1<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN" 2 "http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd"> 3<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en"> 4<head> 5<link rel="shortcut icon" href="myIcon.ico"> 6<meta http-equiv="Content-Type" content="text/html;charset=utf-8" /> 7 8<meta name="keywords" content="Xiaoyi Qu"> 9<meta name="description" content="Chaoyan Huang's home page"> 10<meta name="google-site-verification" content="yy_3iiS_X6pJdegdwitJMrH0LRLHXwpjrV9RKLXxKjg" /> 11<link rel="stylesheet" href="jemdoc.css" type="text/css"> 12<title>Xiaoyi Qu</title>
13<script>
vendor: 345 bytes, lines 13-19
13 14 (function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){ 15 (i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o), 16 m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m) 17 })(window,document,'script','https://www.google-analytics.com/analytics.js','ga'); 18 19 ga('create', '
19UA-87320911-1
vendor: 39 bytes, lines 19-21
19', 'auto'); 20 ga('send', 'pageview'); 21
22</script>
22 23</head> 24<body> 25 26 <!-- <div id="layout-content" style="margin-top:25px"> 27 <a href="https://github.com/CV-JunchengLi" class="github-corner"><svg width="80" height="80" viewBox="0 0 250 250" style="fill:#FD6C6C; color:#fff; position: absolute; top: 0; border: 0; right: 0;"><path d="M0,0 L115,115 L130,115 L142,142 L250,250 L250,0 Z"></path><path d="M128.3,109.0 C113.8,99.7 119.0,89.6 119.0,89.6 C122.0,82.7 120.5,78.6 120.5,78.6 C119.2,72.0 123.4,76.3 123.4,76.3 C127.3,80.9 125.5,87.3 125.5,87.3 C122.9,97.6 130.6,101.9 134.4,103.2" fill="currentColor" style="transform-origin: 130px 106px;" class="octo-arm"></path><path d="M115.0,115.0 C114.9,115.1 118.7,116.5 119.8,115.4 L133.7,101.6 C136.9,99.2 139.9,98.4 142.2,98.6 C133.8,88.0 127.5,74.4 143.8,58.0 C148.5,53.4 154.0,51.2 159.7,51.0 C160.3,49.4 163.2,43.6 171.4,40.1 C171.4,40.1 176.1,42.5 178.8,56.2 C183.1,58.6 187.2,61.8 190.9,65.4 C194.5,69.0 197.7,73.2 200.1,77.6 C213.8,80.2 216.3,84.9 216.3,84.9 C212.7,93.1 206.9,96.0 205.4,96.6 C205.1,102.4 203.0,107.8 198.3,112.5 C181.9,128.9 168.3,122.5 157.7,114.1 C157.9,116.9 156.7,120.9 152.7,124.9 L141.0,136.5 C139.8,137.7 141.6,141.9 141.8,141.8 Z" fill="currentColor" class="octo-body"></path></svg></a><style>.github-corner:hover .octo-arm{animation:octocat-wave 560ms ease-in-out}@keyframes octocat-wave{0%,100%{transform:rotate(0)}20%,60%{transform:rotate(-25deg)}40%,80%{transform:rotate(10deg)}}@media (max-width:500px){.github-corner:hover .octo-arm{animation:none}.github-corner .octo-arm{animation:octocat-wave 560ms ease-in-out}}</style> --> 28 29<table> 30 <tbody> 31 <tr> 32 <td width="670"> 33 <div id="toptitle"> 34 <h1>Xiaoyi Qu (æ²ç¬ä¸)</h1> 35 </div> 36 37 <p> 38 PhD candidate <br> 39 Department of Industrial and Systems Engineering,<br> 40 Lehigh University,<br> 41 200 W Packer Ave. Bethlehem, PA 18015.<br> 42 <br> 43 Email: [email protected] 44 </p> 45 46 <br> 47 <!-- <p> 48 <em>Success is not final, failure is not fatal, it is the courage to continue that counts.--Churchill</em> 49 </p> --> 50 51 <p> 52 <a href="https://scholar.google.com/citations?hl=en&user=zn_q63wAAAAJ", target="_blank"><img src="./img/semantics_scholar.png" height="30px" style="margin-bottom:-3px"></a> 53 <a href="https://github.com/Xiaoyi-Qu", target="_blank"><img src="./img/git.jpeg" height="30px" style="margin-bottom:-3px"></a> 54 <a href="https://www.linkedin.com/in/xiaoyi-qu-20590018b/", target="_blank"><img src="./img/linkedin.png" height="50px" style="margin-bottom:-13px"></a> 55 <!