-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathindex.html
879 lines (763 loc) · 31 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
<!DOCTYPE html>
<html>
<head>
<title>AntGPT</title>
<style>
body {
background-color: white;
font-family: 'Roboto', sans-serif;
font-weight: 100;
}
#title-banner {
background-color: white;
padding-right: 200px;
padding-left: 200px;
font-size: 20px;
color: black;
text-align: center;
}
#bibtex-banner {
margin: 20px;
padding-top: 50px;
background-color: white;
padding-right: 100px;
padding-left: 100px;
color: black;
}
.bibtex {
background-color: #F5F5F5;
padding: 20px;
}
#abstract {
margin: 19px;
padding-right: 300px;
padding-left: 300px;
}
.abstract-title {
font-size: 30px;
text-align: center;
}
#main-visualization {
margin: 20px;
}
figcaption {
font-size: 12px;
font-style: italic;
/* Italicize the caption (optional) */
color: #666;
/* Set the color of the caption (optional) */
}
#result-visualization {
margin: 20px;
padding-top: 50px;
padding-left: 100px;
padding-right: 100px;
}
.image-row {
display: flex;
padding-right: 30px
}
.main-image {
width: 60%;
/*120vh;*/
background-size: cover;
border-radius: 5px;
display: block;
padding-left: 20px;
margin-left: auto;
margin-right: auto;
}
.interpolation-image2 {
width: 100%;
/*120vh;*/
background-size: cover;
border-radius: 5px;
/*display: block;*/
margin-top: auto;
margin-left: auto;
}
.interpolation-image3 {
height: 100%;
/*120vh;*/
width: 90%;
background-size: cover;
border-radius: 5px;
display: block;
margin-left: auto;
margin-right: auto;
}
.interpolation-image4 {
height: 100%;
/*120vh;*/
width: 100%;
/*120vh;*/
background-size: cover;
border-radius: 5px;
display: block;
margin-left: auto;
margin-right: auto;
}
.interpolation-image5 {
height: 100%;
/*120vh;*/
width: 90%;
/*120vh;*/
background-size: cover;
border-radius: 5px;
display: block;
margin-left: auto;
margin-right: auto;
}
.button {
background-color: blue;
border: none;
color: white;
padding: 10px 20px;
text-align: center;
text-decoration: none;
display: inline-block;
font-size: 16px;
margin: 4px 2px;
cursor: pointer;
border-radius: 40px;
/* Rounded corners */
}
#links {
text-align: center;
/* Center the buttons */
}
#authors a {
color: blue;
text-decoration: none;
/* No underline */
}
#authors a:hover {
text-decoration: underline;
/* Underline on hover */
}
</style>
<style type="text/css">
svg:not(:root).svg-inline--fa {
overflow: visible
}
.svg-inline--fa {
display: inline-block;
font-size: inherit;
height: 1em;
overflow: visible;
vertical-align: -.125em
}
.svg-inline--fa.fa-lg {
vertical-align: -.225em
}
.svg-inline--fa.fa-w-1 {
width: .0625em
}
.svg-inline--fa.fa-w-2 {
width: .125em
}
.svg-inline--fa.fa-w-3 {
width: .1875em
}
.svg-inline--fa.fa-w-4 {
width: .25em
}
.svg-inline--fa.fa-w-5 {
width: .3125em
}
.svg-inline--fa.fa-w-6 {
width: .375em
}
.svg-inline--fa.fa-w-7 {
width: .4375em
}
.svg-inline--fa.fa-w-8 {
width: .5em
}
.svg-inline--fa.fa-w-9 {
