1<!DOCTYPE HTML> 2<html lang="en"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8"> 3 4 <title>Xubing Ye</title> 5 6 <meta name="author" content="Xubing Ye"> 7 <meta name="viewport" content="width=device-width, initial-scale=1"> 8 9 <link rel="stylesheet" type="text/css" href="stylesheet.css"> 10 <link rel="icon" type="image/png" href="images/icon.png"> 11</head> 12 13<body> 14 <table style="width:100%;max-width:850px;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;margin-top:20px"><tbody> 15 <tr style="padding:0px"> 16 <td style="padding:0px"> 17 <table style="width:100%;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;"><tbody> 18 <tr style="padding:0px"> 19 <td style="padding:2.5%;width:63%;vertical-align:middle"> 20 <p style="text-align:center"> 21 <name>Xubing Ye (å¶æ ©å°)</name> 22 </p> 23 <p>Hi! I'm Xubing Ye. I'm currently an algorithm engineer at ByteDance Seed. Before that, I obtained my master's degree from <a href="https://www.sigs.tsinghua.edu.cn/">SIGS</a> at <a href="https://www.tsinghua.edu.cn/">Tsinghua University</a> in 2026, supervised by Prof. 24 <a href="https://andytang15.github.io/">Yansong Tang</a>. I obtained my bachelor's degree from <a href="https://www.tongji.edu.cn/">Tongji University</a> in 2023.</p> 25 <!-- I collaborated with Dr. <a href="https://scholar.google.com/citations?hl=zh-CN&user=8rltp9AAAAAJ">Yukang Gan</a>, Dr. <a href="https://geyixiao.com/">Yixiao Ge</a>, Dr. <a href="https://scholar.google.com/citations?user=4oXBp9UAAAAJ&hl=zh-CN&oi=ao">Ying Shan</a> and Dr. <a href="https://scholar.google.com/citations?hl=zh-CN&user=mBZewKcAAAAJ">Zhao Yang</a>. --> 26 <p>My current research focuses on data-centric engineering for agentic LLMs and MLLMs.</p> 27 28 <p style="text-align:center"> 29 <a href="mailto:[email protected]">Email: [email protected]</a>  /  30 <a href="https://github.com/Yxxxb"> Github </a>  /  31 <a href="https://scholar.google.com/citations?user=EERPDDEAAAAJ&hl"> Scholar </a> 32 </p> 33 </td> 34 <td style="padding:2.5%;width:30%;max-width:30%"> 35 <img style="width:80%;max-width:80%" alt="profile photo" src="images/xubing.jpg"> 36 </td> 37 </tr> 38 </tbody></table> 39 <table style="width:100%;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;margin-bottom:-22px;"><tbody> 40 <tr> 41 <td style="padding:20px;width:100%;vertical-align:middle"> 42 <heading>Recent Works</heading> 43 <p>* Indicates Equal Contribution</p> 44 </td> 45 </tr> 46 </tbody></table> 47 <table style="width:100%;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;margin-bottom:26px;"><tbody> 48 49 50 51 <tr> 52 <td style="padding:20px;width:30%;max-width:30%" align="center"> 53 <img style="width:100%;max-width:100%" src="images/seedrealtime.png" alt="SeedRealtime"> 54 </td> 55 <td width="75%" valign="center"> 56 <papertitle>SeedRealtime: An Audio-Visual Full-Duplex LLM</papertitle> 57 <br> 58 <span class="paper-authors">ByteDance Seed Team</span> 59 <br> 60 <em>Tech Blog</em>, 2026 61 <br> 62 <a href="https://seed.bytedance.com/en/SeedRealtime">[Tech Blog]</a> 63 <br> 64 <p>A native audio-visual full-duplex LLM enabling real-time interactions that can watch, listen, and speak.</p> 65 </td> 66 </tr> 67 68 <tr> 69 <td style="padding:20px;width:30%;max-width:30%" align="center"> 70 <img style="width:100%;max-width:100%" src="images/miro.png" alt="MiroThinker benchmark comparison"> 71 </td> 72 <td width="75%" valign="center"> 73 <papertitle>MiroThinker: Pushing the Performance Boundaries of Open-Source Research Agents via Model, Context, and Interactive Scaling</papertitle> 74 <br>
