-
Notifications
You must be signed in to change notification settings - Fork 4
/
index.html
1128 lines (1108 loc) · 58.3 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
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
<!DOCTYPE html>
<html lang="en">
<head>
<title>MMMU</title>
<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="A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI">
<meta name="keywords" content="MMMU, LMM, LMM Evaluation, Vision Language Model, Large Language Model, Large Multimodal Model, artificial intelligence, AI, AGI, artificial general intelligence">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title> MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI</title>
<link rel="icon" href="./static/images/mmmu_icon2.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">
<script src="https://kit.fontawesome.com/fff5b27ec1.js" crossorigin="anonymous"></script>
<!-- <script src="https://kit.fontawesome.com/eaf1856e6f.js" crossorigin="anonymous"></script> -->
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></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>
</head>
<body>
<nav class="navbar" role="navigation" aria-label="main navigation">
<div class="navbar-brand">
<a role="button" class="navbar-burger" aria-label="menu" aria-expanded="false">
<span aria-hidden="true"></span>
<span aria-hidden="true"></span>
<span aria-hidden="true"></span>
</a>
</div>
<div class="navbar-menu">
<div class="navbar-start" style="flex-grow: 1; justify-content: center;">
<div class="navbar-item has-dropdown is-hoverable">
<a class="navbar-link">
More Research
</a>
<div class="navbar-dropdown">
<a class="navbar-item" href="https://huggingface.co/datasets/MMMU/MMMU_Pro">
<b>MMMU-Pro</b> <span style="font-size:18px; display: inline; margin-left: 5px;">🔥</span>
</a>
<a class="navbar-item" href="https://tiger-ai-lab.github.io/MAmmoTH/">
MAmmoTH
</a>
<a class="navbar-item" href="https://osu-nlp-group.github.io/TableLlama/">
TableLlama
</a>
<a class="navbar-item" href="https://osu-nlp-group.github.io/MagicBrush/">
MagicBrush
</a>
<a class="navbar-item" href="https://osu-nlp-group.github.io/Mind2Web/">
Mind2Web
</a>
</div>
</div>
</div>
</div>
</nav>
<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/mmmu_icon2.png" style="width:1em;vertical-align: middle" alt="Logo"/>
<span class="mmmu" style="vertical-align: middle">MMMU</span>
</h1>
<h2 class="subtitle is-3 publication-subtitle">
A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI
</h2>
<div class="is-size-5 publication-authors">
<span class="author-block">
<a href="https://xiangyue9607.github.io/" style="text-decoration: none; color: inherit;">Xiang Yue*†</a>,
</span>
<span class="author-block">
<a href="https://yuanshengni.github.io/" style="text-decoration: none; color: inherit;">Yuansheng Ni*</a>,
</span>
<span class="author-block">
<a href="https://drogozhang.github.io/" style="text-decoration: none; color: inherit;">Kai Zhang*</a>,
</span>
<span class="author-block">
<a href="https://scholar.google.com/citations?hl=en&user=Vq-VZnUAAAAJ" style="text-decoration: none; color: inherit;">Tianyu Zheng*</a>,
</span>
<br>
<span class="author-block">Ruoqi Liu,</span>
<span class="author-block">Ge Zhang,</span>
<span class="author-block">Samuel Stevens,</span>
<span class="author-block">Dongfu Jiang,</span>
<span class="author-block">Weiming Ren,</span>
<span class="author-block">Yuxuan Sun,</span>
<span class="author-block">Cong Wei,</span>
<span class="author-block">Botao Yu,</span>
<span class="author-block">Ruibin Yuan,</span>
<span class="author-block">Renliang Sun,</span>
<span class="author-block">Ming Yin,</span>
<span class="author-block">Boyuan Zheng,</span>
<span class="author-block">Zhenzhu Yang,</span>
<span class="author-block">Yibo Liu,</span>
<span class="author-block">Wenhao Huang,</span><br>
<span class="author-block">
<a href="https://web.cse.ohio-state.edu/~sun.397/" style="text-decoration: none; color: inherit;">Huan Sun*</a>,
</span>
<span class="author-block">
<a href="https://ysu1989.github.io/" style="text-decoration: none; color: inherit;">Yu Su*†</a>,
</span>
<span class="author-block">
<a href="https://wenhuchen.github.io/" style="text-decoration: none; color: inherit;">Wenhu Chen*†</a>
</span>
</div>
<br>
<div class="is-size-5 publication-authors">
<span class="author-block"><b>MMMU Team</b></span>
</div>
<br>
<div class="is-size-5 publication-authors">
<span class="author-block">*Core Contributors</span><br>
<span class="author-block">†Corresponding to:</span>
