-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathindex.html
More file actions
834 lines (747 loc) · 38.8 KB
/
index.html
File metadata and controls
834 lines (747 loc) · 38.8 KB
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
<!DOCTYPE HTML>
<html lang="en">
<head>
<!-- Google analytics tag (gtag.js) -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-WRYQ3GG5Y8"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'G-WRYQ3GG5Y8');
</script>
<link rel="icon" type="image/png" href="images/favicon.png" />
<title>Kay - Physical Intelligence</title>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1" />
<!-- Isotope JS -->
<script src="https://cdnjs.cloudflare.com/ajax/libs/jquery/3.6.1/jquery.min.js"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/jqueryui/1.13.2/jquery-ui.min.js"></script>
<script src="https://unpkg.com/isotope-layout@3/dist/isotope.pkgd.min.js"></script>
<!-- Custom Style -->
<link rel="stylesheet" href="style.css">
<!-- Google Font -->
<link rel="preconnect" href="https://fonts.googleapis.com">
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
<link
href="https://fonts.googleapis.com/css2?family=Asap:ital,wght@0,100;0,200;0,300;0,400;0,500;0,600;0,700;1,100;1,200;1,300;1,400;1,500;1,600;1,700&display=swap"
rel="stylesheet">
</head>
<body id="body">
<div id="main">
<div id="intro">
<div id="intro-text">
<h1>Kay - Liyiming Ke</h1>
<p>
Hi 👋 I work at <a href="https://www.physicalintelligence.company/">Physical Intelligence</a>
researching on Machine Learning for Robot Manipulation.
During my PhD at University of Washington, I built a chopsticks-welding robot to showcase data-driven fine
motor skills.
My path to robotics started unconventionally—I majored in Economics before diving into AI, with internships at
Meta AI, Microsoft Research, and Google Search along the way. I’m driven by curiosity and currently I aim to
design robot policies that master Robustness, Precision, and Dexterity.
<br><br>
<div id="more-bio" style="display: none;">
<br>
<p>Liyiming Ke is a full stack robotist at Physical Intelligence researching on Machine Learning for Robot
Manipulation. She earned her Ph.D. from the University of Washington with her thesis titled "Data-driven
Fine Manipulation". She built a chopsticks-welding robot that demonstrate fine motor skills and developed
theoretical frameworks for robot learning. She has led human-robot interactive demonstration at AAAS in 2020
and has been selected as one of the Rising Stars in EECS 2023.</p>
</div>
<br>
<a href="javascript:toggle_bio()">Formal Bio</a>
<a href="https://scholar.google.com/citations?user=EhOtO3cAAAAJ">G. Scholar</a>
<a href="https://github.com/kelym">Github</a>
<a href="https://www.linkedin.com/in/kelym/">LinkedIn</a>
<a href="https://x.com/xkelym">Twitter</a>
<br><br>
kay at physical intelligence dot company
<br><br>
</p>
</div>
<div id="intro-image">
<img src="images/profile.jpg" alt="Portrait of Liyiming Ke">
</div>
</div>
<div id="filters" class="button-group">
<button class="button is-checked" data-filter=".highlight">Highlights</button>
<button class="button" data-filter=".publication">Research</button>
<button class="button" data-filter=".talk">Talks</button>
<button class="button" data-filter=".misc">Misc</button>
</div>
<div class="grid">
<div class="list-item highlight description" data-category="highlight">
</div>
<!-- Highlights row 1 -->
<div class="list-item highlight previews" data-category="highlight">
<div class="hover-video preview1">
<a href="https://www.pi.website/research/rlt">
<video playsinline muted autoplay loop>
<source src="images/20260317-screw.mp4" type="video/mp4">
</video>
<div class="hover-caption">
RL Tokens connect a VLA to online RL for fast adaptation on fine manipulation tasks.
</div>
</a>
</div>
<div class="hover-video preview2">
<a href="https://www.pi.website/research/rlt">
<video playsinline muted autoplay loop>
<source src="images/20260317-ziptie.mp4" type="video/mp4">
</video>
<div class="hover-caption">
RL Tokens let robots improve fine motor skills from minutes to hours of real-world experience.
</div>
</a>
</div>
<div class="hover-video preview3">
<a href="https://www.pi.website/blog/pistar06">
<video playsinline muted autoplay loop>
<source src="images/20251118-pi06.mp4" type="video/mp4">
</video>
<div class="hover-caption">
π*0.6 improves VLA model during deployment using autonomous rollouts and human coaching.