--<a href="index_zw.html", target="_blank"><img src="./pic/zw.png" height="30px" style="margin-bottom:-3px"></a>--> 56 </p><br> 57 </td> 58 <td> 59 <img src="./img/Profile_Xiaoyi.jpg" border="0" width="200"><br> 60 </td> 61 </tr> 62 <tr> 63 </tr> 64 </tbody> 65</table> 66 67<div id="page_end_html"> 68 <!--white cat--> 69
69<script src=" https://eqcn.ajz.miesnfu.com/wp-content/plugins/wp-3d-pony/live2dw/lib/L2Dwidget.min.js"></script>
69 70
70<script src="https://eqcn.ajz.miesnfu.com/wp-content/plugins/wp-3d-pony/live2dw/lib/L2Dwidget.0.min.js"></script>
70 71
71<script> 72 L2Dwidget.init({ 73 "model": { 74 jsonPath: " https://unpkg.com/[email protected]/assets/tororo.model.json" , 75 "scale": 1 76 }, 77 "display": { 78 "position": "right", 79 "width": 100, 80 "height": 150, 81 "hOffset": -20, 82 "vOffset": -20 83 }, 84 "mobile": { 85 "show": true, 86 "scale": 0.1 87 }, 88 "react": { 89 "opacityDefault": 0.8, 90 "opacityOnHover": 0.3 91 } 92 }); 93 </script>
93 94 95 96 </div> 97 98<h2>Biography</h2> 99<p style="text-align:justify; text-justify:inter-ideograph;"> 100 Xiaoyi Qu is a fourth-year Ph.D. candidate in the Department of Industrial and Systems Engineering at Lehigh University, 101 under the supervision of Prof. <a href="https://coral.ise.lehigh.edu/danielprobinson/" target="blank">Daniel P. Robinson</a>. 102 His research focuses on two primary areas: (i) the development and theoretical analysis of first-order methods 103 for solving constrained structured optimization problems, and (ii) model optimization techniques (e.g., structured pruning and quantization-aware 104 training) for deep neural networks. In 2026, he worked as a machine learning engineer intern at <a href="https://genies.com/" target="blank">Genies Inc</a>, 105 where he focused on inference acceleration and model optimization for an internal 3D generation pipeline. In 2024, he worked as a research intern with Microsoft's 106 Applied Sciences Group (<a href="https://www.microsoft.com/applied-sciences/?msockid=2aaecc043ad26f493eb7d9693b836e27" target="blank">ASG</a>), where 107 his work centered on joint pruning and quantization techniques. Prior to this, he was an acoustic test intern at 108 <a href="https://www.neusoftmedical.com/en/" target="blank">Neusoft Medical</a> in 2019. 109</p> 110 111 112<h2>Education</h2> 113<ul> 114<li><p>Ph.D., Industrial Engineering, Lehigh University, Bethlehem, PA, USA, 2022–current</p> 115</li> 116<li><p>B.S., Mathematical Science, University of Michigan, Ann Arbor, MI, USA, 2019–2022</p> 117</li> 118<li><p> Computational Mathematics, Dalian University of Technology, Dali
118an, China, 2017–2019</p> 119</li> 120</ul> 121 122<h2>Publication and Preprints </h2> 123 124<table id="tbPublications" width="100%"> 125 <tbody> 126 127<tr> 128 <td width="206"> 129 <img src="./img/multi_agent_medical.png" width="200" height = "120" style="box-shadow: 4px 4px 8px #888"> 130 </td> 131 <td> 132 <a href="./img/multi_agent_medical.png" target="_blank">Multi-Agent Consensus Matrix Modeling for Medical Decision-Making: A Role-Specialized LLM Framework for Oncology MDT Consultations 133 </a><br /> 134 Ziyi Ni, Yiming