<span class="author-block"><a href="mailto:[email protected]">[email protected]</a>,</span>
<span class="author-block"><a href="mailto:[email protected]">[email protected]</a>,</span>
<span class="author-block"><a href="mailto:[email protected]">[email protected]</a></span>
</div>
<div class="column has-text-centered">
<div class="publication-links">
<span class="link-block">
<a href="https://arxiv.org/abs/2311.16502" 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/MMMU/MMMU_Pro" class="external-link button is-normal is-rounded is-dark">
<span class="icon" style="font-size:18px">🤗</span>
<span>MMMU-Pro</span>
</a>
</span>
<span class="link-block">
<a href="https://huggingface.co/datasets/MMMU/MMMU" class="external-link button is-normal is-rounded is-dark">
<span class="icon" style="font-size:18px">🤗</span>
<span>MMMU</span>
</a>
</span>
<span class="link-block">
<a href="https://github.com/MMMU-Benchmark/MMMU" class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Code</span>
</a>
</span>
<span class="link-block">
<a href="#leaderboard" class="external-link button is-normal is-rounded is-dark">
<span class="icon has-text-white">
<i class="fa-solid fa-trophy"></i>
</span>
<span>Leaderboard</span>
</a>
</span>
<span class="link-block">
<a href="https://eval.ai/web/challenges/challenge-page/2179/overview" class="external-link button is-normal is-rounded is-dark">
<span class="icon has-text-white">
<i class="fa-solid fa-medal"></i>
</span>
<span>EvalAI</span>
</a>
</span>
<span class="link-block">
<a href="https://twitter.com/xiangyue96/status/1729698316554801358" class="external-link button is-normal is-rounded is-dark">
<span class="icon has-text-white">
<i class="fa-brands fa-x-twitter"></i>
</span>
<span>Twitter</span>
</a>
</span>
<span class="link-block">
<a href="#examples" class="external-link button is-normal is-rounded is-dark">
<span class="icon has-text-white">
<i class="fa-solid fa-book"></i>
</span>
<span>Examples</span>
</a>
</span>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<section class="hero teaser">
<div class="container is-max-desktop has-text-centered">
<img src="static/images/overview_mmlu.Jpeg" alt="geometric reasoning">
<p>Overview of the MMMU dataset. MMMU presents four challenges:
1) <b>comprehensiveness</b>: 11.5K college-level problems across six broad disciplines and 30 college subjects;
2) highly <b>heterogeneous</b> image types;
3) <b>interleaved</b> text and images;
4) <b>expert-level</b> perception and reasoning rooted in deep subject knowledge.
</p>
</div>
</section>
<section class="section">
<div class="container" style="margin-bottom: 2vh;">
<!-- Abstract. -->
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">🔔News</h2>
<div class="content has-text-justified">
<p>
<b>🔥[2024-09-05] Introducing <a href="https://arxiv.org/abs/2409.02813">MMMU-Pro</a>, a robust version of MMMU benchmark for multimodal AI evaluation! 🚀</b>
</p>
<p>
<b>🚀[2024-01-31]: We added Human Expert performance on the <a href="#leaderboard">Leaderboard</a>!🌟</b>
</p>
<p>
<b>🔥[2023-12-04]: Our evaluation server for the test set is now available on <a href="https://eval.ai/web/challenges/challenge-page/2179/overview"><b>EvalAI</b></a>. We welcome all submissions and look forward to your participation! 😆</b>
</p>
</div>
<h2 class="title is-3">Introduction</h2>
<div class="content has-text-justified">
<p>
We introduce MMMU: a new benchmark designed to evaluate multimodal models on massive multi-discipline tasks demanding college-level subject knowledge and deliberate reasoning. MMMU includes <b>11.5K</b> meticulously collected multimodal questions from college exams, quizzes, and textbooks, covering six core disciplines: Art & Design, Business, Science, Health & Medicine, Humanities & Social Science, and Tech & Engineering. These questions span <b>30</b> subjects and <b>183</b> subfields, comprising 30 highly heterogeneous image types, such as charts, diagrams, maps, tables, music sheets, and chemical structures. Unlike existing benchmarks, MMMU focuses on advanced perception and reasoning with domain-specific knowledge, challenging models to perform tasks akin to those faced by experts. Our evaluation of 14 open-source LMMs and the proprietary GPT-4V(ision) highlights the substantial challenges posed by MMMU. Even the advanced GPT-4V only achieves a 56% accuracy, indicating significant room for improvement. We believe MMMU will stimulate the community to build next-generation multimodal foundation models towards expert artificial general intelligence.
