-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathindex.html
More file actions
1166 lines (1066 loc) · 65.1 KB
/
index.html
File metadata and controls
1166 lines (1066 loc) · 65.1 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
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>
<!--
Spatial by TEMPLATED
templated.co @templatedco
Released for free under the Creative Commons Attribution 3.0 license (templated.co/license)
-->
<html>
<head>
<title> Mikhail Yurochkin | PHD, Statistics </title>
<meta http-equiv="content-type" content="text/html; charset=utf-8" />
<meta name="description" content="Personal webpage of Mikhail Yurochkin" />
<meta name="keywords" content="Mikhail Yurochkin" />
<noscript>
<link rel="stylesheet" href="css/style.css" />
<link rel="stylesheet" href="css/skel.css" />
<link rel="stylesheet" href="css/style-xlarge.css" />
</noscript>
<link rel="stylesheet" type="text/css" href="https://fonts.googleapis.com/css?family=Raleway:400,700">
<link rel="stylesheet" type="text/css" href="https://fonts.googleapis.com/css?family=Open+Sans">
<link rel="stylesheet" href="css/font-awesome.min.css">
<link rel="icon" type="image/png" sizes="32x32" href="favicon-32x32.png">
<link rel="icon" type="image/png" sizes="16x16" href="favicon-16x16.png">
<!--[if lte IE 8]><script src="js/html5shiv.js"></script><![endif]-->
<script src="js/jquery.min.js"></script>
<script src="js/skel.min.js"></script>
<script src="js/skel-layers.min.js"></script>
<script src="js/init.js"></script>
<script src='https://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML'></script>
<script>
(function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){
(i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o),
m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m)
})(window,document,'script','https://www.google-analytics.com/analytics.js','ga');
ga('create', 'UA-105100164-1', 'auto');
ga('send', 'pageview');
</script>
</head>
<body class="landing">
<!-- Header -->
<header id="header">
<ul class="icons">
<li>
<a href="https://scholar.google.com/citations?user=QjBF9sUAAAAJ&hl=en&authuser=1" class="icon fa-graduation-cap"></a>
</li>
<li>
<a href="https://x.com/Yurochkin_M" class="icon fa-twitter"></a>
</li>
<li>
<a href="https://github.com/moonfolk" class="icon fa-github"></a>
</li>
<li>
<a href="https://www.linkedin.com/in/mikhail-yurochkin-a45659114/" class="icon fa-linkedin"></a>
</li>
</ul>
<nav id="nav">
<ul>
<li><a href="#one">Home</a></li>
<li><a href="#two">Publications</a></li>
<li><a href="files/CV_MY.pdf">CV</a></li>
<li><a href="#footer">Contact</a></li>
</ul>
</nav>
</header>
<!-- One -->
<section id="one" class="wrapper style1">
<div class="container 60%">
<div class="row 200%">
<div class="5u 5u$$(medium)">
<header class="major">
<h2 style="font-size:2.5em; text-center"> Mikhail Yurochkin</h2>
<p>
<div class="image rounded" style="margin-bottom: 1.5em;"><img src="images/photo.jpg" width="420" alt="" /></div>
</p>
<!--<p>PHD, Statistics</p>-->
</header>
</div>
<!-- <div class="7u$ 5u$(medium)"> -->
<div class="7u$ 12u$(medium)">
<font size="5"><p <br><br><br><br>
<!-- I am a Research Staff Member at IBM Research and MIT-IBM Watson AI Lab in Cambridge, Massachusetts. My research interests are
</p>
<ul>
<li>Foundation Models</li>
<li>Out-of-Distribution Generalization</li>
<li>Algorithmic Fairness</li>
<!-- <li>Applications of Optimal Transport in Machine Learning</li>
<li>Model Fusion and Federated Learning</li>
<!-- <li>Bayesian (nonparametric) modeling and inference</li>
</ul>-->
I am a Staff AI Scientist and LLM Data Team lead at the <a href="https://ifm.mbzuai.ac.ae/" target ="_blank">IFM MBZUAI Silicon Valley Lab</a>.
Before that, I was a research manager at the <a href="https://mitibmwatsonailab.mit.edu/" target ="_blank">MIT‑IBM Watson AI Lab</a> leading the Statistical Large Language Modeling group.
I completed my PhD in Statistics at the University of Michigan, where I worked with Prof. <a href="http://dept.stat.lsa.umich.edu/~xuanlong/" target ="_blank">Long Nguyen</a>.
<p>I am interested in a variety of LLM‑related problems—pre‑ and post‑training, data quality, reasoning, evaluation, routing, and efficient inference—and enjoy exploring statistical modeling approaches to solve them.
