-
Notifications
You must be signed in to change notification settings - Fork 0
/
index.html
484 lines (432 loc) · 20.3 KB
/
index.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<!-- saved from url=(0041)https://people.eecs.berkeley.edu/~barron/ -->
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=windows-1252">
<meta name="viewport" content="width=device-width, initial-scale=1">
<meta name="generator" content="HTML Tidy for Linux/x86 (vers 11 February 2007), see www.w3.org">
<style type="text/css">
/* Color scheme stolen from Sergey Karayev */
a {
color: #07889b;
/*#1772d0;*/
text-decoration: none;
}
a:focus,
a:hover {
color: #e37222;
/*#f7b733;*/
/*f09228;*/
text-decoration: none;
}
body,
td,
th,
tr,
p {
font-family: "TeXGyrePagella", "Palatino Linotype", "Book Antiqua", Palatino, serif;
/*'Lato', Verdana, Helvetica, sans-serif;*/
font-size: 15px;
/*14*/
}
a {
font-family: "TeXGyrePagella", "Palatino Linotype", "Book Antiqua", Palatino, serif;
/*'Lato', Verdana, Helvetica, sans-serif;*/
}
strong {
font-family: "TeXGyrePagella", "Palatino Linotype", "Book Antiqua", Palatino, serif;
/*'Lato', Verdana, Helvetica, sans-serif;*/
font-size: 15px;
/*14*/
}
heading {
font-family: "TeXGyrePagella", "Palatino Linotype", "Book Antiqua", Palatino, serif;
/*'Lato', Verdana, Helvetica, sans-serif;*/
font-size: 22px;
color: #e37222;
/*#fc4a1a;*/
}
heading2 {
font-family: "TeXGyrePagella", "Palatino Linotype", "Book Antiqua", Palatino, serif;
/*'Lato', Verdana, Helvetica, sans-serif;*/
font-size: 18px;
}
papertitle {
font-family: "TeXGyrePagella", "Palatino Linotype", "Book Antiqua", Palatino, serif;
/*'Lato', Verdana, Helvetica, sans-serif;*/
font-size: 15px;
/*14*/
font-weight: 700;
}
name {
font-family: "TeXGyrePagella", "Palatino Linotype", "Book Antiqua", Palatino, serif;
/*'Lato', Verdana, Helvetica, sans-serif;*/
font-size: 42px;
}
li:not(:last-child) {
margin-bottom: 5px;
}
.one {
width: 160px;
height: 160px;
position: relative;
}
.two {
width: 160px;
height: 160px;
position: absolute;
transition: opacity .2s ease-in-out;
-moz-transition: opacity .2s ease-in-out;
-webkit-transition: opacity .2s ease-in-out;
}
.fade {
transition: opacity .2s ease-in-out;
-moz-transition: opacity .2s ease-in-out;
-webkit-transition: opacity .2s ease-in-out;
}
span.highlight {
background-color: #ffffd0;
}
</style>
<link rel="icon" type="image/png" href="https://people.eecs.berkeley.edu/~barron/seal_icon.png">
<title>Junwei Liao</title>
<link href="./files/css" rel="stylesheet" type="text/css">
<style id="dark-reader-style" type="text/css">
@media screen {
/* Leading rule */
/*html {
-webkit-filter: brightness(100%) contrast(100%) grayscale(20%) sepia(10%) !important;
}*/
/* Text contrast */
html {
text-shadow: 0 0 0 !important;
}
/* Full screen */
*:-webkit-full-screen,
*:-webkit-full-screen * {
-webkit-filter: none !important;
}
/* Page background */
html {
background: rgb(255, 255, 255) !important;
}
}
</style>
<script type="text/javascript">
function visibility_on(id) {
var e = document.getElementById(id + "_text");
if (e.style.display == 'none')
e.style.display = 'block';
var e = document.getElementById(id + "_img");
if (e.style.display == 'none')
e.style.display = 'block';
}
function visibility_off(id) {
var e = document.getElementById(id + "_text");
if (e.style.display == 'block')
e.style.display = 'none';
var e = document.getElementById(id + "_img");
if (e.style.display == 'block')
e.style.display = 'none';
}
function toggle_visibility(id) {
var e = document.getElementById(id + "_text");
if (e.style.display == 'inline')
e.style.display = 'block';
else
e.style.display = 'inline';
var e = document.getElementById(id + "_img");
if (e.style.display == 'inline')
e.style.display = 'block';
else
e.style.display = 'inline';
}
function toggle_vis(id) {
var e = document.getElementById(id);
if (e.style.display == 'none')
e.style.display = 'inline';
else
e.style.display = 'none';
}
</script>
</head>
</div>
<table width="800" border="0" align="center" cellspacing="0" cellpadding="0">
<tbody>
<tr>
<td>
<table width="100%" align="center" border="0" cellspacing="0" cellpadding="20">
<tbody>
<tr>
<td width="67%" valign="middle">
<p align="center">
<name>Junwei (Jaden) Liao</name><br>
jwliao dot ai at gmail dot com
</p>
<p>
I am a senior undergrad majoring in Artificial Intelligence at
<a href="http://en.xjtu.edu.cn/">Xi'an Jiaotong University</a>
and was once a visiting student at
<a href="https://www.berkeley.edu/" target="_blank">University of California, Berkeley</a>