75 <span class="paper-authors">MiroMind Team</span> 76 <br> 77 <em>Tech Report</em>, 2026 78 <br> 79 <a href="https://arxiv.org/abs/2511.11793">[arXiv]</a> 80 <a href="https://github.com/MiroMindAI/MiroThinker" data-github-stars="MiroMindAI/MiroThinker">[Code]</a> 81 <a href="https://huggingface.co/models?search=miromind-ai%2FMiroThinker" data-hf-author="miromind-ai" data-hf-search="MiroThinker">[HF]</a> 82 <br> 83 <p>An open-source deep research agent advancing model, context, and interactive scaling.</p> 84 </td> 85 </tr> 86 87 <tr> 88 <td style="padding:20px;width:30%;max-width:30%" align="center"> 89 <img style="width:100%;max-width:100%" src="images/voco.png" alt="dise"> 90 </td> 91 <td width="75%" valign="center"> 92 <papertitle>VoCo-LLaMA: Towards Vision Compression with Large Language Models</papertitle> 93 <br> 94 <span class="paper-authors"> 95 <strong>Xubing Ye</strong>, 96 <a href="https://scholar.google.com/citations?hl=zh-CN&user=8rltp9AAAAAJ">Yukang Gan</a>, 97 <a href="https://xk-huang.github.io/">Xiaoke Huang</a>, 98 <a href="https://geyixiao.com/">Yixiao Ge</a>, 99 <a href="https://andytang15.github.io/">Yansong Tang</a> 100 </span> 101 <br> 102 <em>IEEE/CVF Conference on Computer Vision and Pattern Recognition (<strong>CVPR</strong>)</em>, 2025 103 <br> 104 <a href="https://arxiv.org/abs/2406.12275v2">[arXiv]</a> 105 <a href="https://github.com/Yxxxb/VoCo-LLaMA" data-github-stars="Yxxxb/VoCo-LLaMA">[Code]</a> 106 <br> 107 <p>Compresses hundreds of video tokens into one via attention distillation for efficient long-video understanding.</p> 108 </td> 109 </tr> 110 111 <tr> 112 <td style="padding:20px;width:30%;max-width:30%" align="center"> 113 <img style="width:100%;max-width:100%" src="images/atp.png" alt="dise"> 114 </td> 115 <td width="75%" valign="center"> 116 <papertitle>ATP-LLaVA: Adaptive Token Pruning for Large Vision Language Models</papertitle> 117 <br> 118 <span class="paper-authors"> 119 <strong>Xubing Ye</strong>, 120 <a href="https://scholar.google.com/citations?hl=zh-CN&user=8rltp9AAAAAJ">Yukang Gan</a>, 121 <a href="https://geyixiao.com/">Yixiao Ge</a>, 122 <a href="https://sites.google.com/view/xiaopingzhang/">Xiao-ping Zhang</a>, 123 <a href="https://andytang15.github.io/">Yansong Tang</a> 124 </span> 125 <br> 126 <em>IEEE/CVF Conference on Computer Vision and Pattern Recognition (<strong>CVPR</strong>)</em>, 2025 127 <br> 128 <a href="https://arxiv.org/abs/2412.00447">[arXiv]</a> 129 <br> 130 <p>Adaptively prunes visual tokens inside LLM decoder layers with minimal performance loss.</p> 131 </td> 132 </tr> 133 134 <!-- <tr> 135 <td style="padding:20px;width:30%;max-width:30%" align="center"> 136 <img style="width:100%;max-width:100%" src="images/points-reader.png" alt="POINTS-Reader training pipeline"> 137 </td> 138 <td width="75%" valign="center"> 139 <papertitle>POINTS-Reader: Distillation-Free Adaptation of Vision-Language Models for Document Conversion</papertitle> 140 <br> 141 <span class="paper-authors">WeLM Team</span> 142 <br> 143 <em>Conference on Empirical Methods in Natural Language Processing (<strong>EMNLP</strong>)</em>, 2025 144 <br> 145 <a href="https://arxiv.org/abs/2509.01215">[arXiv]</a> 146 <a href="https://github.com/Tencent/POINTS-Reader" data-github-stars="Tencent/POINTS-Reader">[Code]</a> 147 <a href="https://huggingface.co/tencent/POINTS-Reader" data-hf-model="tencent/POINTS-Reader">[HF]</a> 148 <br> 149 <p>A distillation-free document conversion model for extracting text, tables, and formulas.<