I have also worked on OOD generalization, algorithmic fairness, optimal transport, federated learning, and Bayesian nonparametrics.</p></font>
<!-- <p>Before joining IBM, I completed my PhD in Statistics at the University of Michigan, where I worked with <a href="http://dept.stat.lsa.umich.edu/~xuanlong/" target ="_blank">Long Nguyen</a>. I received my Bachelor's degree in applied mathematics and physics from Moscow Institute of Physics and Technology.</p></font> -->
</div>
</div>
<!--<p style="text-align: center; font-weight: bold; font-size:120%"> text </p>-->
</div>
</section>
<section id="news" class="wrapper style2">
<div class="container 75%">
<h3 style="text-align: center">News</h3>
<ul>
<li>09/2025: Three papers accepted at NeurIPS 2025.</li>
<li>07/2025: Two papers accepted at COLM 2025.</li>
<li>05/2025: Three papers accepted at ICML 2025.</li>
<li>04/2025: Presented our work on serving 1000s of Customized LLMs at <a href="https://orfe.princeton.edu/events/2025/mikhail-yurochkin-university-michigan">Wilks Seminar at Princeton University</a> (<a href="files/LoRA-compression-April7-full-talk.pdf">slides</a>).</li>
<li>04/2025: I joined MBZUAI Institute of Foundation Models as a Staff AI Scientist.</li>
<li>02/2025: Thanks IBM Research for sharing <a href="https://research.ibm.com/blog/misha-yurochkin-ai">my journey as an AI Researcher</a>.</li>
<li>12/2024: Another great year serving as Project Advisor for <a href="https://www.breakthroughtech.org/programs/the-ai-program">Break Through Tech AI</a>!</li>
<li>10/2024: IBM Research blogs featuring our work on <a href="https://research.ibm.com/blog/LLM-routers">LLM Routing</a> and <a href="https://research.ibm.com/blog/LoRAs-explained">LoRA compression</a>.</li>
<li>09/2024: Three papers accepted at NeurIPS 2024 and one at Findings of EMNLP 2024.</li>
<li>07/2024: Two papers accepted at COLM 2024.</li>
<!-- <li>05/2024: Three papers accepted at ICML 2024.</li> -->
<li>03/2024: <a href="https://arxiv.org/abs/2402.14992">tinyBenchmarks</a> are now on <a href="https://huggingface.co/tinyBenchmarks">HuggingFace</a>! Evaluate on MMLU (and others) with just 100 examples.</li>
<!-- <li>02/2024: Three papers and one tiny paper accepted at ICLR 2024.</li> -->
<!-- <li>11/2023: I am promoted to lead the new Statistical Large Language Modeling group at the MIT-IBM AI Lab.</li> -->
<!-- <li>10/2023: Five papers accepted at NeurIPS 2023 Workshops, including one oral!</li> -->
<!-- <li>10/2023: Two new papers on solving NLP tasks by fusing/routing smaller LLMs (<a href="https://arxiv.org/abs/2309.15789">one</a>, <a href="https://arxiv.org/abs/2310.01542">two</a>).</li> -->
<!-- <li>10/2023: Operationalizing Individual Fairness at <a href="https://youngstats.github.io/post/2023/09/19/algorithmic-fairness/">One World YoungStatS Webinar series</a> (<a href="https://www.youtube.com/watch?v=faPooozTKGk">video</a>).</li> -->
<!--<li>09/2023: I am back as a Project Advisor at <a href="https://www.breakthroughtech.org/programs/the-ai-program/">Break Through Tech AI (BTAI)</a>! This time working with two teams.</li>-->
<!--<li>04/2023: One paper accepted at ICML 2023 🌴.</li>-->
<!--<li>02/2023: AI Fairness through Robustness Tutorial @ AAAI 2023, February 7th at 8:30 AM EST (slides and more on our <a href="https://sites.google.com/view/aaai-2023-fairness-robustness">website</a>).-->
<!-- <li>01/2023: Three papers accepted at ICLR 2023 and one at CHI 2023.</li> -->
<!-- <li>10/2022: Our tutorial on AI Fairness through Robustness was accepted for presentation at <a href="https://aaai.org/Conferences/AAAI-23/aaai23tutorials/">AAAI 2023</a>!</li> -->
<!--<li>10/2022: I am co-presenting a tutorial on Fairness of Machine Learning in Search Engines at <a href="https://www.cikm2022.org/tutorials">CIKM 2022</a> (<a href="files/MY_slides_cikm.pdf">slides</a>).</li>-->
<!-- <li>09/2022: Two papers accepted at NeurIPS 2022, including one oral!</li> -->
<!-- <li>09/2022: I will serve as a Project Advisor in the <a href="https://computing.mit.edu/about/diversity-equity-inclusion/break-through-tech-ai/">Break Through Tech AI (BTAI)</a> program at MIT.</li> -->
<!--<li>07/2022: Our team has released <a href="https://github.com/IBM/inFairness/">inFairness</a> - the first open source Python library for Individual Fairness.<br>-->
<!--Learn more about fairness in AI in our <a href="https://video.ibm.com/recorded/131932983">tutorial video</a> (<a href="files/MY_talk_tutorial.pdf">slides</a>).--> <!-- and tryout the <a href="https://fairbert.vizhub.ai/">Fairness Playground</a>! -->
<!-- <li>07/2022: Mentoring a <a href="https://www.cee.org/programs/research-science-institute/">Research Science Institute (RSI)</a> student for the first time! <a href="https://toddtechtalk.org/">Todd Zhou</a>, welcome to the group.</li> -->
<!-- <li>06/2022: Excited to mentor <a href="https://lilianngweta.com/">Lilian Ngweta</a>, <a href="https://yzeng58.github.io/zyc_cv/">Yuchen Zeng</a>, <a href="https://www.linkedin.com/in/hbyang/">Hao Bang (Kevin) Yang</a>, and <a href="https://scholar.google.com/citations?user=eD9vCGMAAAAJ&hl=en">Subha Maity</a> this summer.</li> -->
<!-- <li>05/2022: One paper accepted at ICML 2022.</li> -->
<!-- <li>03/2022: I presented Practical Individual Fairness at the <a href="https://trustmlresearch.github.io/">TrustML Young Scientist Seminars</a> (<a href="https://www.youtube.com/watch?v=6jI7_wlMPX4">video</a>).</li> -->
<!-- <li>02/2022: One paper accepted at Findings of ACL 2022.</li> -->
<!-- <li>01/2022: One paper accepted at AISTATS 2022.</li> -->
<!-- <li>11/2021: I gave invited talks on Individual Fairness at <a href="https://odsc.com/california/">ODSC West</a> and <a href="https://www.torontomachinelearning.com/">Toronto Machine Learning Summit</a> (<a href="files/MY_fairness_slides.pdf">slides</a>).</li> -->