2023 Fall.
</p>
</p>
<p>
<i>I am interested in Reinforcement Learning, Large Decision Models,
specifically LLM-based Agents and unleashing the full potential of Agents
through RL and other advanced techniques.</i>
</p>
<p align="center">
<!-- <a href="http://ai.stanford.edu/~cbfinn/bio.txt">Bio</a> / -->
<a href="mailto:[email protected]">Email</a> /
<a href="https://github.com/jwliao-ai" target="_blank"> GitHub </a> /
<a href="https://www.linkedin.com/in/jwliao" target="_blank"> LinkedIn </a>
</p>
</td>
<td width="33%">
<a href="files/JunweiLiao.jpg"><img width="280" alt="profile photo" src="files/JunweiLiao.jpg"></a>
</td>
</tr>
</tbody>
</table>
<table width="100%" align="center" border="0" cellspacing="0" cellpadding="20">
<tbody>
<tr>
<td>
<heading>Education</heading>
<ul>
<li>
<b>Sep. 2021 ~ Present</b>: B.S. in Artificial Intelligence at
<a href="http://en.xjtu.edu.cn/">Xi'an Jiaotong University</a>.
</li>
<li>
<b>Aug. 2023 ~ Dec. 2023</b>: Visiting Student at
<a href="https://www.berkeley.edu/" target="_blank">University of California, Berkeley</a>.
</li>
</ul>
</td>
</tr>
<tr>
<td>
<heading>Research Experience</heading>
<ul>
<li><b>May. 2024 ~ Present</b>: RA at
<a href="http://en.sjtu.edu.cn/">Shanghai Jiao Tong University</a>,
advised by
<a href="https://wnzhang.net/" target="_blank">Weinan Zhang</a>.
</li>
<p>Focus on Deep Reinforcement Learning, Large Decision Models and Agent Technology.</p>
<li><b>Sep. 2023 ~ Mar. 2024</b>: RA at
<a href="https://www.tsinghua.edu.cn/" target="_blank">Tsinghua University</a>,
advised by
<a href="https://juren1987.github.io/" target="_blank">Ju Ren</a>.
</li>
<p>Focus on Deep Reinforcement Learning and RLHF/RLAIF.</p>
</ul>
</td>
</tr>
<!-- <tr><td>
<heading>News</heading>
<p>See our <a href="https://irislab.stanford.edu/">lab website</a> for up-to-date news.</p>
<ul>
<li>I am honored and thrilled to have received the <a href="https://awards.acm.org/about/2018-doctoral-dissertation">ACM 2018 Doctoral Dissertation Award</a> for my thesis, <a href="_files/dissertation.pdf">Learning to Learn with Gradients</a>.</li>
<li> <a href="javascript:toggle_vis('news')">show more</a> </li>
<div id="news" style="display:none">
<li>I co-organized a workshop at NIPS 2016 on <a href="https://sites.google.com/site/nips16interaction/">Deep Learning for Action and Interaction</a> (<a href="https://www.youtube.com/watch?v=vTgwWobuoFw&list=PL_iWQOsE6TfVCLmikLdaQOBntJuCZLwQY&index=1">videos here</a>). </li>
</div>
</ul>
</td></tr> -->
<!--
<tr><td>
<heading>Blog Posts</heading>
<ul>
<li> <a href="https://ai.stanford.edu/blog/prototransformer/">Meta-Learning Student Feedback to 16,000 Solutions</a>: our work on studying meta-learning for education and how we can scale student feedback.</li>
<li> <a href="javascript:toggle_vis('blogs')">show more</a> </li>
<div id="blogs" style="display:none">
<li><a href="https://research.googleblog.com/2016/10/how-robots-can-acquire-new-skills-from.html">How Robots Can Acquire New Skills from Their Shared Experience</a>:
features some work done by me and my colleagues at Google Brain, X, and DeepMind on learning across multiple robots. It was nicely summarized
by the MIT Technology Review <a href="https://www.technologyreview.com/s/602529/google-builds-a-robotic-hive-mind-kindergarten/">here</a>. </li>
</div>
</ul>
</td></tr>
-->
<!-- <tr><td>
<heading>Recent Talks</heading><br><br>
<b>Robotics Focused Talk (November 2022)</b>
<iframe width="780" height="439" src="https://www.youtube.com/embed/m8pQeXe7J0w" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe>
<br><br>
<b>Machine Learning Focused Talk (June 2022)</b>
<iframe width="780" height="439" src="https://www.youtube.com/embed/gtDo9njJKb8" frameborder="0" allow="accelerometer; autoplay; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe>