</p>
</div>
</div>
</div>
<!--/ Abstract. -->
</div>
</section>
<!-- DATASET SECTION -->
<section class="hero is-light is-small">
<div class="hero-body has-text-centered">
<h1 class="title is-1 mmmu">
<img src="static/images/mmmu_icon2.png" alt="Logo" class="mmmu-logo"/>
<span class="mmmu">MMMU Benchmark</span>
</h1>
</div>
</section>
<section class="section">
<div class="container">
<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>
We introduce the Massive Multi-discipline Multimodal Understanding and Reasoning (MMMU) benchmark, a novel benchmark meticulously curated to assess the expert-level multimodal understanding capability of foundation models across a broad scope of tasks. Covering subjects across disciplines, including Art, Business, Health & Medicine, Science, Humanities & Social Science, and Tech & Engineering, and over subfields. The detailed subject coverage and statistics are detailed in the figure. The questions in our benchmark were manually collected by a team of college students (including coauthors) from various disciplines and subjects, drawing from online sources, textbooks, and lecture materials.
</p>
<img src="static/images/mmlu_example.Jpeg" alt="algebraic reasoning" class="center">
<br>
<p>
MMMU is designed to measure three essential skills in LMMs: perception, knowledge, and reasoning. Our aim is to evaluate how well these models can not only perceive and understand information across different modalities but also apply reasoning with subject-specific knowledge to derive the solution.
</p>
<p>
Our MMMU benchmark introduces key challenges to multimodal foundation models, as detailed in a figure. Among these, we particularly highlight the challenge stemming from the requirement for both expert-level visual perceptual abilities and deliberate reasoning with subject-specific knowledge. This challenge is vividly illustrated through our tasks, which not only demand the processing of various heterogeneous image types but also necessitate a model's adeptness in using domain-specific knowledge to deeply understand both the text and images and to reason. This goes significantly beyond basic visual perception, calling for an advanced approach that integrates advanced multimodal analysis with domain-specific knowledge.
</p>
</div>
</div>
</div>
<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>
To further distinguish the difference between <i>dataset</i> and other existing ones, we elaborate the benchmark details in Figure.
From the <i>breadth</i> perspective, the prior benchmarks are heavily focused on daily knowledge and common sense.
The covered image format is also limited. Our benchmark aims to cover college-level knowledge with 30 image formats including diagrams,
tables, charts, chemical structures, photos, paintings, geometric shapes, music sheets, medical images, etc.
In the <i>depth</i> aspect, the previous benchmarks normally require commonsense knowledge or simple physical or temporal reasoning.
In contrast, our benchmark requires deliberate reasoning with college-level subject knowledge.