<!--<li>10/2021: Three papers accepted at NeurIPS 2021.</li> -->
<!-- <li>08/2021: I gave a keynote talk on Algorithmic Fairness at <a href="https://sites.google.com/view/kdd-mlf-2021/home">KDD Workshop on Machine Learning in Finance</a>.</li> -->
<!-- <li>06/2021: I gave invited talks on Federatel Learning at <a href="https://odsc.com/europe/">ODSC Europe</a> and <a href="https://federatedlearningconference.com/">The 2nd Annual Federated and Distributed/Decentralized Machine Learning Conference</a>.</li> -->
<!-- <li>05/2021: I was an invited panelist at the <a href="https://hubsecurity.io/watch-now-federated-learning-ai-data-security-summit/">Federated Learning, AI & Data Security Summit</a> organized by <a href="https://hubsecurity.io/">HUB Security</a> (<a href="https://www.youtube.com/watch?v=PL39scRsxvE">video here</a>).</li> -->
<!-- <li>05/2021: I am honored to receive IBM's Outstanding Technical Achievement Award.</li> -->
<!-- <li>05/2021: One paper accepted at ICML 2021.</li> -->
<!-- <li>05/2021: IBM <a href="https://research.ibm.com/blog/make-ai-fairer">blog post</a> about our work on Individual Fairness. -->
<!-- <li>03/2021: I gave an invited talk on Individual Fairness at the <a href="https://sites.google.com/view/faf-workshop">Foundations of Algorithmic Fairness Workshop</a> (<a href="https://youtu.be/6v-uDp-EkXE?t=5946">video here</a>).</li> -->
<!-- <li>01/2021: Four papers accepted at ICLR 2021, including one oral and one spotlight!</li> -->
</ul>
</div>
</section>
<!-- Two -->
<section id="two" class="wrapper style1">
<div class="container 75%">
<h2 style="text-align: center">Publications</h2>
<hr />
<h2 style="text-align: right">2025</h2>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/sloth.png" width="210" height="160" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2412.06540">Sloth: scaling laws for LLM skills to predict multi-benchmark performance across families</a></h4>
</header>
<p>Felipe Maia Polo, Seamus Somerstep, Leshem Choshen, Yuekai Sun, Mikhail Yurochkin<br>
Neural Information Processing Systems (NeurIPS), 2025<br>
<a href="https://arxiv.org/pdf/2412.06540">arXiv</a> / <a href="https://github.com/felipemaiapolo/sloth">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/llm-judge.png" width="210" height="160" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2508.12792">Bridging Human and LLM Judgments: Understanding and Narrowing the Gap</a></h4>
</header>
<p>Felipe Maia Polo, Xinhe Wang, Mikhail Yurochkin, Gongjun Xu, Moulinath Banerjee, Yuekai Sun<br>
Neural Information Processing Systems (NeurIPS), 2025<br>
<a href="https://arxiv.org/pdf/2508.12792">arXiv</a> / <a href="https://github.com/felipemaiapolo/bridge">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/guru.png" width="220" height="210" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2506.14965">Revisiting Reinforcement Learning for LLM Reasoning from A Cross-Domain Perspective</a></h4>
</header>
<p>Zhoujun Cheng, Shibo Hao, Tianyang Liu, Fan Zhou, Yutao Xie, Feng Yao, Yuexin Bian, Yonghao Zhuang, Nilabjo Dey, Yuheng Zha, Yi Gu, Kun Zhou, Yuqi Wang, Yuan Li, Richard Fan, Jianshu She, Chengqian Gao, Abulhair Saparov, Haonan Li, Taylor W. Killian, Mikhail Yurochkin, Zhengzhong Liu, Eric P. Xing, Zhiting Hu<br>
Neural Information Processing Systems (NeurIPS), Datasets and Benchmarks Track, 2025<br>
<a href="https://arxiv.org/pdf/2506.14965">arXiv</a> / <a href="https://github.com/LLM360/Reasoning360">Code</a> / <a href="https://huggingface.co/datasets/LLM360/guru-RL-92k">Data</a> / <a href="https://huggingface.co/LLM360/guru-32B">Model</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/ood-synthetic.png" width="220" height="120" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2502.03323">Out-of-Distribution Detection using Synthetic Data Generation</a></h4>
</header>
<p>Momin Abbas, Muneeza Azmat, Raya Horesh, Mikhail Yurochkin<br>
Conference on Language Modeling (COLM), 2025<br>
<a href="https://arxiv.org/pdf/2502.03323">arXiv</a> / <a href="https://github.com/mominabbass/ood_synthetic">Code</a> / <a href="https://huggingface.co/datasets/abbasm2/synthetic_ood">Data</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/weak-to-strong.png" width="220" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2508.17018">Limitations of refinement methods for weak to strong generalization</a></h4>
</header>
<p>Seamus Somerstep, Ya'acov Ritov, Mikhail Yurochkin, Subha Maity, Yuekai Sun<br>
Conference on Language Modeling (COLM), 2025<br>
<a href="https://arxiv.org/pdf/2508.17018">arXiv</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/jd-lora.png" width="220" height="150" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2407.00066">Compress then Serve: Serving Thousands of LoRA Adapters with Little Overhead</a></h4>
</header>
<p>Rickard Brüel-Gabrielsson, Jiacheng Zhu, Onkar Bhardwaj, Leshem Choshen, Kristjan Greenewald, Mikhail Yurochkin, Justin Solomon<br>
International Conference on Machine Learning (ICML), 2025<br>
<a href="https://arxiv.org/pdf/2407.00066">arXiv</a> / <a href="https://huggingface.co/Lots-of-LoRAs">Lots of LoRAs</a> / <a href="https://icml.cc/virtual/2025/poster/46530">Slides and Poster</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/spri.png" width="210" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2502.03397">SPRI: Aligning Large Language Models with Context-Situated Principles</a></h4>
</header>
<p>Hongli Zhan, Muneeza Azmat, Raya Horesh, Junyi Jessy Li, Mikhail Yurochkin<br>
International Conference on Machine Learning (ICML), 2025<br>