149/p> 150 </td> 151 </tr> --> 152 153 <tr> 154 <td style="padding:20px;width:30%;max-width:30%" align="center"> 155 <img style="width:100%;max-width:100%" src='images/lavt-pami.png' alt="dise"> 156 </td> 157 <td width="75%" valign="center"> 158 <papertitle>Language-Aware Vision Transformer for Referring Segmentation</papertitle> 159 <br> 160 <span class="paper-authors"> 161 <strong>Xubing Ye</strong>*, 162 <a href="https://scholar.google.com/citations?hl=zh-CN&user=mBZewKcAAAAJ">Zhao Yang</a>*, 163 <a href="https://myownskyw7.github.io/">Jiaqi Wang</a>*, 164 <a href="https://andytang15.github.io/">Yansong Tang</a>, 165 <a href="https://chenkai.site/">Kai Chen</a>, 166 <a href="https://hszhao.github.io/">Hengshuang Zhao</a>, 167 <a href="https://www.robots.ox.ac.uk/~phst/">Philip H.S. Torr</a> 168 </span> 169 <br> 170 <em>IEEE Transactions on Pattern Analysis and Machine Intelligence (<font color="red"><strong>TPAMI</strong></font>, IF=20.8)</em>, 2024 171 <br> 172 <em>IEEE/CVF Conference on Computer Vision and Pattern Recognition (<strong>CVPR</strong>)</em>, 2022 173 <br> 174 <a href="https://doi.org/10.1109/TPAMI.2024.3468640">[IEEE]</a> 175 <a href="https://arxiv.org/abs/2112.02244">[Conference Version]</a> 176 <a href="https://github.com/yz93/LAVT-RIS" data-github-stars="yz93/LAVT-RIS">[Code]</a> 177 <br> 178 <p>A universal referring image and video segmentation framework with language-aware visual encoding.</p> 179 </td> 180 </tr> 181 182 183 </tbody></table> 184 185 186 187 <table class="section-heading"><tbody> 188 <tr> 189 <td> 190 <heading>Selected Honors and Awards</heading> 191 </td> 192 </tr> 193 </tbody></table> 194 195 <table class="resume-list"><tbody> 196 <tr class="resume-item"> 197 <td> 198 <div class="resume-header"> 199 <span><strong>Nanhu Elite Scholarship</strong> · Tsinghua University · æ¸ å大å¦ç»¼åä¼ç§å¥å¦éï¼æ ¡çº§ä¸çï¼</span> 200 <span class="resume-date">2025</span> 201 </div> 202 </td> 203 </tr> 204 <tr class="resume-item"> 205 <td> 206 <div class="resume-header"> 207 <span><strong>Zhaoyi Scholarship</strong> · Tsinghua University · æ¸ å大å¦ç»¼åä¼ç§å¥å¦éï¼æ ¡çº§ä¸çï¼</span> 208 <span class="resume-date">2024</span> 209 </div> 210 </td> 211 </tr> 212 <tr class="resume-item"> 213 <td> 214 <div class="resume-header"> 215 <span><strong>First Prize Scholarship</strong> · Tongji University · åæµå¤§å¦ç»¼åä¼ç§å¥å¦éï¼æ ¡çº§ä¸çï¼</span> 216 <span class="resume-date">2023</span> 217 </div> 218 </td> 219 </tr> 220 </tbody></table> 221 222 223 224 225 <table class="section-heading"><tbody> 226 <tr> 227 <td> 228 <heading>Internship Experience</heading> 229 </td> 230 </tr> 231 </tbody></table> 232 233 234 <table class="resume-list"><tbody> 235 <tr class="resume-item"> 236 <td> 237 <div class="resume-header"> 238 <span class="resume-title">Shanda · MiroMind Team</span> 239 <span class="resume-date">2025 â 2026</span> 240 </div> 241 <div class="resume-details"> 242 Shanghai, China · Work with <a href="https://songbai.site/">Song Bai</a> and <a href="https://jifengdai.org/">Jifeng Dai</a>. 243 </div> 244 </td> 245 </tr> 246 <tr class="resume-item"> 247 <td> 248 <div class="resume-header"> 249 <span class="resume-title">Tencent · WXG · WeLM Team</span> 250 <span class="resume-date">2025</span> 251 </div> 252 <div class="resume-details"> 253 Shanghai, China · Work with <a href="https://yuanliuuuuuu.github.io/">Yuan Liu</a>, Le Tian, and Xiao Zhou. 254 </div> 255 </td> 256 </tr> 257 <tr class="resume-item"> 258 <td> 259 <div class="resume-header">
260 <span class="resume-title">ByteDance · SEED · Application</span> 261 <span class="resume-date">2024 â 2025</span> 262 </div> 263 <div class="resume-details"> 264 Beijing, China · Work with Baihan Shu. 265 </div> 266 </td> 267 </tr> 268 <tr class="resume-item"> 269 <td> 270 <div class="resume-header"> 271 <span class="resume-title">Tencent · PCG ARC Lab · Applied Research Center</span> 272 <span class="resume-date">2024</span> 273 </div> 274 <div class="resume-details"> 275 Shenzhen, China · Work with Yukang Gan, <a href="https://geyixiao.com/">Yixiao Ge</a>, and <a href="https://scholar.google.com/citations?user=4oXBp9UAAAAJ&hl=en">Ying Shan</a>. 