</td></tr> -->
<!-- <table width="100%" align="center" border="0" cellspacing="0" cellpadding="20">
<heading>Teaching</heading>
</table>
<table width="100%" align="center" border="0" cellpadding="20">
<tbody><tr>
<td width="75%" valign="center">
<p>
<a href="https://stanford-cs221.github.io/spring2021/"><papertitle>Stanford CS221: Artificial Intelligence: Principles and Techniques</papertitle></a> - Spring 2020, Spring 2021 <br>
</p>
</td>
</tr>
</tbody></table> -->
<!-- <tr><td>
<heading>Tutorials and Lectures</heading>
<ul>
<li> In August 2017, I gave guest lectures on model-based reinforcement learning and inverse reinforcement learning at the <a href="https://www.deepbootcamp.io/">Deep RL Bootcamp</a> (slides <a href="_files/mbrl_bootcamp.pdf">here</a> and <a href="_files/bootcamp_inverserl.pdf">here</a>, videos <a href="https://youtu.be/iC2a7M9voYU">here</a> and <a href="https://youtu.be/d9DlQSJQAoI">here</a>). </li>
</ul>
</td></tr> -->
<!-- <tr><td>
<heading>Invited Talks</heading>
<ul>
<li>I gave a talk on meta-learning (<a href="_files/samsung_ai_forum.pdf">slides here</a>, <a href="https://youtu.be/AhvFggX2xow?t=3379">video here</a>) at the Samsung AI Forum in 2020.</li>
<li> <a href="javascript:toggle_vis('talks')">show more</a> </li>
<div id="talks" style="display:none">
<li>At RSS 2019, I gave invited talks in the workshops on Simulation to Real World Transfer (<a href="_files/rss19_sim2real.pdf">slides here</a>), the workshop on Task-Informed Grasping (<a href="_files/rss19_tig.pdf">slides here</a>), and the workshop on Women in Robotics (<a href="_files/rss19_women.pdf">slides here</a>) </li>
</div>
</ul>
</td></tr> -->
<td width="100%" valign="middle">
<heading>Preprints</heading> <br><br>
<div onmouseover="document.getElementById('agenticir').style.display = 'block';"
onmouseout="document.getElementById('agenticir').style.display='none';">
<a href="https://arxiv.org/pdf/2410.09713" target="_blank">
<papertitle>Agentic Information Retrieval</papertitle>
</a><br>
<a href="https://wnzhang.net/" target="_blank">Weinan Zhang</a>,
<i>Junwei Liao</i>,
Ning Li,
Kounianhua Du
<br>
<a href="https://arxiv.org/abs/2410.09713" target="_blank">arXiv</a> / <a href="https://arxiv.org/html/2410.09713"
target="_blank">html</a>
<br>
<div id="agenticir" style="display:none">
In this position paper, we introduce Agentic Information Retrieval (Agentic IR), a novel IR paradigm shaped by the
capabilities of LLM agents.
</div>
</div><br>
<div onmouseover="document.getElementById('hammer').style.display = 'block';"
onmouseout="document.getElementById('hammer').style.display='none';">
<a href="http://arxiv.org/pdf/2410.04587" target="_blank">
<papertitle>Hammer: Robust Function-Calling for On-Device Language Models via Function Masking</papertitle>
</a><br>
Qiqiang Lin*,
Muning Wen*,
Qiuying Peng*,
Guanyu Nie,
<i>Junwei Liao</i>,
Jun Wang,
Xiaoyun Mo,
Jiamu Zhou,
Cheng Cheng,
Yin Zhao,
Jun Wang,
<a href="https://wnzhang.net/" target="_blank">Weinan Zhang</a>
<br>
<a href="https://arxiv.org/abs/2410.04587" target="_blank">arXiv</a> / <a href="https://github.com/MadeAgents/Hammer"
target="_blank">code</a> / <a href="https://arxiv.org/html/2410.04587" target="_blank">html</a> / <a
href="https://huggingface.co/MadeAgents" target="_blank">dataset and
models</a>
</div>
<div id="hammer" style="display:none">
In this paper, we introduce Hammer, a novel family of foundation
models specifically engineered for on-device function calling. Hammer employs
an augmented dataset that enhances models' sensitivity to irrelevant functions and
incorporates function masking techniques to minimize misleading.