<a href="https://arxiv.org/pdf/2502.03397">arXiv</a> / <a href="https://github.com/honglizhan/SPRI-public">Code</a> / <a href="https://icml.cc/virtual/2025/poster/44235">Slides and Poster</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/cosd.png" width="220" height="115" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2502.08020">Speculate, then Collaborate: Fusing Knowledge of Language Models during Decoding</a></h4>
</header>
<p>Ziyao Wang, Muneeza Azmat, Ang Li, Raya Horesh, Mikhail Yurochkin<br>
International Conference on Machine Learning (ICML), 2025<br>
<a href="https://arxiv.org/pdf/2502.08020">arXiv</a> / <a href="https://github.com/ziyaow1010/CoSD">Code</a> / <a href="https://icml.cc/media/PosterPDFs/ICML%202025/44974.png?t=1751054524.065811">Poster</a></p>
</div>
</div>
<hr />
<h2 style="text-align: right">2024</h2>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/prompteval.png" width="220" height="160" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2405.17202">Efficient multi-prompt evaluation of LLMs</a></h4>
</header>
<p>Felipe Maia Polo, Ronald Xu, Lucas Weber, Mírian Silva, Onkar Bhardwaj, Leshem Choshen, Allysson Flavio Melo de Oliveira, Yuekai Sun, Mikhail Yurochkin<br>
ICML Workshop on Efficient Systems for Foundation Models, 2024<br>
Neural Information Processing Systems (NeurIPS), 2024<br>
<a href="https://arxiv.org/pdf/2405.17202">arXiv</a> / <a href="https://github.com/felipemaiapolo/prompteval">Code</a> / <a href="https://huggingface.co/PromptEval">Data</a> / <a href="https://x.com/felipemaiapolo/status/1801251683311202522">Twitter</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/aot.png" width="200" height="160" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2406.05882">Distributional Preference Alignment of LLMs via Optimal Transport</a></h4>
</header>
<p>Igor Melnyk, Youssef Mroueh, Brian Belgodere, Mattia Rigotti, Apoorva Nitsure, Mikhail Yurochkin, Kristjan Greenewald, Jiri Navratil, Jerret Ross<br>
ICML Workshop on Models of Human Feedback for AI Alignment, 2024<br>
Neural Information Processing Systems (NeurIPS), 2024<br>
<a href="https://arxiv.org/pdf/2406.05882">arXiv</a> / <a href="https://huggingface.co/docs/trl/en/dpo_trainer">Available in TRL</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/weak-supervision.png" width="180" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2312.04601">Weak Supervision Performance Evaluation via Partial Identification</a></h4>
</header>
<p>Felipe Maia Polo, Subha Maity, Mikhail Yurochkin, Moulinath Banerjee, Yuekai Sun<br>
Neural Information Processing Systems (NeurIPS), 2024<br>
<a href="https://arxiv.org/pdf/2312.04601">arXiv</a> / <a href="https://github.com/felipemaiapolo/wsbounds">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/aligners.png" width="180" height="140" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2403.04224">Aligners: Decoupling LLMs and Alignment</a></h4>
</header>
<p>Lilian Ngweta, Mayank Agarwal, Subha Maity, Alex Gittens, Yuekai Sun, Mikhail Yurochkin<br>
Tiny Papers at the International Conference on Learning Representations (ICLR), 2024 <span style="font-weight: bold; color: #f24900">(Notable)</span><br>
Findings of the Association for Computational Linguistics: EMNLP, 2024<br>
<a href="https://arxiv.org/pdf/2403.04224">arXiv</a> / <a href="https://github.com/lilianngweta/aligners">Code</a> / <a href="https://huggingface.co/datasets/lilianngweta/aligners-datasets">Data</a> / <a href="https://huggingface.co/lilianngweta/ethical-aligner-7B">Model</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/llmrouting.png" width="220" height="160" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2309.15789">Large Language Model Routing with Benchmark Datasets</a></h4>
</header>
<p>Tal Shnitzer, Anthony Ou, Mírian Silva, Kate Soule, Yuekai Sun, Justin Solomon, Neil Thompson, Mikhail Yurochkin<br>
NeurIPS Workshop on Distribution Shifts (DistShift), 2023 <span style="font-weight: bold; color: #f24900">(Oral)</span><br>
Conference on Language Modeling (COLM), 2024<br>
<a href="https://arxiv.org/pdf/2309.15789">arXiv</a> / <a href="https://openreview.net/forum?id=LyNsMNNLjY">Code (see supplementary material)</a> </p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/pepr.png" width="180" height="140" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2405.11083">Prompt Exploration with Prompt Regression</a></h4>
</header>
<p>Michael Feffer, Ronald Xu, Yuekai Sun, Mikhail Yurochkin<br>
Conference on Language Modeling (COLM), 2024<br>
<a href="https://arxiv.org/pdf/2405.11083">arXiv</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/tinybenchmarks.png" width="200" height="160" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2402.14992">tinyBenchmarks: evaluating LLMs with fewer examples</a></h4>
</header>
<p>Felipe Maia Polo, Lucas Weber, Leshem Choshen, Yuekai Sun, Gongjun Xu, Mikhail Yurochkin<br>
ICLR Workshop on Mathematical and Empirical Understanding of Foundation Models, 2024<br>
International Conference on Machine Learning (ICML), 2024<br>
<a href="https://arxiv.org/pdf/2402.14992">arXiv</a> / <a href="https://github.com/felipemaiapolo/tinyBenchmarks">Code</a> / <a href="https://huggingface.co/tinyBenchmarks">Data</a> / <a href="https://twitter.com/felipemaiapolo/status/1765472691145085165">Twitter</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/lora-assymetry.png" width="220" height="160" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2402.16842">Asymmetry in Low-Rank Adapters of Foundation Models</a></h4>
</header>
<p>Jiacheng Zhu, Kristjan Greenewald, Kimia Nadjahi, Haitz Sáez de Ocáriz Borde, Rickard Brüel-Gabrielsson, Leshem Choshen, Marzyeh Ghassemi, Mikhail Yurochkin, Justin Solomon<br>
ICLR Workshop on Mathematical and Empirical Understanding of Foundation Models, 2024<br>
