forked from jonbarron/jonbarron.github.io
-
Notifications
You must be signed in to change notification settings - Fork 1
Expand file tree
/
Copy pathindex.html
More file actions
executable file
·483 lines (409 loc) · 18 KB
/
index.html
File metadata and controls
executable file
·483 lines (409 loc) · 18 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
<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Qingyi Si (佀庆一)</title>
<style>
:root {
--primary-color: #2c3e50;
--link-color: #2980b9;
--hover-color: #e74c3c;
--bg-color: #ffffff;
--text-color: #333333;
--border-color: #ecf0f1;
--secondary-bg: #f8f9fa;
}
* {
margin: 0;
padding: 0;
box-sizing: border-box;
}
body {
font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", Roboto, "Helvetica Neue", Arial, sans-serif;
line-height: 1.6;
color: var(--text-color);
background-color: var(--bg-color);
max-width: 900px;
margin: 0 auto;
padding: 20px;
}
/* Header */
header {
border-bottom: 2px solid var(--primary-color);
padding-bottom: 20px;
margin-bottom: 30px;
}
h1 {
font-size: 2.2em;
color: var(--primary-color);
margin-bottom: 5px;
}
.subtitle {
font-size: 1.1em;
color: #666;
margin-bottom: 10px;
}
.contact {
font-size: 0.95em;
color: #555;
}
.contact a {
color: var(--link-color);
text-decoration: none;
margin-right: 15px;
}
.contact a:hover {
color: var(--hover-color);
text-decoration: underline;
}
/* Navigation */
nav {
background-color: var(--secondary-bg);
padding: 12px 20px;
margin-bottom: 30px;
border-radius: 5px;
}
nav a {
color: var(--primary-color);
text-decoration: none;
margin-right: 25px;
font-weight: 500;
font-size: 1.05em;
}
nav a:hover {
color: var(--hover-color);
}
/* Sections */
section {
margin-bottom: 35px;
}
h2 {
font-size: 1.5em;
color: var(--primary-color);
border-bottom: 2px solid var(--border-color);
padding-bottom: 8px;
margin-bottom: 15px;
}
h3 {
font-size: 1.2em;
color: var(--primary-color);
margin-top: 20px;
margin-bottom: 10px;
}
/* News */
.news-item {
margin-bottom: 8px;
padding-left: 15px;
position: relative;
}
.news-item::before {
content: "▸";
position: absolute;
left: 0;
color: var(--link-color);
}
.news-date {
font-weight: 600;
color: var(--primary-color);
}
/* Bio */
.bio {
text-align: justify;
line-height: 1.8;
}
.highlight {
background-color: #fff3cd;
padding: 2px 5px;
border-radius: 3px;
}
/* Publications */
.pub-item {
margin-bottom: 15px;
padding-left: 10px;
border-left: 3px solid var(--border-color);
}
.pub-title {
font-weight: 600;
color: var(--primary-color);
}
.pub-authors {
color: #555;
}
.pub-venue {
font-style: italic;
color: #777;
}
.pub-links a {
color: var(--link-color);
text-decoration: none;
margin-right: 10px;
font-size: 0.9em;
}
.pub-links a:hover {
text-decoration: underline;
}
.venue-tag {
display: inline-block;
background-color: var(--primary-color);
color: white;
padding: 2px 8px;
border-radius: 3px;
font-size: 0.85em;
margin-right: 8px;
}
/* Lists */
ul {
list-style: none;
padding-left: 0;
}
li {
margin-bottom: 8px;
padding-left: 15px;
position: relative;
}
li::before {
content: "•";
position: absolute;
left: 0;
color: var(--link-color);
}
/* Footer */
footer {
margin-top: 50px;
padding-top: 20px;
border-top: 1px solid var(--border-color);
text-align: center;
color: #777;
font-size: 0.9em;
}
/* Responsive */
@media (max-width: 600px) {
body {
padding: 15px;
}
h1 {
font-size: 1.8em;
}
nav a {
display: block;
margin-bottom: 8px;
}
}
</style>
<base target="_blank">
</head>
<body>
<header>