width: .5625em
}
.svg-inline--fa.fa-w-10 {
width: .625em
}
.svg-inline--fa.fa-w-11 {
width: .6875em
}
.svg-inline--fa.fa-w-12 {
width: .75em
}
.svg-inline--fa.fa-w-13 {
width: .8125em
}
.svg-inline--fa.fa-w-14 {
width: .875em
}
.svg-inline--fa.fa-w-15 {
width: .9375em
}
.svg-inline--fa.fa-w-16 {
width: 1em
}
.svg-inline--fa.fa-w-17 {
width: 1.0625em
}
.svg-inline--fa.fa-w-18 {
width: 1.125em
}
.svg-inline--fa.fa-w-19 {
width: 1.1875em
}
.svg-inline--fa.fa-w-20 {
width: 1.25em
}
.svg-inline--fa.fa-pull-left {
margin-right: .3em;
width: auto
}
.svg-inline--fa.fa-pull-right {
margin-left: .3em;
width: auto
}
.svg-inline--fa.fa-border {
height: 1.5em
}
.svg-inline--fa.fa-li {
width: 2em
}
.svg-inline--fa.fa-fw {
width: 1.25em
}
.fa-layers svg.svg-inline--fa {
bottom: 0;
left: 0;
margin: auto;
position: absolute;
right: 0;
top: 0
}
.fa-layers {
display: inline-block;
height: 1em;
position: relative;
text-align: center;
vertical-align: -.125em;
width: 1em
}
.fa-layers svg.svg-inline--fa {
-webkit-transform-origin: center center;
transform-origin: center center
}
.fa-layers-counter,
.fa-layers-text {
display: inline-block;
position: absolute;
text-align: center
}
.fa-layers-text {
left: 50%;
top: 50%;
-webkit-transform: translate(-50%, -50%);
transform: translate(-50%, -50%);
-webkit-transform-origin: center center;
transform-origin: center center
}
.fa-layers-counter {
background-color: #ff253a;
border-radius: 1em;
-webkit-box-sizing: border-box;
box-sizing: border-box;
color: #fff;
height: 1.5em;
line-height: 1;
max-width: 5em;
min-width: 1.5em;
overflow: hidden;
padding: .25em;
right: 0;
text-overflow: ellipsis;
top: 0;
-webkit-transform: scale(.25);
transform: scale(.25);
-webkit-transform-origin: top right;
transform-origin: top right
}
.fa-layers-bottom-right {
bottom: 0;
right: 0;
top: auto;
-webkit-transform: scale(.25);
transform: scale(.25);
-webkit-transform-origin: bottom right;
transform-origin: bottom right
}
.fa-layers-bottom-left {
bottom: 0;
left: 0;
right: auto;
top: auto;
-webkit-transform: scale(.25);
transform: scale(.25);
-webkit-transform-origin: bottom left;
transform-origin: bottom left
}
.fa-layers-top-right {
right: 0;
top: 0;
-webkit-transform: scale(.25);
transform: scale(.25);
-webkit-transform-origin: top right;
transform-origin: top right
}
.fa-layers-top-left {
left: 0;
right: auto;
top: 0;
-webkit-transform: scale(.25);
transform: scale(.25);
-webkit-transform-origin: top left;
transform-origin: top left
}
.fa-lg {
font-size: 1.3333333333em;
line-height: .75em;
vertical-align: -.0667em
}
.fa-xs {
font-size: .75em
}
.fa-sm {
font-size: .875em
}
.fa-1x {
font-size: 1em
}
.fa-2x {
font-size: 2em
}
.fa-3x {
font-size: 3em
}
.fa-4x {
font-size: 4em
}
.fa-5x {
font-size: 5em
}
.fa-6x {
font-size: 6em
}
.fa-7x {
font-size: 7em
}
.fa-8x {
font-size: 8em
}
.fa-9x {
font-size: 9em
}
.fa-10x {
font-size: 10em
}
.fa-fw {
text-align: center;
width: 1.25em
}
.fa-ul {
list-style-type: none;
margin-left: 2.5em;