International Conference on Machine Learning (ICML), 2024<br>
<a href="https://arxiv.org/pdf/2402.16842">arXiv</a> / <a href="https://github.com/Jiacheng-Zhu-AIML/AsymmetryLoRA">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/risk-aware.png" width="200" height="160" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2310.07132">Risk Assessment and Statistical Significance in the Age of Foundation Models</a></h4>
</header>
<p>Apoorva Nitsure, Youssef Mroueh, Mattia Rigotti, Kristjan Greenewald, Brian Belgodere, Mikhail Yurochkin, Jiri Navratil, Igor Melnyk, Jerret Ross<br>
NeurIPS Workshop on Socially Responsible Language Modelling Research (SoLaR), 2023<br>
International Conference on Machine Learning (ICML), 2024<br>
<a href="https://arxiv.org/pdf/2310.07132">arXiv</a> / <a href="https://openreview.net/forum?id=fvse7bMkAs">Code (see supplementary material)</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/foe.png" width="200" height="155" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2310.01542">Fusing Models with Complementary Expertise</a></h4>
</header>
<p>Hongyi Wang, Felipe Maia Polo, Yuekai Sun, Souvik Kundu, Eric Xing, Mikhail Yurochkin<br>
NeurIPS Workshop on Distribution Shifts (DistShift), 2023<br>
International Conference on Learning Representations (ICLR), 2024<br>
<a href="https://arxiv.org/pdf/2310.01542">arXiv</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/stereotyping.png" width="220" height="140" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2310.01583">An Investigation of Representation and Allocation Harms in Contrastive Learning</a></h4>
</header>
<p>Subha Maity, Mayank Agarwal, Mikhail Yurochkin, Yuekai Sun<br>
International Conference on Learning Representations (ICLR), 2024<br>
<a href="https://arxiv.org/pdf/2310.01583">arXiv</a> / <a href="https://github.com/smaityumich/CL-representation-harm">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/dist-prop.png" width="200" height="155" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2402.08324">Uncertainty Quantification via Stable Distribution Propagation</a></h4>
</header>
<p>Felix Petersen, Aashwin Ananda Mishra, Hilde Kuehne, Christian Borgelt, Oliver Deussen, Mikhail Yurochkin<br>
International Conference on Learning Representations (ICLR), 2024<br>
<a href="https://arxiv.org/pdf/2402.08324">arXiv</a> / <a href="https://github.com/Felix-Petersen/distprop">Code</a></p>
</div>
</div>
<hr />
<h2 style="text-align: right">2023</h2>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/grasp.png" width="170" height="180" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2210.06759">Outlier-Robust Group Inference via Gradient Space Clustering</a></h4>
</header>
<p>Yuchen Zeng, Kristjan Greenewald, Luann Jung, Kangwook Lee, Justin Solomon, Mikhail Yurochkin<br>
NeurIPS Workshop on Distribution Shifts (DistShift), 2023<br>
<a href="https://arxiv.org/pdf/2210.06759.pdf">arXiv</a> / <a href="https://github.com/yzeng58/private_demographics">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/gcn-positions.png" width="190" height="140" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2201.12674">Rewiring with Positional Encodings for Graph Neural Networks</a></h4>
</header>
<p>Rickard Brüel-Gabrielsson, Mikhail Yurochkin, Justin Solomon<br>
Transactions on Machine Learning Research (TMLR), 2023<br>
<a href="https://arxiv.org/pdf/2201.12674.pdf">arXiv</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/pisco.png" width="220" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2302.09795">Simple Disentanglement of Style and Content in Visual Representations</a></h4>
</header>
<p>Lilian Ngweta, Subha Maity, Alex Gittens, Yuekai Sun, Mikhail Yurochkin<br>
International Conference on Machine Learning (ICML), 2023<br>
<a href="https://arxiv.org/pdf/2302.09795.pdf">arXiv</a> / <a href="https://github.com/lilianngweta/PISCO">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/kmixup.jpg" width="170" height="150" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2106.02933">k-Mixup Regularization for Deep Learning via Optimal Transport</a></h4>
</header>
<p>Kristjan Greenewald, Anming Gu, Mikhail Yurochkin, Justin Solomon, Edward Chien<br>
Transactions on Machine Learning Research (TMLR), 2023<br>
<a href="https://browse.arxiv.org/pdf/2106.02933.pdf">arXiv</a> / <a href="https://github.com/AnmingGu/kmixup-cifar10">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/birds.png" width="200" height="170" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2205.13577">Understanding new tasks through the lens of training data via exponential tilting</a></h4>
</header>
<p>Subha Maity, Mikhail Yurochkin, Moulinath Banerjee, Yuekai Sun<br>
International Conference on Learning Representations (ICLR), 2023<br>
<a href="https://arxiv.org/pdf/2205.13577.pdf">arXiv</a> / <a href="https://github.com/smaityumich/exponential-tilting">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/prox.png" width="190" height="160" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2201.11945">Learning Proximal Operators to Discover Multiple Optima</a></h4>
</header>
<p>Lingxiao Li, Noam Aigerman, Vladimir Kim, Jiajin Li, Kristjan Greenewald, Mikhail Yurochkin, Justin Solomon<br>
International Conference on Learning Representations (ICLR), 2023<br>
<a href="https://arxiv.org/pdf/2201.11945.pdf">arXiv</a> / <a href="https://github.com/lingxiaoli94/POL">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/MIED.png" width="190" height="160" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2210.13400">Sampling with Mollified Interaction Energy Descent</a></h4>
</header>
<p>Lingxiao Li, Qiang Liu, Anna Korba, Mikhail Yurochkin, Justin Solomon<br>