</div>
</a>
</div>
</div>
<!-- Highlights row 2 -->
<div class="list-item highlight previews" data-category="highlight">
<div class="hover-video preview1">
<a href="https://www.pi.website/blog/pi05">
<video playsinline muted autoplay loop>
<source src="images/20250330-pi05-makebed.mp4" type="video/mp4">
</video>
<div class="hover-caption">
π0.5 generalizes household chores across different real homes and layouts.
</div>
</a>
</div>
<div class="hover-video preview2">
<a href="https://www.pi.website/blog/pi05">
<video playsinline muted autoplay loop>
<source src="images/20250330-pi05-putsink.mp4" type="video/mp4">
</video>
<div class="hover-caption">
π0.5 performs open-world household tasks in unfamiliar environments.
</div>
</a>
</div>
<div class="hover-video preview3">
<a href="https://www.physicalintelligence.company/research/hirobot">
<video playsinline muted autoplay loop>
<source src="images/20250110-hirobot.mp4" type="video/mp4">
</video>
<div class="hover-caption">
A hierarchical VLA reasons through long-horizon instructions and executes them step by step.
</div>
</a>
</div>
</div>
<!-- Highlights row 3 -->
<div class="list-item highlight previews" data-category="highlight">
<div class="hover-video preview1">
<a href="https://personalrobotics.github.io/CCIL/">
<video playsinline muted autoplay loop>
<source src="images/20241101-ccil-applied.mp4" type="video/mp4">
</video>
<div class="hover-caption">
CCIL boosts imitation learning robustness with synthetic corrective labels derived from continuity.
</div>
</a>
</div>
<div class="hover-video preview2">
<a href="https://www.physicalintelligence.company/blog/pi0">
<video playsinline muted autoplay loop>
<source src="images/20241101-pizero_bussing_trashpile.mp4" type="video/mp4">
</video>
<div class="hover-caption">
π0 shows that large-scale cross-embodiment pretraining can transfer across robots and tasks.
</div>
</a>
</div>
<div class="hover-video preview3">
<a href="https://goodcherrybot.github.io/">
<video playsinline muted autoplay loop>
<source src="images/20230315-cherrybot-granola_retry.mp4" type="video/mp4">
</video>
<div class="hover-caption">
Reinforcement learning enables chopsticks robot to do precise manipulation under disturbances.
</div>
</a>
</div>
</div>
<div id="main-highlights">
</div>
<!--
<div id="more-highlights" style="display: none;">
<div class="list-item highlight" data-category="highlight">
<p class="date">2024</p> <a href="https://personalrobotics.github.io/CCIL/">Can we improve robustness of
Imitation Learning by generating synthetic corrective labels?</a>
</div>
<div class="list-item highlight" data-category="highlight">
<p class="date">2024</p> <a href="https://www.physicalintelligence.company/blog/pi0">Can imitation learning
benefits from large-scale pre-training on diverse task and robot embodiment?</a>
</div>
<div class="list-item highlight" data-category="highlight">
<p class="date">2023</p> <a href="https://goodcherrybot.github.io/">Can we learn fine motor
skills like picking up cherries with chopsticks using reinforcement learning?</a>
</div>
<div class="list-item highlight" data-category="highlight">
<p class="date">2020</p> <a href="https://arxiv.org/abs/1905.12888">Viewing imitation learning from the frame
of
divergence minimization</a>
</div>
</div>
<div class="list-item highlight toggle-button" data-category="highlight">
<a id="toggle_highlights_button" href="javascript:toggle_highlights()">Show more</a>
</div>
-->
<!-- Publications -->
<div class="list-item publication" data-category="publication">
<a href="https://www.pi.website/research/rlt" class="thumbnail">
<video playsinline muted autoplay loop width="180">
<source src="images/20260317-screw.mp4" type="video/mp4">
</video>
</a>
<div class="project-description">
<h3><a href="https://www.pi.website/research/rlt">RL Token: Bootstrapping Online RL with Vision-Language-Action Models</a></h3>
<p>
Charles Xu, Jost Tobias Springenberg, Michael Equi, Ali Amin, Adnan Esmail, Sergey Levine, <b>Liyiming Ke</b><br>
<a href="https://www.pi.website/research/rlt">Webpage</a> •
<a href="https://www.pi.website/download/rlt.pdf">PDF</a> •
<span class="excerpt-container">
<a href="#" onclick="event.preventDefault();">Summary</a>
<span class="excerpt-content">
Introduces RL tokens (RLT), a lightweight interface between a VLA and an online RL policy,
enabling fast adaptation for precise, delicate manipulation tasks from minutes or hours of
real-world experience.