<h1>Qingyi Si (佀庆一)</h1>
<div class="subtitle">LLM Researcher @ JD Explore Academy｜ VLM & RL & Inference Efficiency</div>
<div class="contact">
<a href="mailto:siqingyi.phoebus@jd.com">Email</a>
<a href="https://scholar.google.com/citations?view_op=list_works&hl=en&hl=en&user=escx2WMAAAAJ" target="_blank">Google Scholar</a>
<a href="https://github.com/PhoebusSi" target="_blank">GitHub</a>
<a href="https://www.xiaohongshu.com/user/profile/5b55441ef7e8b93a75f833b1?xsec_token=YByIa-PWL1OGGTFisk4-lTyW-Eot2OKtO9HiyNWRJHLXs=&xsec_source=app_share&xhsshare=CopyLink&shareRedId=N0g3NTc5NU5MPko6S0EzRz04Tz83OEg6&apptime=1769350852&share_id=46701201e6574612b4809bfc0cdedcc3" target="_blank">Xiaohongshu</a>
</div>
</header>
<section id="about">
<h2>About Me</h2>
<p class="bio">
I am currently an LLM researcher at <strong>JD Explore Academy</strong> (京东探索研究院), through the <strong>TGT</strong> Program, working with <a href="https://scholar.google.com/citations?user=GDvt570AAAAJ&hl=zh-CN" target="_blank">Jiaqi Wang</a>, and <a href="https://scholar.google.com/citations?user=Qaa6OxIAAAAJ&hl=en" target="_blank">Nan Duan</a>, focusing on <span class="highlight">Multimodal Understanding</span>, <span class="highlight">Post-training (RL & SFT)</span> and <span class="highlight">Inference Efficiency</span>.
</p>
<p class="bio" style="margin-top: 10px;">
Previously, I was with Huawei ICT BG as a member of TopMinds program (天才少年), and received my Ph.D. from the <strong>Institute of Information Engineering, Chinese Academy of Sciences</strong> in 2024, advised by <a href="https://people.ucas.ac.cn/~linzheng" target="_blank">Prof. Zheng Lin</a>
and <a href="https://people.ucas.ac.cn/~wwp" target="_blank"> Prof. Weiping Wang</a>.
I have published 30+ peer-reviewed papers in top-tier AI conferences including <strong>ACL, EMNLP, ICLR, NeurIPS, ICRA, IROS, AAAI, IJCAI, MM</strong>. My open-source projects have accumulated <strong>3000+ stars</strong> on GitHub.
</p>
<p class="bio" style="margin-top: 10px;">
We're hiring interns for the TGT program! If you have top-tier publications and are passionate about video understanding and VLMs, we'd love to hear from you.
</p>
</section>
<section id="news">
<h2>Recent News</h2>
<div class="news-item">
<span class="news-date">[2026.2]</span> One paper accepted to <strong>CVPR 2026</strong>.
</div>
<div class="news-item">
<span class="news-date">[2026.1]</span> Two papers accepted to <strong>ICLR 2026</strong>.
</div>
<div class="news-item">
<span class="news-date">[2025.11]</span> Three papers accepted to <strong>AAAI 2026</strong>.
</div>
<div class="news-item">
<span class="news-date">[2025.10]</span> One paper accepted to <strong>IROS workshop 2025</strong>.
</div>
<div class="news-item">
<span class="news-date">[2025.09]</span> Two papers accepted to <strong>NeurIPS 2025</strong>.
</div>
<div class="news-item">
<span class="news-date">[2025.06]</span> Awarded <strong>Major Contribution Special Award</strong> at Huawei.
</div>
<div class="news-item">
<span class="news-date">[2025.01]</span> One paper accepted to <strong>ACL 2025</strong>.
</div>
</section>
<section id="publications">
<h2>Selected Publications</h2>
<p style="margin-bottom: 15px; color: #666; font-size: 0.95em;">
* indicates co-first author, † indicates corresponding author. Full list available on <a href="https://scholar.google.com/citations?view_op=list_works&hl=en&hl=en&user=escx2WMAAAAJ" style="color: var(--link-color);">Google Scholar</a>.
