="carousel results-carousel">
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/case_1.jpg" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/case_2.jpg" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/case_3.jpg" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/case_4.jpg" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/case_5.jpg" alt="grade-lv" width="60%"/>
</div>
</div>
</div>
</div>
</div>
<!--/ Examples -->
</div>
</section>
<!--/ Evaluation -->
<!-- Citation -->
<section class="section" id="citation">
<div class="container is-max-desktop content">
<h2 class="title is-3 has-text-centered">Citation</h2>
<pre><code>
@article{luo2024codis,
title={CODIS: Benchmarking Context-Dependent Visual Comprehension for Multimodal Large Language Models},
author={Fuwen Luo and Chi Chen and Zihao Wan and Zhaolu Kang and Qidong Yan and Yingjie Li and Xiaolong Wang and Siyu Wang and Ziyue Wang and Xiaoyue Mi and Peng Li and Ning Ma and Maosong Sun and Yang Liu},
journal={arXiv preprint arXiv:2402.13607},
year={2024}
}
</code></pre>
</div>
</section>
<!--/ Citation -->
<!-- Footer -->
<footer class="footer">
<div class="content has-text-centered">
</div>
<div class="columns is-centered">
<div class="column is-8">
<div class="content">
<p>This website is adapted from <a href="https://mmmu-benchmark.github.io/">MMMU</a>, licensed under a <a rel="license" href="http://creativecommons.org/licenses/by-sa/4.0/">Creative Commons Attribution-ShareAlike 4.0 International License</a>.</p>
</div>
</div>
</div>
</footer>
<!--/ Footer -->
<style>
#toggleButton {
background-color: #ffffff;
border: 1px solid #dddddd;
color: #555555;
padding: 10px 20px;
text-align: center;
text-decoration: none;
display: inline-block;
font-size: 14px;
margin: 4px 2px;
cursor: pointer;
border-radius: 25px;
box-shadow: 0 4px 8px 0 rgba(0,0,0,0.2);
transition-duration: 0.4s;
}
#toggleButton:hover {
box-shadow: 0 12px 16px 0 rgba(0,0,0,0.24), 0 17px 50px 0 rgba(0,0,0,0.19);
}
table {
border-collapse: collapse;
width: 100%;
margin-top: 5px;
border: 1px solid #ddd;
font-size: 14px;
}
td {
text-align: left;
padding: 8px;
}
td:hover {background-color: #ffffff;}
</style>
</body>
</html>