</span>
</span>
</p>
</div>
</div>
<div class="list-item publication" data-category="publication">
<a href="https://arxiv.org/abs/2511.14759" class="thumbnail">
<img src="images/20251118-pi06.jpg" alt="π*0.6 teaser image" />
</a>
<div class="project-description">
<h3><a href="https://www.pi.website/blog/pistar06">π<sup>*</sup><sub>0.6</sub>: a VLA That Learns From Experience</a></h3>
<p>
Ali Amin, Raichelle Aniceto, Ashwin Balakrishna, Kevin Black, Ken Conley,
Grace Connors, James Darpinian, Karan Dhabalia, Jared DiCarlo, Danny Driess, Michael Equi,
Adnan Esmail, Yunhao Fang, Chelsea Finn, Catherine Glossop, Thomas Godden, Ivan Goryachev,
Lachy Groom, Hunter Hancock, Karol Hausman, Gashon Hussein, Brian Ichter, Szymon Jakubczak,
Rowan Jen, Tim Jones, Ben Katz, <b>Liyiming Ke</b>, Chandra Kuchi, Marinda Lamb, Devin LeBlanc,
Sergey Levine, Adrian Li-Bell, Yao Lu, Vishnu Mano, Mohith Mothukuri, Suraj Nair, Karl Pertsch,
Allen Z. Ren, Charvi Sharma, Lucy Xiaoyang Shi, Laura Smith, Jost Tobias Springenberg,
Kyle Stachowicz, Will Stoeckle, Alex Swerdlow, James Tanner, Marcel Torne, Quan Vuong,
Anna Walling, Haohuan Wang, Blake Williams, Sukwon Yoo, Lili Yu, Ury Zhilinsky, Zhiyuan Zhou<br>
<a href="https://www.pi.website/blog/pistar06">Webpage</a> •
<a href="https://arxiv.org/abs/2511.14759">PDF</a> •
<span class="excerpt-container">
<a href="#" onclick="event.preventDefault();">Summary</a>
<span class="excerpt-content">
Enable VLA to improve from real-world autonomous rollout and human coaching during deployment time, via reinforcement learning.
</span>
</span>
</p>
</div>
</div>
<div class="list-item publication" data-category="publication">
<a href="https://www.pi.website/blog/pi05" class="thumbnail">
<video playsinline muted autoplay loop width="180">
<source src="images/20250330-pi05-makebed.mp4" type="video/mp4">
</video>
</a>
<div class="project-description">
<h3><a href="https://www.pi.website/blog/pi05">π0.5: A Vision-Language-Action
Model with Open World Generalization</a></h3>
<p>
Kevin Black, Noah Brown, James Darpinian, Karan Dhabalia, Danny Driess, Adnan Esmail, Michael Equi, Chelsea
Finn, Niccolo Fusai, Manuel Y Galliker, Dibya Ghosh, Lachy Groom, Karol Hausman, Brian Ichter, Szymon
Jakubczak, Tim Jones, <b>Liyiming Ke</b>, Devin LeBlanc, Sergey Levine, Adrian Li-Bell, Mohith Mothukuri,
Suraj Nair, Karl Pertsch, Allen Z Ren, Lucy Xiaoyang Shi, Laura Smith, Jost Tobias Springenberg, Kyle
Stachowicz, James Tanner, Quan Vuong, Homer Walke, Anna Walling, Haohuan Wang, Lili Yu, Ury Zhilinsky<br>
<a href="https://www.pi.website/blog/pi05">Webpage</a> •
<a href="https://www.physicalintelligence.company/download/pi05.pdf">PDF</a> •
<span class="excerpt-container">
<a href="#" onclick="event.preventDefault();">Summary</a>
<span class="excerpt-content">
We send mobile robots to many AirBnB houses to generalize tasks across diverse, real-world environments.
Our robots can perform some household chores like cleaning kitchens in unseen houses.