</p>
<h3>LLM Test-time Optimization</h3>
<div class="pub-item">
<span class="venue-tag">Preprint</span>
<span class="pub-title">System 1&2 Synergy via Dynamic Model Interpolation</span>
<div class="pub-authors">Chenxu Yang*, <u>Qingyi Si*</u>, Chong Tian, Xiyu Liu, Dingyu Yao, Chuanyu Qin, Zheng Lin, Weiping Wang, Jiaqi Wang</div>
<div class="pub-venue">Preprint</div>
<div class="pub-links">
<a href="https://arxiv.org/abs/2601.21414">[PDF]</a>
</div>
</div>
<div class="pub-item">
<span class="venue-tag">AAAI'26</span>
<span class="pub-title">Test-time Prompt Intervention</span>
<div class="pub-authors">Chenxu Yang*, <u>Qingyi Si*</u>, Mz Dai*, Dingyu Yao, Mingyu Zheng, Minghui Chen, Zheng Lin, Weiping Wang</div>
<div class="pub-venue">AAAI 2026 </div>
<div class="pub-links">
<a href="https://arxiv.org/abs/2508.02511">[PDF]</a>
</div>
</div>
<div class="pub-item">
<span class="venue-tag">ICLR'26</span>
<span class="pub-title">Dynamic Early Exit in Reasoning Models</span>
<div class="pub-authors">Chenxu Yang*, <u>Qingyi Si*</u>, Yongjie Duan, Zheliang Zhu, Chenyu Zhu, Qiaowei Li, Minghui Chen, Zheng Lin, Weiping Wang</div>
<div class="pub-venue">ICLR 2026</div>
<div class="pub-links">
<a href="https://arxiv.org/abs/2504.15895">[PDF]</a>
<a href="https://github.com/iie-ycx/DEER">[Code]</a>
</div>
</div>
<h3>Reinforcement Learning and Post-training</h3>
<div class="pub-item">
<span class="venue-tag">Preprint</span>
<span class="pub-title">Outcome-Grounded Advantage Reshaping for Fine-Grained Credit Assignment in Mathematical Reasoning</span>
<div class="pub-authors">Ziheng Li, Liu Kang, Feng Xiao, Luxi Xing, <u>Qingyi Si</u>, Zhuoran Li, Weikang Gong, Deqing Yang, Yanghua Xiao, Hongcheng Guo</div>
<div class="pub-venue">Preprint</div>
<div class="pub-links">
<a href="https://arxiv.org/abs/2601.07408">[PDF]</a>
</div>
</div>
<div class="pub-item">
<span class="venue-tag">NeurIPS'25</span>
<span class="pub-title">S-GRPO: Early Exit via Reinforcement Learning in Reasoning Models</span>
<div class="pub-authors">Muzhi Dai*, Chenxu Yang*, <u>Qingyi Si†</u></div>
<div class="pub-venue">NeurIPS 2025</div>
<div class="pub-links">
<a href="https://arxiv.org/abs/2505.07686">[PDF]</a>
</div>
</div>
<div class="pub-item">
<span class="venue-tag">NeurIPS'25 workshop</span>
<span class="pub-title">Stable Reinforcement Learning for Efficient Reasoning</span>
<div class="pub-authors">Muzhi Dai*, Shixuan Liu*, <u>Qingyi Si†</u></div>
<div class="pub-venue">NeurIPS workshop 2025</div>
<div class="pub-links">
<a href="https://arxiv.org/abs/2505.18086">[PDF]</a>
</div>
</div>
<div class="pub-item">
<span class="venue-tag">ACL'24</span>
<span class="pub-title">Multimodal Table Understanding</span>
<div class="pub-authors">Mingyu Zheng*, Xinwei Feng*, <u>Qingyi Si*</u>, Qiaoqiao She, Zheng Lin, Wenbin Jiang, Weiping Wang</div>
<div class="pub-venue">ACL 2024</div>
<div class="pub-links">
<a href="https://arxiv.org/abs/2406.08100">[PDF]</a>
<a href="https://github.com/SpursGoZmy/Table-LLaVA">[Code]</a>
</div>
</div>
<div class="pub-item">
<span class="venue-tag">EMNLP'23</span>
<span class="pub-title">An Empirical Study of Instruction-tuning Large Language Models in Chinese</span>
<div class="pub-authors"><u>Qingyi Si*</u>, Tong Wang*, Zheng Lin, Xu Zhang, Yanan Cao, Weiping Wang</div>
<div class="pub-venue">EMNLP 2023</div>
<div class="pub-links">
<a href="https://arxiv.org/abs/2310.07328">[PDF]</a>
<a href="https://github.com/PhoebusSi/Alpaca-CoT">[Code]</a>
</div>
</div>
<div class="pub-item">
<span class="venue-tag">ACL'23</span>
<span class="pub-title">Combo of Thinking and Observing for Outside-Knowledge VQA</span>