padding-left: 0
}
.fa-ul>li {
position: relative
}
.fa-li {
left: -2em;
position: absolute;
text-align: center;
width: 2em;
line-height: inherit
}
.fa-border {
border: solid .08em #eee;
border-radius: .1em;
padding: .2em .25em .15em
}
.fa-pull-left {
float: left
}
.fa-pull-right {
float: right
}
.fa.fa-pull-left,
.fab.fa-pull-left,
.fal.fa-pull-left,
.far.fa-pull-left,
.fas.fa-pull-left {
margin-right: .3em
}
.fa.fa-pull-right,
.fab.fa-pull-right,
.fal.fa-pull-right,
.far.fa-pull-right,
.fas.fa-pull-right {
margin-left: .3em
}
.fa-spin {
-webkit-animation: fa-spin 2s infinite linear;
animation: fa-spin 2s infinite linear
}
.fa-pulse {
-webkit-animation: fa-spin 1s infinite steps(8);
animation: fa-spin 1s infinite steps(8)
}
@-webkit-keyframes fa-spin {
0% {
-webkit-transform: rotate(0);
transform: rotate(0)
}
100% {
-webkit-transform: rotate(360deg);
transform: rotate(360deg)
}
}
@keyframes fa-spin {
0% {
-webkit-transform: rotate(0);
transform: rotate(0)
}
100% {
-webkit-transform: rotate(360deg);
transform: rotate(360deg)
}
}
.fa-rotate-90 {
-webkit-transform: rotate(90deg);
transform: rotate(90deg)
}
.fa-rotate-180 {
-webkit-transform: rotate(180deg);
transform: rotate(180deg)
}
.fa-rotate-270 {
-webkit-transform: rotate(270deg);
transform: rotate(270deg)
}
.fa-flip-horizontal {
-webkit-transform: scale(-1, 1);
transform: scale(-1, 1)
}
.fa-flip-vertical {
-webkit-transform: scale(1, -1);
transform: scale(1, -1)
}
.fa-flip-both,
.fa-flip-horizontal.fa-flip-vertical {
-webkit-transform: scale(-1, -1);
transform: scale(-1, -1)
}
:root .fa-flip-both,
:root .fa-flip-horizontal,
:root .fa-flip-vertical,
:root .fa-rotate-180,
:root .fa-rotate-270,
:root .fa-rotate-90 {
-webkit-filter: none;
filter: none
}
.fa-stack {
display: inline-block;
height: 2em;
position: relative;
width: 2.5em
}
.fa-stack-1x,
.fa-stack-2x {
bottom: 0;
left: 0;
margin: auto;
position: absolute;
right: 0;
top: 0
}
.svg-inline--fa.fa-stack-1x {
height: 1em;
width: 1.25em
}
.svg-inline--fa.fa-stack-2x {
height: 2em;
width: 2.5em
}
.fa-inverse {
color: #fff
}
.sr-only {
border: 0;
clip: rect(0, 0, 0, 0);
height: 1px;
margin: -1px;
overflow: hidden;
padding: 0;
position: absolute;
width: 1px
}
.sr-only-focusable:active,
.sr-only-focusable:focus {
clip: auto;
height: auto;
margin: 0;
overflow: visible;
position: static;
width: auto
}
.svg-inline--fa .fa-primary {
fill: var(--fa-primary-color, currentColor);
opacity: 1;
opacity: var(--fa-primary-opacity, 1)
}
.svg-inline--fa .fa-secondary {
fill: var(--fa-secondary-color, currentColor);
opacity: .4;
opacity: var(--fa-secondary-opacity, .4)
}
.svg-inline--fa.fa-swap-opacity .fa-primary {
opacity: .4;
opacity: var(--fa-secondary-opacity, .4)
}
.svg-inline--fa.fa-swap-opacity .fa-secondary {
opacity: 1;
opacity: var(--fa-primary-opacity, 1)
}
.svg-inline--fa mask .fa-primary,
.svg-inline--fa mask .fa-secondary {
fill: #000
}
.fad.fa-inverse {
color: #fff
}
</style>
<link href="./VoxPoser_files/css" rel="stylesheet">