Yan, <b>Xiaoyi Qu</b>, Yanzhan Chen, Chuang Liu. <br /> 135 <i style="color: #555;">ICLR 2026 Workshop on MALGAI</i> <br /> 136 [<a href="https://openreview.net/pdf?id=em8vGhlUxU">Paper</a>] 137 </td> 138</tr> 139<tr></tr> 140<tr></tr> 141<tr></tr> 142 143<tr> 144 <td width="206"> 145 <img src="./img/PG_equality.png" width="200" height = "140" style="box-shadow: 4px 4px 8px #888"> 146 </td> 147 <td> <a href="./img/PG_equality.png" target="_blank"> A Proximal Gradient Method for Equality Constrained Optimization</a> <br /> 148 [<img class="eq" src="./img/7227645438722938594-130.png" alt="alpha" style="vertical-align: -1px" />] Yutong Dai, <b>Xiaoyi Qu</b>, Daniel P. Robinson. <br /> 149 <i style="color: #555;">SIAM Journal on Optimization (<i><b>SIOPT</b></i>), 2025</i> <br /> 150 <i style="color: Tomato;"> 2025 Van Hoesen Family Best Publication Award (Honorable Mention) </i> <br /> 151 [<a href="https://epubs.siam.org/doi/10.1137/24M1653586">Paper</a>] 152 [<a href="https://arxiv.org/abs/2404.07460">arXiv</a>] <br /> 153 </td> 154</tr> 155<tr></tr> 156<tr></tr> 157<tr></tr> 158 159<tr> 160 <td width="206"> 161 <img src="./img/geta.png" width="200" height = "120" style="box-shadow: 4px 4px 8px #888"> 162 </td> 163 <td> 164 <a href="./img/geta.png" target="_blank">Automated Joint Structured Pruning and Quantization for Efficient Neural Network Training and Compression</a><br /> 165 <b>Xiaoyi Qu</b>, David Aponte, Colby Banbury, Daniel P. Robinson, Tianyu Ding, Kazuhito Koishida, Ilya Zharkov, Tianyi Chen. <br /> 166 <i style="color: #555;">Computer Vision and Pattern Recognition (<i><b>CVPR</b></i>), 2025</i> <br /> 167 [<a href="https://cvpr.thecvf.com/virtual/2025/poster/34889">Paper</a>] 168 [<a href="https://arxiv.org/abs/2502.16638">arXiv</a>] 169 [<a href="https://github.com/microsoft/geta">Code</a>] <br /> 170 </td> 171</tr> 172<tr></tr> 173<tr></tr> 174<tr></tr> 175 176<tr> 177 <td width="206"> 178 <img src="./img/qat_flow-1.png" width="200" height = "120" style="box-shadow: 4px 4px 8px #888"> 179 </td> 180 <td> 181 <a href="./img/qat_flow-1.png" target="_blank">StableQAT: Stable Quantization-Aware Training at Ultra-Low Bitwidths</a><br /> 182 Tianyi Chen, Sihan Chen, <b>Xiaoyi Qu</b>, Dan Zhao, Ruomei Yan, Jongwoo Ko, Luming Liang, Pashmina Cameron. <br /> 183 <i>arXiv preprint, 2026</i> <br /> 184 [<a href="https://arxiv.org/abs/2601.19320">arXiv</a>] 185 [<a href="https://github.com/microsoft/StableQAT">Code</a>] <br /> 186 </td> 187</tr> 188<tr></tr> 189<tr></tr> 190<tr></tr> 191 192<tr> 193 <td width="206"> 194 <img src="./img/PG_general.png" width="200" height = "160" style="box-shadow: 4px 4px 8px #888"> 195</td> 196<td> <a href="./img/PG_general.png" target="_blank">A Proximal Gradient Method for Solving Regularized Optimization with General Constraints</a><br> 197 [<img class="eq" src="./img/7227645438722938594-130.png" alt="alpha" style="vertical-align: -1px" />] Frank E. Curtis, <b>Xiaoyi Qu </b>, Daniel P. Robinson. <br> 198 <i>arXiv preprint, 2025</i> <br /> 199 [<a href="https://arxiv.org/abs/2512.23166">arXiv</a>] 200 </td> 201 </tr> 202 <tr></tr> 203 <tr></tr> 204 <tr></tr> 205 206<tr> 207 <td width="206"> 208 <img src="./img/hesso_new.png" width="200" height = "120" style="box-shadow: 4px 4px 8px #888"> 209 </td> 210 <td> 211 <a href="./img/hesso_new.png" target="_blank">HESSO: Towards Automatic Efficient and User Friendly Any Neural Network Training and Pruning</a><br /> 212 <b>Xiaoyi Qu</b>, David Aponte, Colby Banbury, Jongwoo Ko, Tianyu Ding, Yong Ma, Vladimir Lyapunov, Ilya Zharkov, Luming Liang, Tianyi Chen. <br /> 213 <i>arXiv preprint, 2024</i> <br /> 214 [<a href="https://arxiv.org/abs/2409.09085">arXiv</a>] 215 [<a href="https://github.com/microsoft/only_train_once">