276 </div> 277 </td> 278 </tr> 279 </tbody></table> 280 281 282 <table class="section-heading"><tbody> 283 <tr> 284 <td> 285 <heading>Academic Services</heading> 286 </td> 287 </tr> 288</tbody></table> 289 290<table class="resume-list services-list"><tbody> 291 <tr class="resume-item"> 292 <td> 293 <div class="resume-header"> 294 <span class="resume-title">Reviewer</span> 295 </div> 296 <div class="resume-details">TPAMI 2026 · TIP 2026 · CVPR 2025 · JVCIR 2024, 2025</div> 297 </td> 298 </tr> 299</tbody></table> 300 301 <div style="width: 30%; margin: 0 auto; overflow: hidden;margin-bottom:30px;"> 302 <!--
302<script type="text/javascript" id="clstr_globe" src="//clustrmaps.com/globe.js?d=FFhe8Qvtc-UmO9uKqymHR_GNpAF3U-1wHzxYpXsVK5E"></script>
302 --> 303 <!--
303<script type="text/javascript" id="clustrmaps" src="//clustrmaps.com/map_v2.js?d=FFhe8Qvtc-UmO9uKqymHR_GNpAF3U-1wHzxYpXsVK5E&cl=ffffff&w=a"></script>
303 --> 304
304<script type='text/javascript' id='clustrmaps' src='//cdn.clustrmaps.com/map_v2.js?cl=ffffff&w=a&t=tt&d=FFhe8Qvtc-UmO9uKqymHR_GNpAF3U-1wHzxYpXsVK5E'></script>
304 305 </div> 306 307 308 309 <p><center> 310 © Xubing Ye | Last updated: Mar. 17, 2024 | <a href="https://jonbarron.info/">Website Template</a> 311 </center></p> 312 313 <!-- <table style="width:100%;border:0px;border-spacing:0px;border-collapse:separate;margin-right:auto;margin-left:auto;"><tbody> 314 <tr> 315 <td style="padding:0px"> 316 <br> 317 <p style="text-align:right;font-size:small;"> 318 <a href="https://jonbarron.info/">Website Template</a> 319 </p> 320 </td> 321 </tr> 322 </tbody></table> 323<a href="https://clustrmaps.com/site/1bevc" title="Visit tracker"><img src="//www.clustrmaps.com/map_v2.png?d=pSqGxdyauRrVjsHq2YSTApeSzMqOewuzLgOt0g6Ow6Y&cl=ffffff" /></a> 324 325 </td> 326 </tr> 327 </table> --> 328 329
329<script> 330 const compactNumber = new Intl.NumberFormat('en', { 331 notation: 'compact', 332 maximumFractionDigits: 1 333 }); 334 335 document.querySelectorAll('[data-github-stars]').forEach(async (link) => { 336 try { 337 const response = await fetch(`https://api.github.com/repos/${link.dataset.githubStars}`); 338 if (!response.ok) return; 339 const repository = await response.json(); 340 link.textContent = `[Code: ${compactNumber.format(repository.stargazers_count)} Stars]`; 341 } catch (error) { 342 // Keep the plain Code link when the API is unavailable. 343 } 344 }); 345 346 document.querySelectorAll('[data-hf-model]').forEach(async (link) => { 347 try { 348 const model = link.dataset.hfModel; 349 const response = await fetch(`https://huggingface.co/api/models/${model}?expand[]=downloadsAllTime`); 350 if (!response.ok) return; 351 const data = await response.json(); 352 link.textContent = `[HF: ${compactNumber.format(data.downloadsAllTime)} Downloads]`; 353 } catch (error) { 354 // Keep the plain HF link when the API is unavailable. 355 } 356 }); 357 358 document.querySelectorAll('[data-hf-author]').forEach(async (link) => { 359 try { 360 const params = new URLSearchParams({ 361 author: link.dataset.hfAuthor, 362 search: link.dataset.hfSearch, 363 limit: '100' 364 }); 365 params.append('expand[]', 'downloadsAllTime'); 366 const response = await fetch(`https://huggingface.co/api/models?${params}`); 367 if (!response.ok) return; 368 const models = await response.json(); 369 const downloads = models.reduce((total, model) => total + (model.downloadsAllTime || 0), 0); 370 link.textContent = `[HF: ${compactNumber.format(downloads)} Downloads]`; 371 } catch (error) { 372 // Keep the plain HF link when the API is unavailable. 373 } 374 }); 375 376 </script>
376 377</body> 378 379</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.