</head>
<body>
<div id="title-banner">
<h1 id="paper-title">AntGPT: Can Large Language Models Help Long-term Action Anticipation from Videos?</h1>
<p id="authors" style="padding-right: 180px; padding-left: 180px">
<a href="https://kevinz8866.github.io/">Qi Zhao<sup>*</sup><sup>1</sup></a>,
<a href="https://wang-sj16.github.io/">Shijie Wang<sup>*</sup><sup>1</sup></a>,
<a href="mailto:[email protected]">Ce Zhang<sup>1</sup></a>,
<a href="https://www.linkedin.com/in/changcheng-fu/">Changcheng Fu<sup>1</sup></a>,
<a href="https://www.linkedin.com/in/minh-quan-do-887419195/">Minh Quan Do<sup>1</sup></a>,
<a href="https://lukan94.github.io/">Nakul Agarwal<sup>2</sup></a>,
<a href="https://scholar.google.com/citations?user=C6Wu8M0AAAAJ&hl=en">Kwonjoon Lee<sup>2</sup></a>,
<a href="https://chensun.me/">Chen Sun<sup>1</sup></a>
</p>
<p id="affiliations">
<sup>1</sup>Brown University,
<sup>2</sup>Honda Research Institute
</p>
</div>
<div id="links">
<a href="https://arxiv.org/abs/2307.16368" class="button">
<span class="icon">
<svg class="svg-inline--fa fa-file fa-w-12" aria-hidden="true" focusable="false" data-prefix="fas"
data-icon="file" role="img" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 384 512"
data-fa-i2svg="">
<path fill="currentColor"
d="M224 136V0H24C10.7 0 0 10.7 0 24v464c0 13.3 10.7 24 24 24h336c13.3 0 24-10.7 24-24V160H248c-13.2 0-24-10.8-24-24zm160-14.1v6.1H256V0h6.1c6.4 0 12.5 2.5 17 7l97.9 98c4.5 4.5 7 10.6 7 16.9z">
</path>
</svg><!-- <i class="fas fa-file"></i> Font Awesome fontawesome.com -->
</span>
<span>Paper</span>
</a>
<a href="https://youtu.be/UiHnVakrE-0" class="button">
<span class="icon">
<svg class="svg-inline--fa fa-youtube fa-w-16" aria-hidden="true" focusable="false" data-prefix="fab"
data-icon="youtube" role="img" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 496 512"
data-fa-i2svg="">
<path fill="currentColor"
d="M549.7 124.1c-6.3-23.7-24.8-42.3-48.3-48.6C458.8 64 288 64 288 64S117.2 64 74.6 75.5c-23.5 6.3-42 24.9-48.3 48.6-11.4 42.9-11.4 132.3-11.4 132.3s0 89.4 11.4 132.3c6.3 23.7 24.8 41.5 48.3 47.8C117.2 448 288 448 288 448s170.8 0 213.4-11.5c23.5-6.3 42-24.2 48.3-47.8 11.4-42.9 11.4-132.3 11.4-132.3s0-89.4-11.4-132.3zm-317.5 213.5V175.2l142.7 81.2-142.7 81.2z">
</path>
</svg><!-- <i class="fas fa-file"></i> Font Awesome fontawesome.com -->
</span>
<span>Video</span>
</a>
<a href="https://github.com/brown-palm/AntGPT/tree/main" class="button">
<span class="icon">
<svg class="svg-inline--fa fa-github fa-w-16" aria-hidden="true" focusable="false" data-prefix="fab"
data-icon="github" role="img" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 496 512"
data-fa-i2svg="">
<path fill="currentColor"