</span>
</span>
</p>
</div>
</div>
<div class="list-item publication" data-category="publication">
<a href="https://www.physicalintelligence.company/research/hirobot" class="thumbnail">
<video playsinline muted autoplay loop width="180">
<source src="images/20250110-hirobot.mp4" type="video/mp4">
</video>
</a>
<div class="project-description">
<h3><a href="https://www.physicalintelligence.company/research/hirobot">Hi Robot: Open-Ended Instruction
Following
with Hierarchical
Vision-Language-Action Models</a></h3>
<p>
Lucy Xiaoyang Shi, Brian Ichter, Michael Equi, <b>Liyiming Ke</b>, Karl Pertsch, Quan Vuong, James Tanner,
Anna Walling, Haohuan Wang, Niccolo Fusai, Adrian Li-Bell, Danny Driess, Lachy Groom, Sergey Levine, Chelsea
Finn<br>
<i>ICML 2025</i><br>
<a href="https://www.physicalintelligence.company/research/hirobot">Webpage</a> •
<a href="https://arxiv.org/abs/2502.19417">PDF</a> •
<span class="excerpt-container">
<a href="#" onclick="event.preventDefault();">Summary</a>
<span class="excerpt-content">
We introduce a hierarchical system enabling robots to “think aloud” and deconstruct complex tasks ("make
me a sandwich") into
manageable steps ("pick up bread, pick up tomato, put tomato on the bread ..."). By combining a
low-level action
model for execution and a
high-level
vision-language model for reasoning and interaction with human inputs, we allow robots to follow complex
instructions and perform tasks with high
precision and adaptability.
</span>
</span>
</p>
</div>
</div>
<div class="list-item publication" data-category="publication">
<a href="https://www.physicalintelligence.company/blog/pi0" class="thumbnail">
<video playsinline muted autoplay loop width="180">
<source src="images/20241101-pizero.mp4" type="video/mp4">
</video>
</a>
<div class="project-description">
<h3><a href="https://www.physicalintelligence.company/blog/pi0">π0: A Vision-Language-Action Flow
Model for
General Robot Control</a></h3>
<p>
Kevin Black, Noah Brown, Danny Driess, Adnan Esmail, Michael Equi, Chelsea Finn, Niccolo Fusai,
Lachy Groom, Karol Hausman, Brian Ichter, Szymon Jakubczak, Tim Jones, <b>Liyiming Ke</b>, Sergey Levine,
Adrian Li-Bell, Mohith Mothukuri, Suraj Nair, Karl Pertsch, Lucy Xiaoyang Shi, James Tanner, Quan Vuong,
Anna Walling, Haohuan Wang, Ury Zhilinsky<br>
<i>RSS 2025</i><br>
<a href="https://www.physicalintelligence.company/blog/pi0">Webpage</a> •
<a href="https://www.physicalintelligence.company/download/pi0.pdf">PDF</a> •
<span class="excerpt-container">
<a href="#" onclick="event.preventDefault();">Summary</a>
<span class="excerpt-content">
Can you train cross-embodiment robotic policies over many many tasks and expect it to work? We show that
it is promising: a big pre-training model can be finetuned on a single task and outperform
dedicated policy that has only seen task-specific data.
</span>
</span>
</p>
</div>
</div>
<div class="list-item publication" data-category="publication">
<a href="https://arxiv.org/abs/2410.20254" class="thumbnail">
<img src="images/20241101-sim2real.jpg" alt="Sim-to-real exploration teaser image" />
</a>
<div class="project-description">
<h3><a href="https://arxiv.org/abs/2410.20254">Overcoming the Sim-to-Real Gap: Leveraging Simulation to Learn
to Explore for Real-World RL</a></h3>
<p>
Andrew Wagenmaker, Kevin Huang, <b>Liyiming Ke</b>, Byron Boots, Kevin Jamieson, Abhishek Gupta<br>
<i>NeurIPS 2024</i><br>
<a href="https://arxiv.org/abs/2410.20254">PDF</a> •
<span class="excerpt-container">
<a href="#" onclick="event.preventDefault();">Summary</a>
<span class="excerpt-content">
We show that, learning an exploration policy in simulation can boost the real-world reinforcement
learning
finetuning efficiency (versus learning an optimal policy in the sim and transfer the policy).