</p>
<div class="content has-text-centered">
<img src="static/images/compare.Jpeg" alt="algebraic reasoning" class="center">
<p> Sampled MMMU examples from each discipline. The questions and images need expert-level knowledge to understand and reason.</p>
</div>
</div>
</div>
</div>
<div class="columns is-centered m-6">
<div class="column is-full has-text-centered content">
<h2 class="title is-3">Statistics</h2>
<div class="carousel results-carousel">
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/mmmu_subject_distribution.Jpeg" alt="algebraic reasoning" width="95%"/>
<p> Sampled MMMU examples from each discipline. The questions and images need expert-level knowledge to understand and reason.</p>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/statistics.png" alt="arithmetic reasoning" width="40%"/>
<p> Key statistics of the MMMU benchmark</p>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/image_type_count.png" alt="arithmetic reasoning" width="80%"/>
<p> Distribution of image types in the MMMU dataset</p>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- RESULTS SECTION -->
<section class="hero is-light is-small">
<div class="hero-body has-text-centered">
<h1 class="title is-1 mmmu">Experiment Results</h1>
</div>
</section>
<section class="section">
<div class="container">
<!-------------------------------------------------------------------- RESULTS SECTION -------------------------------------------------------------------->
<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 has-text-justified">
<p>
We evaluate various models including LLMs and LMMs.
In each type, we consider both closed- and open-source models.
Our evaluation is conducted under a zero-shot setting to assess the capability of models to generate accurate answers without fine-tuning or few-shot demonstrations on our benchmark.
For all models, we use the default prompt provided by each model for multi-choice or open QA, if available.
If models do not provide prompts for task types in MMMU, we conduct prompt engineering on the validation set and use the most effective prompt for the later zero-shot experiment.
</p>
</div>
<br>
<div class="model-labels-container">
<span class="leaderboard-label human_expert">Human Expert</span>
<span class="leaderboard-label open_source">Open-Source</span>
<span class="leaderboard-label proprietary">Proprietary</span>
</div>
<br>
<div class="content has-text-centered">
<p>
Click on MMMU-Pro, MMMU (Val) or MMMU (Test) to expand detailed results.
</p>
</div>
<div class="leaderboard-container">
<div class="table-wrapper">
<table id="mmmu-table">
<thead>
<tr>
<th colspan="3" class="reset-cell clickable" style="text-align: center;">Reset</th>
<th class="pro-details-cell clickable" colspan="1">MMMU-Pro</th>
<th class="val-details-cell clickable" colspan="1">MMMU(Val)</th>
<th class="test-details-cell clickable" colspan="1">MMMU(Test)</th>
</tr>
<tr>
<th class="sortable clickable" data-sort="string">Name</th>
<th class="clickable" data-sort="string">Size</th>
<th class="sortable clickable" data-sort="date">Date</th>
<th class="sortable clickable pro-overall" data-sort="number">Overall</th>
<th class="hidden pro-details sortable clickable" data-sort="number">Vision</th>
<th class="hidden pro-details sortable clickable" data-sort="number">Standard</th>
<th class="sortable clickable val-overall" data-sort="number">Overall</th>
<th class="hidden val-details sortable clickable" data-sort="number">Art & Design</th>
<th class="hidden val-details sortable clickable" data-sort="number">Business</th>
<th class="hidden val-details sortable clickable" data-sort="number">Science</th>
<th class="hidden val-details sortable clickable" data-sort="number">Health & Medicine</th>
<th class="hidden val-details sortable clickable" data-sort="number">Human. & Social Sci.</th>
<th class="hidden val-details sortable clickable" data-sort="number">Tech & Eng.</th>
<th class="sortable clickable test-overall" data-sort="number">Overall</th>
<th class="hidden test-details sortable clickable" data-sort="number">Art & Design</th>
<th class="hidden test-details sortable clickable" data-sort="number">Business</th>
<th class="hidden test-details sortable clickable" data-sort="number">Science</th>
<th class="hidden test-details sortable clickable" data-sort="number">Health & Medicine</th>
<th class="hidden test-details sortable clickable" data-sort="number">Human. & Social Sci.</th>
<th class="hidden test-details sortable clickable" data-sort="number">Tech & Eng.</th>
</tr>
</thead>
<tbody>
<!-- Table body will be populated dynamically -->
</tbody>
</table>
<p class="test-desc"> Overall results of different models on the MMMU leaderboard. The best-performing model in each category is <b>in-bold</b>, and the second best is <u>underlined</u>. *: results provided by the authors.</p>
</div>
</div>
</div>
</div>
<!-------------------------------------------------------------------- Image Type SECTION -------------------------------------------------------------------->
<div class="columns is-centered m-6">
<div class="column is-full has-text-centered content">
<h2 class="title is-3">Different Image Types</h2>
<div class="content has-text-justified">
<p>
We compare the performance of various models across top frequent image types.