d="M165.9 397.4c0 2-2.3 3.6-5.2 3.6-3.3.3-5.6-1.3-5.6-3.6 0-2 2.3-3.6 5.2-3.6 3-.3 5.6 1.3 5.6 3.6zm-31.1-4.5c-.7 2 1.3 4.3 4.3 4.9 2.6 1 5.6 0 6.2-2s-1.3-4.3-4.3-5.2c-2.6-.7-5.5.3-6.2 2.3zm44.2-1.7c-2.9.7-4.9 2.6-4.6 4.9.3 2 2.9 3.3 5.9 2.6 2.9-.7 4.9-2.6 4.6-4.6-.3-1.9-3-3.2-5.9-2.9zM244.8 8C106.1 8 0 113.3 0 252c0 110.9 69.8 205.8 169.5 239.2 12.8 2.3 17.3-5.6 17.3-12.1 0-6.2-.3-40.4-.3-61.4 0 0-70 15-84.7-29.8 0 0-11.4-29.1-27.8-36.6 0 0-22.9-15.7 1.6-15.4 0 0 24.9 2 38.6 25.8 21.9 38.6 58.6 27.5 72.9 20.9 2.3-16 8.8-27.1 16-33.7-55.9-6.2-112.3-14.3-112.3-110.5 0-27.5 7.6-41.3 23.6-58.9-2.6-6.5-11.1-33.3 2.6-67.9 20.9-6.5 69 27 69 27 20-5.6 41.5-8.5 62.8-8.5s42.8 2.9 62.8 8.5c0 0 48.1-33.6 69-27 13.7 34.7 5.2 61.4 2.6 67.9 16 17.7 25.8 31.5 25.8 58.9 0 96.5-58.9 104.2-114.8 110.5 9.2 7.9 17 22.9 17 46.4 0 33.7-.3 75.4-.3 83.6 0 6.5 4.6 14.4 17.3 12.1C428.2 457.8 496 362.9 496 252 496 113.3 383.5 8 244.8 8zM97.2 352.9c-1.3 1-1 3.3.7 5.2 1.6 1.6 3.9 2.3 5.2 1 1.3-1 1-3.3-.7-5.2-1.6-1.6-3.9-2.3-5.2-1zm-10.8-8.1c-.7 1.3.3 2.9 2.3 3.9 1.6 1 3.6.7 4.3-.7.7-1.3-.3-2.9-2.3-3.9-2-.6-3.6-.3-4.3.7zm32.4 35.6c-1.6 1.3-1 4.3 1.3 6.2 2.3 2.3 5.2 2.6 6.5 1 1.3-1.3.7-4.3-1.3-6.2-2.2-2.3-5.2-2.6-6.5-1zm-11.4-14.7c-1.6 1-1.6 3.6 0 5.9 1.6 2.3 4.3 3.3 5.6 2.3 1.6-1.3 1.6-3.9 0-6.2-1.4-2.3-4-3.3-5.6-2z">
</path>
</svg><!-- <i class="fab fa-github"></i> Font Awesome fontawesome.com -->
</span>
<span>Code</span>
</a>
<!-- Add more buttons as needed -->
</div>
<div id="main-visualization">
<img src="./assets/main.gif" class="main-image">
</div>
<div id="abstract">
<h2 class="abstract-title">Abstract</h2>
<p>Can we better anticipate an actor's future actions (e.g. <i>mix eggs</i>) by knowing what commonly
happens after the current action (e.g. <i>crack eggs</i>)? What if the actor also shares the goal
(e.g. <i>make fried rice</i>) with us? The long-term action anticipation (LTA) task aims to predict
an actor's future behavior from video observations in the form of verb and noun sequences, and it
is crucial for human-machine interaction. We propose to formulate the LTA task from two perspectives:
a <i>bottom-up</i> approach that predicts the next actions autoregressively by modeling temporal dynamics;
and a <i>top-down</i> approach that infers the goal of the actor and <i>plans</i> the needed procedure
to accomplish the goal. We hypothesize that large language models (LLMs), which have been pretrained
on procedure text data (e.g. recipes, how-tos), have the potential to help LTA from both perspectives.
It can help provide the prior knowledge on the possible next actions, and infer the goal given the observed
part of a procedure, respectively. We propose <strong>AntGPT</strong>, which represents video observations
as sequences
of human actions, and uses the action representation for an LLM to infer the goals and model temporal
dynamics.
<strong>AntGPT</strong> achieves state-of-the-art performance on Ego4D LTA v1 and v2, EPIC-Kitchens-55, as
well
as EGTEA GAZE+, thanks to LLMs' goal inference and temporal dynamics modeling capabilities. We further
demonstrate that these capabilities can be effectively distilled into a compact neural network 1.3% of
the original LLM model size.