</span>
</span>
</p>
</div>
</div>
<div class="list-item publication" data-category="publication">
<a href="https://arxiv.org/abs/2405.19307" class="thumbnail">
<video playsinline muted autoplay loop width="180">
<source src="images/20241101-ccil-applied.mp4" type="video/mp4">
</video>
</a>
<div class="project-description">
<h3><a href="https://arxiv.org/abs/2405.19307">Data Efficient Behavior Cloning for Fine Manipulation via
Continuity-based Corrective Labels</a></h3>
<p>
Abhay Deshpande, <b>Liyiming Ke</b>, Quinn Pfeifer, Abhishek Gupta, Siddhartha S. Srinivasa<br>
<i>IROS 2024</i><br>
<a href="https://personalrobotics.github.io/CCIL/">Webpage</a> •
<a href="https://arxiv.org/abs/2405.19307">PDF</a> •
<span class="excerpt-container">
<a href="#" onclick="event.preventDefault();">Summary</a>
<span class="excerpt-content">
We apply CCIL to real world robotic manipulation tasks and it kinda worked after some design tweak. The
most juice comes from setting up trust threshold for the generated labels in a task-agnostic way.
</span>
</span>
</p>
</div>
</div>
<div class="list-item publication" data-category="publication">
<a href="https://personalrobotics.github.io/CCIL/" class="thumbnail">
<img src="images/20231019-ccil.png" alt="CCIL teaser image" />
</a>
<div class="project-description">
<h3><a href="https://personalrobotics.github.io/CCIL/">CCIL: Continuity-based Data Augmentation for Corrective
Imitation Learning</a></h3>
<p>
<b>Liyiming Ke*</b>, Yunchu Zhang*, Abhay Deshpande, Siddhartha Srinivasa, Abhishek Gupta<br>
<i>ICLR 2024</i><br>
<a href="https://personalrobotics.github.io/CCIL/">Webpage</a> •
<a href="https://github.com/personalrobotics/CCIL">Code</a> •
<a href="https://arxiv.org/abs/2310.12972v1">PDF</a> •
<span class="excerpt-container">
<a href="#" onclick="event.preventDefault();">Summary</a>
<span class="excerpt-content">
Enhances robustness of imitation learning by generating synthetic corrective labels:
The trick is to leverage local continuity in the environment dynamics - and for regions that are
discontinuous, quantify the confidence and skip them.
</span>
</span>
</p>
</div>
</div>
<div class="list-item publication" data-category="publication">
<a href="https://goodcherrybot.github.io/" class="thumbnail">
<video playsinline muted autoplay loop width="180">
<source src="images/20230315-cherry-picking.mp4" type="video/mp4">
</video>
</a>
<div class="project-description">
<h3><a href="https://goodcherrybot.github.io/">Cherry Picking with Reinforcement Learning</a></h3>
<p>
Yunchu Zhang*, <b>Liyiming Ke*</b>, Abhay Deshpande, Abhishek Gupta, Siddhartha Srinivasa<br>
<i>RSS 2023</i><br>
<a href="https://goodcherrybot.github.io/">Webpage</a> •
<a href="https://arxiv.org/abs/2303.05508">PDF</a> •
<span class="excerpt-container">
<a href="#" onclick="event.preventDefault();">Summary</a>
<span class="excerpt-content">
Use reinforcement learning to learn fine motor skills: pick up slippery cherries with chopsticks under
wind or human disturbances. And I refuse to do parameter sweeping or random seed cherry picking.
</span>
</span>
</p>
</div>
</div>
<div class="list-item publication" data-category="publication">
<a href="https://sites.google.com/view/real-orl" class="thumbnail">
<img src="images/20220930-real-offlinerl.png" alt="Real-world offline RL teaser image" />
</a>
<div class="project-description">
<h3><a href="https://sites.google.com/view/real-orl">Real World Offline Reinforcement Learning with Realistic
Data Sources</a></h3>
<p>
Gaoyue Zhou*, <b>Liyiming Ke*</b>, Siddhartha Srinivasa, Abhinav Gupta, Aravind Rajeswaran, Vikash Kumar<br>
<i>ICRA 2023</i><br>
<a href="https://sites.google.com/view/real-orl">Webpage</a> •
<a href="https://arxiv.org/abs/2210.06479">PDF</a> •
<span class="excerpt-container">
<a href="#" onclick="event.preventDefault();">Summary</a>
<span class="excerpt-content">
Eval offline RL in real-world: emphasize on data being "kinda good" but not perfect.