215Code</a>] <br /> 216 </td> 217</tr> 218<tr></tr> 219<tr></tr> 220<tr></tr> 221 222 223 224</tbody></table> 225 226<!-- <h2>Professional Service</h2> 227<ul> 228<li><p>Journal Reviewer: Optimization and Enigneering, Computational Optimization and Applications, IEEE Transactions on Circuits and Systems for Video Technology</p> 229</li> 230<li><p>Conference Reviewer: Constrained Optimization for Machine Learning (COML) Workshop, NeurIPS 2025; ICASSP 2026; Workshop on Multi-Agent Learning and Its Opportunities in the Era of Generative AI, ICLR 2026</p> 231</li> 232</ul> --> 233 234<h2>Professional Service</h2> 235 236<p><strong>Journal Reviewer:</strong></p> 237<ul> 238 <li>Optimization and Engineering</li> 239 <li>Computational Optimization and Applications</li> 240 <li>IEEE Transactions on Circuits and Systems for Video Technology</li> 241 <li>INFORMS Journal on Optimization</li> 242 <li>Machine Learning</li> 243</ul> 244 245<p><strong>Conference Reviewer:</strong></p> 246<ul> 247 <li>Constrained Optimization for Machine Learning (COML) Workshop, NeurIPS 2025</li> 248 <li>ICASSP 2026</li> 249 <li>Workshop on Multi-Agent Learning and Its Opportunities in the Era of Generative AI, ICLR 2026</li> 250</ul> 251 252<h2>Notes</h2> 253<ul> 254<li><p>Inference System Part I: <a href="continuous_batching.html">Continuous Batching (è¿ç»æ¹å¤ç)</a></p> 255<li><p>Inference System Part II: <a href="inference_scheduling.html">Inference Scheduling (æ¨çè°åº¦)</a></p> 256<li><p>Inference System Part III: <a href="pd_disaggregation.html">PD Disaggregation (PDå离)</a></p> 257<li><p>Inference System Part IV: <a href="inference_parallelism.html">Inference Parallelism (æ¨ç侧并è¡)</a></p> 258<li><p>Inference System Part V: <a href="profile_runtime.html">Profiling and Runtime Optimization (ProfilingåRuntimeä¼å)</a></p> 259<li><p>Inference System Part VI: <a href="speculative_decoding.html">Speculative Decoding (ææºè§£ç )</a></p> 260<li><p>Quantization Part I: <a href="dequantization_w4a16.html">Dequantization Inference-W4A16 (åéåæ¨ç)</a></p> 261<li><p>Long context : Long Context Extrapolation (é¿ä¸ä¸æå¤æ¨)</p> 262</li> 263</ul> 264 265 266<h2>Non-academic Stuff</h2> 267<ul> 268<li><p>I am originally from northeastern China.</p> 269<li><p>I once drove cross-country and back, making the full round trip from the East Coast to the West Coast.</p> 270<!-- <li><p>Since childhood, I have dreamed of becoming a geologist, partly for the chance to travel across the nation for free.</p> --> 271</li> 272</ul> 273 274<!-- <h2>Useful Links</h2> 275<ul> 276<li><p><a href="https://www.educatum.com/engineering-blogs-in-ai-ml-system-design">System Design in Tech Industry</a></p> 277</li> 278</ul> --> 279 280 281<div id="footer"> 282 <div id="footer-text"></div> 283</div> 284 285<p> 286<center> 287<div> 288 289<br> 290© Xiaoyi | Last updated: 2026-4 291</center> 292</p> 293 294</div> 295</body></html>
Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.