International Conference on Learning Representations (ICLR), 2023<br>
<a href="https://arxiv.org/pdf/2210.13400.pdf">arXiv</a> / <a href="https://github.com/lingxiaoli94/MIED">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/chi-if.png" width="220" height="220" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2303.00673">Fairness Evaluation in Text Classification: Machine Learning Practitioner Perspectives of Individual and Group Fairness</a></h4>
</header>
<p>Zahra Ashktorab, Benjamin Hoover, Mayank Agarwal, Casey Dugan, Werner Geyer, Hao Bang Yang, Mikhail Yurochkin<br>
CHI Conference on Human Factors in Computing Systems, 2023<br>
<a href="https://arxiv.org/pdf/2303.00673.pdf">arXiv</a></p> <!-- / <a href="https://fairbert.vizhub.ai/">Fairness Playground</a></p> -->
</div>
</div>
<hr />
<h2 style="text-align: right">2022</h2>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/constraints.png" width="220" height="160" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2301.06195">Calibrated Data-Dependent Constraints with Exact Satisfaction Guarantees</a></h4>
</header>
<p>Songkai Xue, Yuekai Sun, Mikhail Yurochkin<br>
Neural Information Processing Systems (NeurIPS), 2022 <span style="font-weight: bold; color: #f24900">(Oral)</span><br>
<a href="https://arxiv.org/pdf/2301.06195.pdf">arXiv</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/ifda.png" width="220" height="110" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2205.00504">Domain Adaptation meets Individual Fairness. And they get along</a></h4>
</header>
<p>Debarghya Mukherjee, Felix Petersen, Mikhail Yurochkin, Yuekai Sun<br>
Neural Information Processing Systems (NeurIPS), 2022</span><br>
<a href="https://arxiv.org/pdf/2205.00504.pdf">arXiv</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/fl_book.png" width="170" height="180" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://link.springer.com/chapter/10.1007/978-3-030-96896-0_7">Communication-Efficient Model Fusion</a></h4>
</header>
<p>Mikhail Yurochkin and Yuekai Sun<br>
Chapter 7 of Federated Learning: A Comprehensive Overview of Methods and Applications (edited by Heiko Ludwig and Nathalie Baracaldo), 2022</span><br>
<a href="https://link.springer.com/content/pdf/10.1007/978-3-030-96896-0_7.pdf">PDF</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/fl_book.png" width="170" height="180" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://link.springer.com/chapter/10.1007/978-3-030-96896-0_4">Personalization in Federated Learning</a></h4>
</header>
<p>Mayank Agarwal, Mikhail Yurochkin, Yuekai Sun<br>
Chapter 4 of Federated Learning: A Comprehensive Overview of Methods and Applications (edited by Heiko Ludwig and Nathalie Baracaldo), 2022</span><br>
<a href="https://link.springer.com/content/pdf/10.1007/978-3-030-96896-0_4.pdf">PDF</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/LES.png" width="220" height="100" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2202.01671">Log-Euclidean Signatures for Intrinsic Distances Between Unaligned Datasets</a></h4>
</header>
<p>Tal Shnitzer, Mikhail Yurochkin, Kristjan Greenewald, Justin Solomon<br>
International Conference on Machine Learning (ICML), 2022</span><br>
<a href="https://arxiv.org/pdf/2202.01671.pdf">arXiv</a> / <a href="https://github.com/shnitzer/les-distance">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/nlp-fair.png" width="220" height="150" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2108.01250">Your fairness may vary: Pretrained language model fairness in toxic text classification</a></h4>
</header>
<p>Ioana Baldini, Dennis Wei, Karthikeyan Natesan Ramamurthy, Mikhail Yurochkin, Moninder Singh<br>
Findings of ACL, 2022</span><br>
<a href="https://arxiv.org/pdf/2108.01250.pdf">arXiv</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/gp_sensitivity.png" width="220" height="150" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2106.06510">Measuring the sensitivity of Gaussian processes to kernel choice</a></h4>
</header>
<p>William Stephenson, Soumya Ghosh, Tin Nguyen, Mikhail Yurochkin, Sameer Deshpande, Tamara Broderick<br>
International Conference on Artificial Intelligence and Statistics (AISTATS), 2022 </span><br>
<a href="https://arxiv.org/pdf/2106.06510.pdf">arXiv</a> / <a href="https://github.com/checkGP/checkGP">Code</a></p>
</div>
</div>
<hr />
<h2 style="text-align: right">2021</h2>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/maml-sensitive-wide.png" width="220" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2110.13953">On sensitivity of meta-learning to support data</a></h4>
</header>
<p>Mayank Agarwal, Mikhail Yurochkin, Yuekai Sun<br>
Neural Information Processing Systems (NeurIPS), 2021</span><br>
<a href="https://arxiv.org/pdf/2110.13953.pdf">arXiv</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/if-pp.png" width="195" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2110.13796">Post-processing for Individual Fairness</a></h4>
</header>
<p>Felix Petersen, Debarghya Mukherjee, Yuekai Sun, Mikhail Yurochkin<br>
Neural Information Processing Systems (NeurIPS), 2021</span><br>
<a href="https://arxiv.org/pdf/2110.13796.pdf">arXiv</a> / <a href="https://github.com/Felix-Petersen/fairness-post-processing">Code</a> / <a href="https://www.youtube.com/watch?v=9PyKODDewPA">Video</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/subpop.png" width="220" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2011.03173">Does enforcing fairness mitigate biases caused by subpopulation shift?</a></h4>
</header>
<p>Subha Maity, Debarghya Mukherjee, Mikhail Yurochkin, Yuekai Sun<br>
Neural Information Processing Systems (NeurIPS), 2021</span><br>