</span>
</span>
</p>
</div>
</div>
<div class="list-item publication" data-category="publication">
<a href="https://personalrobotics.cs.washington.edu/publications/ke2021grasping.pdf" class="thumbnail">
<img src="images/20201101-chopsticks-grasping.jpg" alt="Grasping with chopsticks teaser image" />
</a>
<div class="project-description">
<h3><a href="https://personalrobotics.cs.washington.edu/publications/ke2021grasping.pdf">Grasping with
Chopsticks: Combating Covariate Shift in Model-free Imitation Learning for Fine Manipulation</a></h3>
<p>
<b>Liyiming Ke</b>, Jingqiang Wang, Tapomayukh Bhattacharjee, Byron Boots, Siddhartha S. Srinivasa<br>
<i>ICRA 2021</i><br>
<a href="https://arxiv.org/abs/2011.06719">PDF</a> •
<span class="excerpt-container">
<a href="#" onclick="event.preventDefault();">Summary</a>
<span class="excerpt-content">
Teach a robot to use chopsticks for precise manipulation tasks through human demonstrations: Addresses
covariate shift in imitation learning by noise-injection, object-centric transformation and
bunch of hacks.
</span>
</span>
</p>
</div>
</div>
<div class="list-item publication" data-category="publication">
<a href="https://personalrobotics.cs.washington.edu/publications/ke2020teleop.pdf" class="thumbnail">
<img src="images/20200630-chopsticks-teleop.gif" alt="Telemanipulation with chopsticks teaser image" />
</a>
<div class="project-description">
<h3><a href="https://personalrobotics.cs.washington.edu/publications/ke2020teleop.pdf">Telemanipulation with
Chopsticks: Analyzing Human Factors in User Demonstrations</a></h3>
<p>
<b>Liyiming Ke</b>, Ajinkya Kamat, Jingqiang Wang, Tapomayukh Bhattacharjee, Christoforos Mavrogiannis,
Siddhartha S. Srinivasa<br>
<i>IROS 2020</i><br>
<a href="https://arxiv.org/abs/2008.00101">PDF</a> •
<span class="excerpt-container">
<a href="#" onclick="event.preventDefault();">Summary</a>
<span class="excerpt-content">
Built a chopsticks robot and a fun human-interactive demo collection interface: turns out that tracking
a
wand and commmand the robot can be really easy.
</span>
</span>
</p>
</div>
</div>
<div class="list-item publication" data-category="publication">
<a href="https://arxiv.org/abs/1905.12888" class="thumbnail">
<img src="images/20190608-fimitation-teaser.png" alt="f-divergence imitation learning teaser image" />
</a>
<div class="project-description">
<h3><a href="https://arxiv.org/abs/1905.12888">Imitation Learning as f-Divergence Minimization</a></h3>
<p>
<b>Liyiming Ke</b>, Sanjiban Choudhury, Matt Barnes, Wen Sun, Gilwoo Lee, Siddhartha Srinivasa<br>
<i>WAFR 2020</i><br>
<a href="https://arxiv.org/abs/1905.12888">PDF</a> •
<span class="excerpt-container">
<a href="#" onclick="event.preventDefault();">Summary</a>
<span class="excerpt-content">
A unified theoretical framework for imitation learning! Turns out some SOTA algorithms are using
f-divergence. We show how different divergence measures lead to different imitation learning approaches.
</span>
</span>
</p>
</div>
</div>
<div class="list-item publication" data-category="publication">
<a href="http://openaccess.thecvf.com/content_CVPR_2019/html/Ke_Tactical_Rewind_Self-Correction_via_Backtracking_in_Vision-And-Language_Navigation_CVPR_2019_paper.html"
class="thumbnail">
<img src="images/20190607-vln-teaser.png" alt="Tactical Rewind teaser image" />
</a>
<div class="project-description">
<h3><a
href="http://openaccess.thecvf.com/content_CVPR_2019/html/Ke_Tactical_Rewind_Self-Correction_via_Backtracking_in_Vision-And-Language_Navigation_CVPR_2019_paper.html">Tactical
Rewind: Self-Correction via Backtracking in Vision-and-Language Navigation</a></h3>
<p>
<b>Liyiming Ke</b>, Xiujun Li, Yonatan Bisk, Ari Holtzman, Zhe Gan, Jingjing Liu, Jianfeng Gao, Yejin Choi,
Siddhartha Srinivasa<br>
<i>CVPR 2019</i>
<font color="49bf9"><i>★ Oral Presentation, CVPR (5.6%) ★</i></font><br>
<a href="https://arxiv.org/abs/1903.02547">PDF</a> •
<span class="excerpt-container">
<a href="#" onclick="event.preventDefault();">Summary</a>
<span class="excerpt-content">
Baking Search and Planning into ML-based navigation: We propose a new framework for VL navigation,
enabling agents to recover from mistakes by maintaining internal search tree and returning to previous
positions and trying alternative
paths.