</div><br>
<div onmouseover="document.getElementById('etpo').style.display = 'block';"
onmouseout="document.getElementById('etpo').style.display='none';">
<a href="https://arxiv.org/pdf/2402.06700v4" target="_blank">
<papertitle>Entropy-Regularized Token-Level Policy Optimization for Language Agent Reinforcement</papertitle>
</a><br>
Muning Wen*,
<i>Junwei Liao</i>*,
<a href="https://www.cdeng.net/" target="_blank">Cheng Deng</a>,
<a href="http://www0.cs.ucl.ac.uk/staff/Jun.Wang/" target="_blank">Jun Wang</a>,
<a href="https://wnzhang.net/" target="_blank">Weinan Zhang</a>,
<a href="https://yingwen.io/" target="_blank">Ying Wen</a>
<br>
<a href="https://arxiv.org/abs/2402.06700" target="_blank">arXiv</a> /
<a href="https://github.com/morning9393/ETPO" target="_blank">code</a> / <a href="https://arxiv.org/html/2402.06700v4"
target="_blank">html</a>
<br>
<div id="etpo" style="display:none">
In this paper, we introduce Entropy Regularized Token-level Policy Optimization
(ETPO), an entropy-augmented RL method tailored for optimizing LLMs at the token level.
</div>
</div><br>
</td>
<!-- <td width="100%" valign="middle">
<heading>Selected Publications (<a href="https://irislab.stanford.edu/publications.html">See all</a>)</heading> <br><br>
<div onmouseover="document.getElementById('tag').style.display = 'block';"
onmouseout="document.getElementById('tag').style.display='none';">
<a href="https://arxiv.org/pdf/2109.04617">
<papertitle>Efficiently Identifying Task Groupings for Multi-Task Learning</papertitle></a><br>
<a href="https://www.linkedin.com/in/christopher-fifty/">Christopher Fifty</a>,
<a href="https://scholar.google.com/citations?user=F6omR3gAAAAJ&hl=en">Ehsan Amid</a>,
<a href="http://www-personal.umich.edu/~zhezhao/">Zhe Zhao</a>,
<a href="https://cs.stanford.edu/~tianheyu/">Tianhe Yu</a>,
<a href="https://scholar.google.com/citations?user=Mv71-IcAAAAJ/">Rohan Anil</a>,
<i>Chelsea Finn</i><br>
<em>Neural Information Processing Systems (NeurIPS)</em>, 2021 <font color="#e37222"><strong>(Spotlight)</strong></font> <br>
<a href="https://arxiv.org/abs/2109.04617">arXiv</a> / <a href="https://github.com/google-research/google-research/tree/master/tag">code</a>
</div>
<div id="tag" style="display:none">
</div><br>
<div onmouseover="document.getElementById('e2e').style.display = 'block';"
onmouseout="document.getElementById('e2e').style.display='none';">
<a href="http://www.jmlr.org/papers/volume17/15-522/15-522.pdf">
<papertitle>End-to-End Training of Deep Visuomotor Policies</papertitle></a><br>
<a href="https://people.eecs.berkeley.edu/~svlevine/">Sergey Levine*</a>,
<i>Chelsea Finn</i>*, <a href="https://people.eecs.berkeley.edu/~trevor/">Trevor Darrell</a>,
<a href="https://people.eecs.berkeley.edu/~pabbeel/">Pieter Abbeel</a><br>
<strong style="color:#e37222">CCC Blue Sky Ideas <a href="http://www.cccblog.org/2015/08/03/blue-sky-ideas-aaai-rss-special-workshop-on-the-50th-anniversary-of-shakey/">Award</a></strong><br>
<em>Journal of Machine Learning Research (JMLR)</em>, 2016 <br>
<a href="https://arxiv.org/abs/1504.00702">arXiv</a> /
<a href="https://sites.google.com/site/visuomotorpolicy/">video</a> /
<a href="http://rll.berkeley.edu/deeplearningrobotics">project page</a> /
<a href="http://rll.berkeley.edu/gps">code</a>
<br>
<div id="e2e" style="display:none">
We demonstrate a deep neural network trained end-to-end, from perception to controls, for robotic manipulation tasks.
</div>
</div><br>
</td> -->
</tbody>
</table>
<table width="100%" align="center" border="0" cellspacing="0" cellpadding="20">
<tbody>
<tr>
<td>
<br>
<p align="right">
<font size="2">
Last Updated: Oct. 15, 2024
</font>
</p>
<p align="right">
<font size="2">
<a href="https://people.eecs.berkeley.edu/~barron/">This guy makes a nice webpage.</a>
</font>
</p>
</td>
</tr>
</tbody>
</table>
<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-59618557-1', 'auto');
ga('send', 'pageview');
</script>
</td>
</tr>
</tbody>
</table>
</body>
</html>