<a href="https://arxiv.org/pdf/2011.03173.pdf">arXiv</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/ot-clustering-jmlr.png" width="220" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://www.jmlr.org/papers/v22/19-782.html">On Efficient Multilevel Clustering via Wasserstein Distances</a></h4>
</header>
<p>Viet Huynh, Nhat Ho, Nhan Dam, XuanLong Nguyen, Mikhail Yurochkin, Hung Bui, Dinh Phung<br>
Journal of Machine Learning Research (JMLR), 2021</span><br>
<a href="https://www.jmlr.org/papers/volume22/19-782/19-782.pdf">PDF</a> / <a href="https://github.com/viethhuynh/wasserstein-means">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/robot.png" width="220" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2012.07363">Outlier-Robust Optimal Transport</a></h4>
</header>
<p>Debarghya Mukherjee, Aritra Guha, Justin Solomon, Yuekai Sun, Mikhail Yurochkin<br>
International Conference on Machine Learning (ICML), 2021</span><br>
<a href="https://arxiv.org/pdf/2012.07363.pdf">arXiv</a> / <a href="https://github.com/debarghya-mukherjee/Robust-Optimal-Transport">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/sensei.png" width="220" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2006.14168">SenSeI: Sensitive Set Invariance for Enforcing Individual Fairness</a></h4>
</header>
<p>Mikhail Yurochkin and Yuekai Sun<br>
International Conference on Learning Representations (ICLR), 2021 <span style="font-weight: bold; color: #f24900">(Oral)</span><br>
<a href="https://arxiv.org/pdf/2006.14168.pdf">arXiv</a> / <a href="https://openreview.net/forum?id=DktZb97_Fx">Code (see supplementary material)</a> / <a href="https://iclr.cc/virtual/2021/oral/3499">Video</a> / <a href="https://www.research.ibm.com/blog/make-ai-fairer">Blog</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/ranking.png" width="220" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2103.11023">Individually Fair Rankings</a></h4>
</header>
<p>Amanda Bower, Hamid Eftekhari, Mikhail Yurochkin, Yuekai Sun<br>
International Conference on Learning Representations (ICLR), 2021<br>
<a href="https://arxiv.org/pdf/2103.11023.pdf">arXiv</a> / <a href="https://openreview.net/forum?id=71zCSP_HuBN">Code (see supplementary material)</a> / <a href="https://iclr.cc/virtual/2021/poster/2627">Video</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/testing2-pic.png" width="170" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2103.16714">Statistical inference for individual fairness</a></h4>
</header>
<p>Subha Maity, Songkai Xue, Mikhail Yurochkin, Yuekai Sun<br>
International Conference on Learning Representations (ICLR), 2021<br>
<a href="https://arxiv.org/pdf/2103.16714.pdf">arXiv</a> / <a href="https://github.com/smaityumich/individual-fairness-testing">Code</a> / <a href="https://iclr.cc/virtual/2021/poster/2966">Video</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/boosting.png" width="220" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2103.16785">Individually Fair Gradient Boosting</a></h4>
</header>
<p>Alexander Vargo, Fan Zhang, Mikhail Yurochkin, Yuekai Sun<br>
International Conference on Learning Representations (ICLR), 2021 <span style="font-weight: bold; color: #f24900">(Spotlight)</span><br>
<a href="https://arxiv.org/pdf/2103.16785.pdf">arXiv</a> / <a href="https://openreview.net/forum?id=JBAa9we1AL">Code (see supplementary material)</a> / <a href="https://iclr.cc/virtual/2021/spotlight/3502">Video</a></p>
</div>
</div>
<hr />
<h2 style="text-align: right">2020</h2>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/cwb.png" width="180" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2008.12534">Continuous Regularized Wasserstein Barycenters</a></h4>
</header>
<p>Lingxiao Li, Aude Genevay, Mikhail Yurochkin, Justin Solomon<br>
Neural Information Processing Systems (NeurIPS), 2020<br>
<a href="https://arxiv.org/pdf/2008.12534.pdf">arXiv</a> / <a href="https://github.com/lingxiaoli94/CWB">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/bloans.png" width="180" height="150" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2012.01193">Black Loans Matter: Distributionally Robust Fairness for Fighting Subgroup Discrimination</a></h4>
</header>
<p>Mark Weber, Mikhail Yurochkin, Sherif Botros, Vanio Markov<br>
NeurIPS Fair AI in Finance Workshop, 2020 <span style="font-weight: bold; color: #f24900">(Spotlight Talk)</span><br>
<a href="https://arxiv.org/pdf/2012.01193.pdf">arXiv</a> / <a href="https://mitibmwatsonailab.mit.edu/research/blog/black-loans-matter-fighting-bias-for-ai-fairness-in-lending/">Blog</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/kl-fusion.png" width="220" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2007.06168">Model Fusion with Kullback–Leibler Divergence</a></h4>
</header>
<p>Sebastian Claici, Mikhail Yurochkin, Soumya Ghosh, Justin Solomon<br>
International Conference on Machine Learning (ICML), 2020<br>
<a href="https://arxiv.org/pdf/2007.06168.pdf">arXiv</a> / <a href="https://github.com/IBM/KL-fusion">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/fair-metric.png" width="220" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2006.11439">Two Simple Ways to Learn Individual Fairness Metric from Data</a></h4>
</header>
<p>Debarghya Mukherjee, Mikhail Yurochkin, Moulinath Banerjee, Yuekai Sun<br>
International Conference on Machine Learning (ICML), 2020<br>
<a href="https://arxiv.org/pdf/2006.11439.pdf">arXiv</a> / <a href="https://github.com/debarghya-mukherjee/Fair_metric_learning">Code</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/testing1.png" width="220" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2003.05048">Auditing ML models for individual bias and unfairness</a></h4>
</header>
<p>Songkai Xue, Mikhail Yurochkin, Yuekai Sun<br>