</span>
</span>
</p>
</div>
</div>
<div class="list-item publication" data-category="publication">
<a href="https://ojs.aaai.org/index.php/AAAI/article/view/10061" class="thumbnail">
<img src="images/20160101-email-teaser.png" alt="Email filter evasion teaser image" />
</a>
<div class="project-description">
<h3><a href="https://ojs.aaai.org/index.php/AAAI/article/view/10061">Behavioral Experiments in Email Filter
Evasion</a></h3>
<p>
<b>Liyiming Ke</b>, Bo Li, Yevgeniy Vorobeychik<br>
<i>AAAI 2016</i><br>
<a href="https://ojs.aaai.org/index.php/AAAI/article/view/10061">PDF</a> •
<span class="excerpt-container">
<a href="#" onclick="event.preventDefault();">Summary</a>
<span class="excerpt-content">
Studies how humans attempt to evade email spam filters.
Provides insights into adversarial behavior and implications for security system design.
</span>
</span>
</p>
</div>
</div>
<!-- Talks -->
<div class="list-item talk" data-category="talk">
<p class="date">2026</p>Upcoming<a href=""></a>
</div>
<div class="list-item talk" data-category="talk">
<p class="date">2025</p>RSS Workshop <a href="https://wcbm-workshop.github.io/">WCBM</a> and <a href="https://hitl-robot-learning.github.io/">HILT</a>
</div>
<div class="list-item talk" data-category="talk">
<p class="date"></p>Actuate SF (<a href="https://www.youtube.com/watch?v=s4go3bAB1q8">recording</a>)
</div>
<div class="list-item talk" data-category="talk">
<p class="date">2024</p>OpenAI Reading Group
</div>
<div class="list-item talk" data-category="talk">
<p class="date"></p>University of Washington, Robotics Seminar (<a
href="https://www.youtube.com/watch?v=LeHYQVR8a8k">video</a>)
</div>
<div class="list-item talk" data-category="talk">
<p class="date">2023</p>Stanford University, <a href="https://iliad.stanford.edu/">ILIAD Lab</a>
</div>
<div class="list-item talk" data-category="talk">
<p class="date"></p>University of California Berkeley, <a href="https://bair.berkeley.edu/">BAIR Lab</a>
</div>
<div class="list-item talk" data-category="talk">
<p class="date"></p>Carnegie Mellon University, <a href="https://www.cs.cmu.edu/~cga/">Atkeson Lab</a>
</div>
<div class="list-item talk" data-category="talk">
<p class="date"></p>Shanghai Jiaotong University, <a href="https://automation.sjtu.edu.cn/">Department of
Automation</a>
</div>
<div class="list-item talk" data-category="talk">
<p class="date"></p>Stanford University, <a href="https://iprl.stanford.edu/">Interactive Perception and
Robot
Learning Lab</a>
</div>
<div class="list-item talk" data-category="talk">
<p class="date">2022</p>Cornell University, <a href="https://emprise.cs.cornell.edu/">EmPRISE
Lab</a>
</div>
<div class="list-item talk" data-category="talk">
<p class="date"></p> <a href="https://mila.quebec/en/">Mila - Quebec AI Institute</a> (<a
href="https://www.youtube.com/watch?v=LeHYQVR8a8k">video</a>)
</div>
<div class="list-item talk" data-category="talk">
<p class="date">2021</p>MetaAI Reading Group
</div>
<div class="list-item talk" data-category="talk">
<p class="date">2018</p>Microsoft Research Dialogue Group Reading Group
</div>
<!-- Misc -->
<div class="list-item misc" data-category="misc">
<p class="date"> </p>Reviewer of RSS, CoRL, ICLR, NeurIPS, ICRA, IJRR, IROS, RA-L, HRI, AAMAS, IJCAI
</div>
<div class="list-item misc" data-category="misc">
<p class="date">2026</p>I had <a href="https://chinaresearchcollective.substack.com/p/exclusive-interview-with-physical">a casual interview with Harry Xu</a> to chat about robot reinforcement learning.
</div>
<div class="list-item misc" data-category="misc">
<p class="date">2025</p>We <a href="https://github.com/Physical-Intelligence/openpi">open source
π 0 on Github</a> to empower the community by sharing our foundation models.