Across all types, GPT-4V consistently outperforms the other models by a huge margin.
Open-source models demonstrate relatively strong performance in categories like Photos and Paintings, which are more frequently seen during training.
However, for less common image categories like Geometric shapes, Music sheets and Chemical structures, all models obtain very low scores (some are close to random guesses).
This indicates that the existing models are generalizing poorly towards these image types.
</p>
</div>
<div class="model-labels-container">
<span class="model-label" style="background-color: rgba(196, 123, 160, 0.5);">Adept Fuyu-8B</span>
<span class="model-label" style="background-color: rgba(245, 123, 113, 0.5);">Qwen-VL-7B-Chat</span>
<span class="model-label" style="background-color: rgba(255, 208, 80, 0.5);">LLaVA-1.5-13B</span>
<span class="model-label" style="background-color: rgba(110, 194, 134, 0.5);">InstructBLIP-T5-XXL</span>
<span class="model-label" style="background-color: rgba(255, 153, 78, 0.5);">BLIP-2 FLAN-T5-XXL</span>
<span class="model-label" style="background-color: rgba(42, 149, 235, 0.5);">Yi-VL-34B</span>
<span class="model-label" style="background-color: rgba(183, 156, 220, 0.5);">LLaVA-1.6-34B</span>
<span class="model-label" style="background-color: rgba(143, 169, 209, 0.5);">InternVL-Chat-V1.2</span>
<span class="model-label" style="background-color: rgba(72, 199, 176, 0.5);">VILA1.5</span>
<span class="model-label" style="background-color: rgba(117, 209, 215, 0.5);">GPT-4V</span>
</div>
<div class="content has-text-centered">
<div class="chart-grid">
<!-- Chart 1: Diagrams -->
<div class="chart-item">
<canvas id="chart_Diagrams"></canvas>
<p class="chart-label">Diagrams (3184)</p>
</div>
<!-- Chart 2: Tables -->
<div class="chart-item">
<canvas id="chart_Tables"></canvas>
<p class="chart-label">Tables (2267)</p>
</div>
<!-- Chart 3: Plots and Charts -->
<div class="chart-item">
<canvas id="chart_PlotsAndCharts"></canvas>
<p class="chart-label">Plots and Charts (840)</p>
</div>
<!-- Chart 4: Chemical Structures -->
<div class="chart-item">
<canvas id="chart_ChemicalStructures"></canvas>
<p class="chart-label">Chemical Structures (573)</p>
</div>
<!-- Chart 5: Photographs -->
<div class="chart-item">
<canvas id="chart_Photographs"></canvas>
<p class="chart-label">Photographs (770)</p>
</div>
<!-- Chart 6: Paintings -->
<div class="chart-item">
<canvas id="chart_Paintings"></canvas>
<p class="chart-label">Paintings (453)</p>
</div>
<!-- Chart 7: Geometric Shapes -->
<div class="chart-item">
<canvas id="chart_GeometricShapes"></canvas>
<p class="chart-label">Geometric Shapes (336)</p>
</div>
<!-- Chart 8: Sheet Music -->
<div class="chart-item">
<canvas id="chart_SheetMusic"></canvas>
<p class="chart-label">Sheet Music (335)</p>
</div>
<!-- Chart 9: Medical Images -->
<div class="chart-item">
<canvas id="chart_MedicalImages"></canvas>
<p class="chart-label">Medical Images (272)</p>
</div>
<!-- Chart 10: Pathological Images -->
<div class="chart-item">
<canvas id="chart_PathologicalImages"></canvas>
<p class="chart-label">Pathological Images (253)</p>
</div>
<!-- Chart 11: Microscopic Images -->
<div class="chart-item">
<canvas id="chart_MicroscopicImages"></canvas>
<p class="chart-label">Microscopic Images (226)</p>
</div>
<!-- Chart 12: MRI, CT scans, and X-rays -->
<div class="chart-item">
<canvas id="chart_MRIsCTScansXrays"></canvas>