</p>
</div>
<div id="result-visualization">
<h2>AntGPT and its novel capabilities</h2>
<div class="image-row">
<div style="text-align: center; margin-right: 10px;">
<img src="./assets/antgpt_model2.png" class="interpolation-image2" alt="Result Visualization">
<figcaption>A illustration of how AntGPT utilize language models for LTA.</figcaption>
</div>
<div style="text-align: center;">
<img src="./assets/antgpttransformer.png" class="interpolation-image2" alt="Result Visualization">
<figcaption>An illustration of AntGPT's top-down framework.</figcaption>
</div>
</div>
<p><strong>AntGPT</strong> is a vision-language framework that aims to explore how to incorporate the emergent
capabilities of large language models into video long-term action anticipation (LTA). The LTA task is
essentially given a video observation of human actions, to anticipate what is the future actions of the
actor. To represent video information for LLMs, we use action recognition models to represent video
observations into discrete action labels. They bridge visual information and language, allowing LLMs to
perform the downstream reasoning tasks. We first query the LLM to infer the goals behind the observed
actions. Then we incorporate the goal information into a vision-only pipeline to see if such
goal-conditioned prediction is helpful. We also used LLM to directly model the temporal dynamics of human
activity to see if LLM has built-in knowledge about action priors. Last but not least, we tested the
capability of LLMs to perform the prediction tasks in a few-shot settings, using popular prompting
strategies such as "Chain-of-Thought".</p>
<p><strong>AntGPT</strong> shows novel capabilities:
<li><strong>Predicting goals through few-shot observations</strong>: We observe that LLMs are very capable
of predicting the goals of the actor even given imperfect observed human actions. In context, we
demonstrate a few successful examples in which the correct actions and goals are given. Then in the
query, we input the observed actions sequences and let LLM output the goals. </li>
<li><strong>Augmenting vision frameworks with goal information</strong>: To testify whether the output goals
are helpful for the LTA task, we encoder goal information into textual features and incorporate it into
a vision framework to perform "goal-conditioned" future prediction and observed improvement to the
state-of-the-art.</li>
<li><strong>Modeling action temporal dynamics</strong>: We then explored if LLMs can directly act as a
reasoning backbone to model the temporal action dynamics. To this end, we fine-tuned LLMs on the
in-domain action sequences and observed that LLMs can bring additional improvement than a transformer
trained from-scratch. </li>
<li><strong>Predicting future actions in few-shot setting</strong>: We further explored how do LLMs perform
on the LTA task in a few-shot setting. When only demonstrating a few examples in the context, LLMs can
still predict the future action sequences. Furthermore, we experimented with popular prompting
strategies.</li>
</div>
<div id="result-visualization">
<h2>Can LLMs Infer Goals to Assist Top-down ?</h2>
<div class="image-row">
<div style="text-align: center; margin-top: 0px; margin-right: 10px; margin-bottom: 10px">
<img src="./assets/goal_visualization_fig2.jpg" class="interpolation-image3" alt="Result Visualization">
<figcaption>Examples of the fine-tuned model outputs.</figcaption>
</div>
<div style="text-align: center; margin-bottom: 10px">
<img src="./assets/cf1.png" class="interpolation-image4" alt="Result Visualization">
<figcaption>Examples of in-context learning outputs.</figcaption>
</div>
</div>
<p></p>
<li><strong>What is a good interface between videos and LLMs for the LTA task?</strong>: We experimented with
various preprocessing techniques and found that representing video segments as discrete action labels are
quite performant to interact with the LLMs, allowing LLMs to perform downstream reasoning from video
observations.</li>
<li><strong>Can LLMs infer the goals and are they helpful for top-down LTA?</strong>: We find that LLMs can
infer the goals and they are particularly helpful for goal-conditioned top-down LTA. As demonstrated in our
experiment, our goal-conditioned framework consistently performs better than our vision-only frameworks.