</div>
<div class="list-item misc" data-category="misc">
<p class="date">2024</p>Our first generalist robotic model, π 0,
is featured on <a href="https://www.nytimes.com/2024/11/04/business/dealbook/selzer-poll-trump-trade.html">
New York Times</a> and <a href="https://www.wired.com/story/physical-intelligence-home-robot/"> Wired</a>.
</div>
<div class="list-item misc" data-category="misc">
<p class="date">2023</p>Honored to be selected as one of the <a
href="https://www.eecs.mit.edu/community-equity/rising-stars-in-eecs/">Rising Stars in EECS</a>
</div>
<div class="list-item misc" data-category="misc">
<p class="date">2020</p>Chopsticks Robot featured on <a
href="https://spectrum.ieee.org/video-friday-agility-robotics-robot-production">IEEE Spectrum Video
Friday</a>
</div>
<div class="list-item misc" data-category="misc">
<p class="date">2020</p>Led a human-robot interactive demo at the <a
href="https://www.aaas.org/events/2020-aaas-annual-meeting">AAAS
gathering</a>
</div>
<div class="list-item misc" data-category="misc">
<p class="date">2017</p>Graduated as one of the <a href="https://my.vanderbilt.edu/collegescholars/">Honor
Scholars</a> from Vanderbilt University
</div>
<div class="list-item misc" data-category="misc">
<p class="date">2015</p>First prize in the Vanderbilt Student Consulting for Non-profit Organization
</div>
<div class="list-item misc" data-category="misc">
<p class="date"> - </p>Inspired by
<ul
style="margin-top: 5px; margin-bottom: 0px; margin-left: 0px; list-style-type: none; padding-left: 0; display: inline;">
<li style="display: inline;"> • </li>
<li style="display: inline;"><a href="https://distill.pub/">Distill</a></li>
<li style="display: inline;"> • </li>
<li style="display: inline;"><a href="https://lilianweng.github.io/">Lil' Log</a></li>
<li style="display: inline;"> • </li>
<li style="display: inline;"><a href="https://colah.github.io/">Colah's Blog</a></li>
<li style="display: inline;"> • </li>
</ul>
</div>
</div>
<div id="footer">
Website template by <a href="https://andyzeng.github.io/">Andy Zeng</a> and <a
href="https://jonbarron.info/">Jon's
website</a>.
</div>
</div>
<script>
var $grid = $('.grid').isotope({
itemSelector: '.list-item',
layoutMode: 'fitRows',
transitionDuration: 0,
stagger: 10,
initLayout: false,
getSortData: {
name: '.name',
symbol: '.symbol',
number: '.number parseInt',
category: '[data-category]',
weight: function (itemElem) {
var weight = $(itemElem).find('.weight').text();
return parseFloat(weight.replace(/[\(\)]/g, ''));
}
}
});
$('#filters').on('click', 'button', function () {
var filterValue = $(this).attr('data-filter');
localStorage.setItem('filterValue', filterValue);
$grid.isotope({ filter: filterValue });
});
$('.button-group').each(function (i, buttonGroup) {
var $buttonGroup = $(buttonGroup);
$buttonGroup.on('click', 'button', function () {
$buttonGroup.find('.is-checked').removeClass('is-checked');
$(this).addClass('is-checked');
});
});
function update_isotope() {
var defaultFilterValue = localStorage.getItem('filterValue');
if (defaultFilterValue == null) {
defaultFilterValue = ".highlight";
}
$grid.isotope({ filter: defaultFilterValue });
var buttons = document.getElementsByClassName("button");
for (var currButton of buttons) {
if (currButton.getAttribute('data-filter') == defaultFilterValue) {
currButton.classList.add('is-checked');
} else {
currButton.classList.remove('is-checked');
}
}
}
function toggle_bio() {
var x = document.getElementById("more-bio");
if (x.style.display === "none") {
x.style.display = "block";
} else {
x.style.display = "none";
}
}
function toggle_highlights() {
var x = document.getElementById("main-highlights");
var y = document.getElementById("more-highlights");
var b = document.getElementById("toggle_highlights_button");
if (y.style.display === "none") {
x.style.display = "none";
y.style.display = "block";
b.innerHTML = "Show less";
update_isotope();
} else {
x.style.display = "block";
y.style.display = "none";
b.innerHTML = "Show more";
update_isotope();
}
}
update_isotope();
</script>
</body>
</html>