<div class="pub-authors"><u>Qingyi Si</u>, Yuchen Mo, Zheng Lin, Huishan Ji, Weiping Wang</div>
<div class="pub-venue">ACL 2023</div>
<div class="pub-links">
<a href="https://arxiv.org/abs/2305.06407">[PDF]</a>
<a href="https://github.com/PhoebusSi/Thinking-while-Observing">[Code]</a>
</div>
</div>
<h3>Long/Streaming Context and Sparse Attention</h3>
<div class="pub-item">
<span class="venue-tag">ICLR'26</span>
<span class="pub-title">LouisKV: Efficient KV Cache Retrieval for Long Input-Output Sequences</span>
<div class="pub-authors">Wenbo Wu*, <u>Qingyi Si*</u>, Xiurui Pan, Ye Wang, Jie Zhang</div>
<div class="pub-venue">ICLR 2026</div>
<div class="pub-links">
<a href="https://arxiv.org/abs/2510.11292">[PDF]</a>
</div>
</div>
<div class="pub-item">
<span class="venue-tag">AAAI'26</span>
<span class="pub-title">Sparse Attention across Multiple-context KV Cache</span>
<div class="pub-authors">Ziyi Cao*, <u>Qingyi Si*</u>, Jingbin Zhang, Bingquan Liu</div>
<div class="pub-venue">AAAI 2026</div>
<div class="pub-links">
<a href="https://arxiv.org/abs/2508.11661">[PDF]</a>
</div>
</div>
<div class="pub-item">
<span class="venue-tag">ACL'25</span>
<span class="pub-title">AdaReTaKe: Adaptive Redundancy Reduction to Perceive Longer for Video-language Understanding</span>
<div class="pub-authors">Xiao Wang*, <u>Qingyi Si*</u> Shiyu Zhu, Jianlong Wu, Li Cao, Liqiang Nie</div>
<div class="pub-venue">ACL 2025</div>
<div class="pub-links">
<a href="https://arxiv.org/abs/2503.12559">[PDF]</a>
<a href="https://github.com/SCZwangxiao/video-FlexReduc">[Code]</a>
</div>
</div>
</section>
<section id="projects">
<h2>Open Source Projects</h2>
<ul>
<li>
<strong>Unified Cache Manager (UCM)</strong> - UCM addresses the challenges of low efficiency and high costs in long-context inference through a multi-tier KV Cache and inference memory management system. <a href="https://mp.weixin.qq.com/s/rCMjm-DMXJmOd2KjTv1QYw" style="color: var(--link-color);">[huawei]</a>
</li>
<li>
<strong>Alpaca-CoT</strong> (2.8k+ stars) - A unified instruction-tuning platform for LLMs with chain-of-thought data integration.
Used by 60+ models including Qwen. <a href="https://github.com/PhoebusSi/Alpaca-CoT" style="color: var(--link-color);">[GitHub]</a>
</li>
<li>
<strong>DEER</strong> - Training-free efficient reasoning algorithms for early exit or intervention in reasoning models. <a href=https://github.com/iie-ycx/DEER" style="color: var(--link-color);">[GitHub]</a>
</li>
<li>
<strong>AdaReTaKe</strong> - Sparse attention for long video understanding, achieving 8x length extension.
Ranked #1 on VideoMME, MLVU, LongVideoBench, LVBench leaderboards. <a href=https://github.com/SCZwangxiao/video-FlexReduc" style="color: var(--link-color);">[GitHub]</a>
</li>
</li>
</ul>
</section>
<section id="services">
<h2>Academic Services</h2>
<ul>
<li><strong>Area Chair:</strong> ACL, EMNLP</li>
<li><strong>Reviewer:</strong> NeurIPS, ICLR, ICML, CVPR, ICCV, AAAI, MM</li>
</ul>
</section>
<section>
<h2>Honors & Awards</h2>
<ul>
<li><strong>Tech Genius Team</strong>, JD Explore Academy(2025.11)</li>
<li><strong>Major Contribution Special Award</strong>, Huawei ICT BG (2025.6)</li>
<li><strong>Chinese Academy of Sciences President Award</strong> (top 1%, 2024.7)</li>
<li><strong>Outstanding Graduate</strong>, University of Chinese Academy of Sciences (2024.6)</li>
<li><strong>Top Minds</strong>, Huawei (2024.3)</li>
<li><strong>Doctoral National Scholarship</strong>, Chinese Academy of Sciences (top 2%, 2021)</li>
</ul>
</section>
<footer>
<p>© 2025 Qingyi Si (佀庆一) | Last updated: January 2026</p>
</footer>
</body>
</html>