</li>
<li><strong>Would knowing the goals affect the action predicted by LLMs?</strong>: We did an interesting
qualitative experiment to see how about would goal affect the action prediction after we conclude goals
inferred by LLMs are useful. If we give an alternative goal instead of the truly inferred goal, how would
the output of LLM be affected? We observed that LLMs do respond according to the goal. For example, when we
switch the inferred goal "fix machines" into "examine machines", LLMs will predict some actions that are
exclusively related to "examine machines" like "read gauge", "record data", etc.</li>
</div>
<div id="result-visualization">
<h2>Do LLMs Model Temporal Dynamics? </h2>
<div class='image-row'>
<div style="text-align: center; margin-bottom: 10px; margin-right: 10px">
<img src="./assets/ft1.png" class="interpolation-image3" alt="Result Visualization">
<figcaption>Examples of the fine-tuned model outputs.</figcaption>
</div>
<div style="text-align: center; margin-bottom: 10px">
<img src="./assets/ft2.png" class="interpolation-image3" alt="Result Visualization">
<figcaption>Examples of in-context learning outputs.</figcaption>
</div>
</div>
<p></p>
<li><strong>LLMs are able to model temporal dynamics</strong>: We found that fine-tuned LLM are better reasoning
backbones than similar transformers trained from-scratch. Even with imperfect output structure and rough
post-processing, LLMs still outperform their transformer peers.</li>
<li><strong>LLM-based temporal model performs implicit goal inference</strong>: We found that fine-tuned LLMs do
implicit goal inference as when we add inferred goal to the fine-tuned LLMs the performance does not
increase compare to their bottom-up counterparts.</li>
<li><strong>Language prior encoded by LLMs benefit LTA</strong>:Through semantic perbutation experiments, we
found that with other symbolic representation of actions, the temporal dynamics modeling ability decreases
indicating that LLMs leverage the language prior to achieve better temporal dynamics modeling.
<li><strong>LLM-encoded knowledge can be condensed into a compact model</strong>: We conduct knowledge
distillation to train a small student model using the fine-tuned LLM as a teacher model and observed the
knowledge can be transferred and the student model can outperform the teacher model. </li>
</div>
<div id="result-visualization">
<h2>Few-shot predictions </h2>
<div class='image-row'>
<!-- <div style="text-align: center; margin-right: 10px; margin-bottom: 10px">
<img src="./assets/cf1.png" class="interpolation-image5" alt="Result Visualization">
<figcaption>Examples of counterfactual predictions.</figcaption>
</div> -->
<div style="text-align: center; margin-bottom: 10px;">
<img src="./assets/icl.png" class="interpolation-image5" alt="Result Visualization">
<figcaption>Illustration of few-shot goal inference and LTA with LLMs.</figcaption>
</div>
</div>
<p>
Beyond fine-tuning, we are also interested in understanding if LLM's in-context learning capability
generalizes to the LTA task. Compared with fine-tuning the model with the whole training set, in-context
learning avoids updating the weights of a pre-trained LLM.
</p>
<p>
An ICL prompt consists of three parts:
</p>
<ol>
<li>An instruction that specifies the anticipating action task, the output format, and the verb and noun
vocabulary.</li>
<li>The in-context examples randomly sampled from the training set. They are in the format of
<code>"<observed actions> => <future actions>"</code> with ground-truth actions.</li>
<li>Finally, the query in the format <code>"<observed actions> => "</code> with recognized actions.
</li>
</ol>
<p>
We also attempt to leverage chain-of-thoughts prompts (CoT) to ask the LLM to first infer the goal, then
perform LTA conditioned on the inferred goal. We observe that all LLM-based methods perform much better than
the transformer baseline for few-shot LTA. Among all the LLM-based methods, top-down prediction with CoT
prompts achieves the best performance on both verb and noun. However, the gain of explicitly using goal
conditioning is marginal, similar to what we have observed when training on the full set.</p>
</div>
<div id="bibtex-banner">
<h2>BibTeX</h2>
<pre class="bibtex"><code>@article{zhao2023antgpt,
title={AntGPT: Can Large Language Models Help Long-term Action Anticipation from Videos?},
author={Qi Zhao and Shijie Wang and Ce Zhang and Changcheng Fu and Minh Quan Do and Nakul Agarwal and Kwonjoon Lee and Chen Sun},
journal={ICLR},
year={2024}
}</code></pre>
</div>
</body>
</html>