<p class="chart-label">MRI, CT scans, and X-rays (198)</p>
</div>
<!-- Chart 13: Sketches and Drafts -->
<div class="chart-item">
<canvas id="chart_SketchesAndDrafts"></canvas>
<p class="chart-label">Sketches and Drafts (184)</p>
</div>
<!-- Chart 14: Maps -->
<div class="chart-item">
<canvas id="chart_Maps"></canvas>
<p class="chart-label">Maps (170)</p>
</div>
<!-- Chart 15: Technical Blueprints -->
<div class="chart-item">
<canvas id="chart_TechnicalBlueprints"></canvas>
<p class="chart-label">Technical Blueprints (162)</p>
</div>
<!-- Chart 16: Trees and Graphs -->
<div class="chart-item">
<canvas id="chart_TreesAndGraphs"></canvas>
<p class="chart-label">Trees and Graphs (146)</p>
</div>
<!-- Chart 17: Mathematical Notations -->
<div class="chart-item">
<canvas id="chart_MathematicalNotations"></canvas>
<p class="chart-label">Mathematical Notations (133)</p>
</div>
<!-- Chart 18: Comics and Cartoons -->
<div class="chart-item">
<canvas id="chart_ComicsAndCartoons"></canvas>
<p class="chart-label">Comics and Cartoons (131)</p>
</div>
<!-- Chart 19: Sculpture -->
<div class="chart-item">
<canvas id="chart_Sculpture"></canvas>
<p class="chart-label">Sculpture (117)</p>
</div>
<!-- Chart 20: Portraits -->
<div class="chart-item">
<canvas id="chart_Portraits"></canvas>
<p class="chart-label">Portraits (91)</p>
</div>
<!-- Chart 21: Screenshots -->
<div class="chart-item">
<canvas id="chart_Screenshots"></canvas>
<p class="chart-label">Screenshots (70)</p>
</div>
<!-- Chart 22: Other -->
<div class="chart-item">
<canvas id="chart_Other"></canvas>
<p class="chart-label">Other(60)</p>
</div>
<!-- Chart 23: Poster -->
<div class="chart-item">
<canvas id="chart_Poster"></canvas>
<p class="chart-label">Poster(57)</p>
</div>
<!-- Chart 24: Icons and Symbols -->
<div class="chart-item">
<canvas id="chart_IconsAndSymbols"></canvas>
<p class="chart-label">Icons and Symbols (42)</p>
</div>
<!-- Chart 25: Historical Timelines -->
<div class="chart-item">
<canvas id="chart_HistoricalTimelines"></canvas>
<p class="chart-label">Historical Timelines (30)</p>
</div>
<!-- Chart 26: 3D Renderings -->
<div class="chart-item">
<canvas id="chart_3DRenderings"></canvas>
<p class="chart-label">3D Renderings (21)</p>
</div>
<!-- Chart 27: DNA Sequences -->
<div class="chart-item">
<canvas id="chart_DNASequences"></canvas>
<p class="chart-label">DNA Sequences (20)</p>
</div>
<!-- Chart 28: Landscapes -->
<div class="chart-item">
<canvas id="chart_Landscapes"></canvas>
<p class="chart-label">Landscapes (16)</p>
</div>
<!-- Chart 29: Logos and Branding -->
<div class="chart-item">
<canvas id="chart_LogosAndBranding"></canvas>
<p class="chart-label">Logos and Branding(14)</p>
</div>
<!-- Chart 30: Advertisements -->
<div class="chart-item">
<canvas id="chart_Advertisements"></canvas>
<p class="chart-label">Advertisements (10)</p>
</div>
</div>
<p class="bottom-text"> Selected models' performance on 30 different image types. Note that a single image may have multiple image types.</p>
</div>
</div>
</div>
<!-------------------------------------------------------------------- Difficulty Levels SECTION -------------------------------------------------------------------->
<div class="columns is-centered m-6">
<div class="column is-full has-text-centered content">
<h2 class="title is-3">Different Difficulty Levels</h2>
<div class="content has-text-justified">
<p>
we compares the performance of selected models across three difficulty levels.