International Conference on Artificial Intelligence and Statistics (AISTATS), 2020<br>
<a href="https://arxiv.org/pdf/2003.05048.pdf">arXiv</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/fedma.png" width="180" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/2002.06440">Federated Learning with Matched Averaging</a></h4>
</header>
<p>Hongyi Wang, Mikhail Yurochkin, Yuekai Sun, Dimitris Papailiopoulos, Yasaman Khazaeni<br>
International Conference on Learning Representations (ICLR), 2020 <span style="font-weight: bold; color: #f24900">(Oral)</span><br>
<a href="https://arxiv.org/pdf/2002.06440.pdf">arXiv</a> / <a href="https://github.com/IBM/FedMA">Code</a> / <a href="https://iclr.cc/virtual_2020/poster_BkluqlSFDS.html">Video</a> / <a href="https://mitibmwatsonailab.mit.edu/research/blog/fedma-layer-wise-federated-learning-with-the-potential-to-fight-ai-bias/">Blog</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/sensr.png" width="220" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/1907.00020">Training individually fair ML models with sensitive subspace robustness</a></h4>
</header>
<p>Mikhail Yurochkin, Amanda Bower, Yuekai Sun<br>
International Conference on Learning Representations (ICLR), 2020 <span style="font-weight: bold; color: #f24900">(Spotlight)</span><br>
<a href="https://arxiv.org/pdf/1907.00020.pdf">arXiv</a> / <a href="https://github.com/IBM/sensitive-subspace-robustness">Code</a> / <a href="https://iclr.cc/virtual_2020/poster_B1gdkxHFDH.html">Video</a> / <a href="https://mitibmwatsonailab.mit.edu/research/blog/training-individually-fair-ml-models-with-sensitive-subspace-robustness/">Blog</a></p>
</div>
</div>
<hr />
<h2 style="text-align: right">2019</h2>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/hott.png" width="220" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/1906.10827">Hierarchical Optimal Transport for Document Representation</a></h4>
</header>
<p>Mikhail Yurochkin, Sebastian Claici, Edward Chien, Farzaneh Mirzazadeh, Justin Solomon<br>
Neural Information Processing Systems (NeurIPS), 2019<br>
<a href="https://arxiv.org/pdf/1906.10827.pdf">arXiv</a> / <a href="https://github.com/IBM/HOTT">Code</a> / <a href="https://mitibmwatsonailab.mit.edu/research/blog/hierarchical-optimal-transport-for-document-representation/">Blog</a> / <a href="https://news.mit.edu/2019/finding-good-read-among-billions-of-choices-1220">MIT News</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/label-switching.png" width="190" height="150" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/1911.02053">Alleviating Label Switching with Optimal Transport</a></h4>
</header>
<p>Pierre Monteiller, Sebastian Claici, Edward Chien, Farzaneh Mirzazadeh, Justin Solomon, Mikhail Yurochkin<br>
Neural Information Processing Systems (NeurIPS), 2019<br>
<a href="https://arxiv.org/pdf/1911.02053.pdf">arXiv</a> / <a href="https://github.com/pierremon/label-switching">Code</a> / <a href="https://www.ibm.com/blogs/research/2019/12/optimal-transport-for-label-switching/">Blog</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/spahm.png" width="220" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/1911.00218">Statistical Model Aggregation via Parameter Matching</a></h4>
</header>
<p>Mikhail Yurochkin, Mayank Agarwal, Soumya Ghosh, Kristjan Greenewald, Trong Nghia Hoang<br>
Neural Information Processing Systems (NeurIPS), 2019<br>
<a href="https://arxiv.org/pdf/1911.00218.pdf">arXiv</a> / <a href="https://github.com/IBM/SPAHM">Code</a> / <a href="https://mitibmwatsonailab.mit.edu/research/blog/statistical-model-aggregation-via-parameter-matching/">Blog</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/sddm.png" width="180" height="150" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/1809.08738">Scalable inference of topic evolution via models for latent geometric structures</a></h4>
</header>
<p>Mikhail Yurochkin, Zhiwei Fan, Aritra Guha, Paraschos Koutris, XuanLong Nguyen<br>
Neural Information Processing Systems (NeurIPS), 2019<br>
<a href="https://arxiv.org/pdf/1809.08738.pdf">arXiv</a> / <a href="https://github.com/moonfolk/SDDM">Code</a> / <a href="https://mitibmwatsonailab.mit.edu/research/blog/scalable-inference-of-topic-evolution-via-models-for-latent-geometric-structures/">Blog</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/vlad.png" width="180" height="130" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/1905.11009">Dirichlet Simplex Nest and Geometric Inference</a></h4>
</header>
<p>Mikhail Yurochkin, Aritra Guha, Yuekai Sun, XuanLong Nguyen<br>
International Conference on Machine Learning (ICML), 2019 <span style="font-weight: bold; color: #f24900">(Long Talk)</span><br>
<a href="https://arxiv.org/pdf/1905.11009.pdf">arXiv</a> / <a href="https://github.com/moonfolk/VLAD">Code</a> / <a href="https://slideslive.com/38917849/bayesian-methods?t=0">Video</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/pfnm-long.png" width="200" height="150" alt="" style="border:none;" /></div>
</div>
<div class="9u$ 8u$(large) 12u$(medium)">
<header>
<h4><a href="https://arxiv.org/abs/1905.12022">Bayesian Nonparametric Federated Learning of Neural Networks</a></h4>
</header>
<p>Mikhail Yurochkin, Mayank Agarwal, Soumya Ghosh, Kristjan Greenewald, Trong Nghia Hoang, Yasaman Khazaeni<br>
International Conference on Machine Learning (ICML), 2019<br>
<a href="https://arxiv.org/pdf/1905.12022.pdf">arXiv</a> / <a href="https://github.com/IBM/probabilistic-federated-neural-matching">Code</a> / <a href="https://youtube.videoken.com/embed/-COX4Por6H0?tocitem=95">Video</a></p>
</div>
</div>
<div class="row 200%">
<div class="3u 4u(large) 12u$(medium)">
<div class="image"><img src="images/gcnucb.png" width="180" height="150" alt="" style="border:none;" /></div>
</div>