GPT-4V demonstrates a significantly higher proficiency, with a success rate of 76.1%, compared to opensource models in the “Easy” category.
When it comes to the “Medium” category, while the gap narrows, GPT-4V still leads at 55.6%.
The further diminishing performance gap in the “Hard” category across models indicates that as the complexity of tasks increases, the advantage of more advanced models like GPT-4V almost disappears.
This might reflect a current limitation in handling expert-level challenging queries even for the most advanced models.
</p>
</div>
<div class="content has-text-centered">
<p>Click legend to switch the comparison chart.</p>
</div>
<div class="content has-text-centered">
<canvas id="difficulty_level_chart"></canvas>
<p>Result decomposition across question difficulty levels.</p>
</div>
</div>
</div>
<!-------------------------------------------------------------------- Single VS Multiple image SECTION -------------------------------------------------------------------->
<div class="columns is-centered m-6">
<div class="column is-full has-text-centered content">
<h2 class="title is-3">Single Image VS Multiple Image</h2>
<div class="content has-text-centered">
<div class="content has-text-centered">
<p>Click legend to switch the comparison chart.</p>
</div>
<canvas id="single_vs_multiple_chart"></canvas>
<p>Result decomposition across single image and multiple image tasks.</p>
</div>
</div>
</div>
<!-------------------------------------------------------------------- Error Analysis SECTION -------------------------------------------------------------------->
<div class="columns is-centered m-6">
<div class="column is-full has-text-centered content">
<h2 class="title is-3">Error Analysis</h2>
<div class="content has-text-justified">
<p>
We delve into the analysis of errors by GPT-4V, a pivotal aspect for understanding its operational capabilities and limitations.
This analysis serves not only to identify the model's current shortcomings but also to guide future enhancements in its design and training.
We meticulously examine 150 randomly sampled error instances from GPT-4V's predictions.
These instances are analyzed by expert annotators who identify the root causes of mispredictions based on their knowledge and the golden explanations if available.
The distribution of these errors is illustrated in Figure, and a selection of 100 notable cases, along with detailed analyses, is included in the Appendix.
</p>
</div>
<div class="content has-text-centered">
<img src="static/images/error_distribution_1.Jpeg" alt="error distribution" width="25%">
<p> Error distribution over 150 annotated GPT-4V errors.</p>
</div>
</div>
</div>
<!-------------------------------------------------------------------- Error Example -------------------------------------------------------------------->
<div class="columns is-centered m-6">
<div class="column is-full has-text-centered content">
<h2 class="title is-3" id="examples">Error Examples</h2>
<div class="carousel results-carousel">
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/1.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/2.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/3.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/4.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/5.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/6.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/7.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/8.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/9.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/10.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/11.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/12.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/13.png" alt="grade-lv" width="60%"/>
</div>
</div>
<!-- <div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/14.png" alt="grade-lv" width="60%"/>
</div>
</div> -->
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/15.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/16.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/17.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/18.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/19.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/20.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/21.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/22.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/23.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/24.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/25.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/26.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/27.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/28.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/29.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/30.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/31.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/32.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/33.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/34.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/35.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/36.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/37.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/38.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/39.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/40.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/41.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/42.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/43.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/44.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/45.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/46.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/47.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/48.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/49.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/error/50.png" alt="grade-lv" width="60%"/>
</div>
</div>
</div>
</div>
</div>
<!-------------------------------------------------------------------- Correct Example -------------------------------------------------------------------->
<div class="columns is-centered m-6">
<div class="column is-full has-text-centered content">
<h2 class="title is-3">Correct 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/correct/1.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/2.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/3.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/4.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/5.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/6.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/7.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/8.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/9.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/10.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/11.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/12.png" alt="grade-lv" width="60%"/>
</div>
</div>
<!-- <div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/13.png" alt="grade-lv" width="60%"/>
</div>
</div> -->
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/14.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/15.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/16.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/17.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/18.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/19.png" alt="grade-lv" width="60%"/>
</div>
</div>
<div class="box m-5">
<div class="content has-text-centered">
<img src="static/images/correct/20.png" alt="grade-lv" width="60%"/>
</div>