1<!DOCTYPE html> 2<html lang="en-US"> 3 <head> 4 <title>Fei Wang</title> 5 6 <meta charset="UTF-8"> 7 <meta http-equiv="X-UA-Compatible" content="IE=edge"> 8 <meta name="viewport" content="width=device-width, initial-scale=1"> 9 <meta name="description" content=""> 10 11 12 13 <link rel="icon" media="(prefers-color-scheme:dark)" href="" type="image/png" /> 14 <link rel="icon" media="(prefers-color-scheme:light)" href="" type="image/png" /> 15
15<script src="./assets/js/favicon-switcher.js" type="application/javascript"></script>
15 16 17 <link rel=stylesheet href=https://cdnjs.cloudflare.com/ajax/libs/academicons/1.8.6/css/academicons.min.css integrity="sha256-uFVgMKfistnJAfoCUQigIl+JfUaP47GrRKjf6CTPVmw=" crossorigin=anonymous> 18 <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.4.2/css/all.min.css" integrity="sha512-z3gLpd7yknf1YoNbCzqRKc4qyor8gaKU1qmn+CShxbuBusANI9QpRohGBreCFkKxLhei6S9CQXFEbbKuqLg0DA==" crossorigin="anonymous" referrerpolicy="no-referrer" /> 19 20 21 <link rel="stylesheet" href="./assets/css/font_sans_serif.css"> 22 23 24 <link rel="stylesheet" href="./assets/css/style.css"> 25 <link rel="stylesheet" href="./assets/css/publications.css"> 26 27 28 </head> 29 <body> 30 <div class="wrapper"> 31 <header> 32 33 34 <a class="image avatar"><img src="assets/img/profile.jpg" alt="avatar" /></a> 35 36 37 <h1>Fei Wang</h1> 38 39 40 41 42 <email>feiwangnlp (at) google [dot] com</email> 43 44 45 <br> 46 <br> 47 <div class="social-icons"> 48 49 <a style="margin: 0 5px 0 0" href="https://scholar.google.com/citations?user=N1O2KT8AAAAJ"> 50 <i class="ai ai-google-scholar" style="font-size:1.2rem"></i> 51 </a> 52 53 54 55 <a style="margin: 0 5px 0 0" href="https://github.com/FeiWang96"> 56 <i class="fab fa-github"></i> 57 </a> 58 59 60 61 <a style="margin: 0 5px 0 0" href="https://www.linkedin.com/in/fei-wang-99b2341a0/"> 62 <i class="fab fa-linkedin"></i> 63 </a> 64 65 66 67 <a style="margin: 0 0 0 0" href="https://x.com/fwang_nlp"> 68 <i class="fab fa-x-twitter"></i> 69 </a> 70 71 </div> 72 <br> 73 74 </header> 75 <section> 76 77 <h2 id="about-me">About Me</h2> 78 79<p>I am a Research Scientist at Google. Previously, I received my Ph.D. in Computer Science from the University of Southern California, co-advised by <a href="https://muhaochen.github.io/">Prof. Muhao Chen</a> and <a href="https://scholar.google.com/citations?user=rJTwW0MAAAAJ">Prof. Aram Galstyan</a>. I also worked closely with <a href="https://web.cs.ucla.edu/~kwchang/">Prof. Kai-Wei Chang</a> and <a href="https://www.cis.upenn.edu/~danroth/">Prof. Dan Roth</a>.</p> 80 81<p>My research focuses on developing post-training methods to make LLMs and multimodal LLMs more reliable and robust.</p> 82 83<p><br /></p> 84 85<h2 id="highlights">Highlights</h2> 86 87<ul> 88 <li><strong>EMNLP 2024 Tutorial</strong>: <a href="https://www.wenpengyin.org/publications/beyond-llm-scaling-emnlp24">Enhancing LLM Capabilities 89Beyond Scaling Up</a></li> 90 <li><strong>Amazon ML Fellowship</strong> (<a href="https://trustedai.usc.edu/20222023-amazon-ml-fellows-1">2023-2024</a>)</li> 91 <li>Annenberg Fellowship (2022-2023)</li> 92</ul> 93 94<p><br /></p> 95 96<h2 id="publications" style="margin: 2px 0px -15px;">Selected Publications</h2> 97 98<div class="publications"> 99<ol class="bibliography"> 100 101<div> 102<p> Please refer to my <a href="https://scholar.google.com/citations?user=N1O2KT8AAAAJ">Google Scholar profile</a> for the full list. </p> 103</div> 104 105 106 107<li> 108<div class="pub-row"> 109 <!-- <div class="col-sm-3 abbr" style="position: relative;padding-right: 15px;padding-left: 15px;"> 110 111 </div> --> 112 <div class="col-sm-9" style="position: relative;padding-right: 5px;padding-left: 5px;"> 113 <div class="title"><a href="https://arxiv.org/abs/2406.09411">MuirBench: A Comprehensive Benchmark for Robust Multi-image Understanding</a></div> 114 <div class="author"><u>Fei Wang</u>, Xingyu Fu*, James Y. Huang, Zekun Li, Qin Liu, Xiaogeng Liu, Mingyu Derek Ma, Nan Xu, Wenxuan Zhou, and others</div> 115 <div class="periodical"><i class="fa-regular fa-bookmark"></i> <em>ICLR 2025</em> 116 </div> 117 <div class="links"> 118 119 <a href="https://arxiv.org/abs/2406.09411" class="btn btn-sm z-depth-0" role="button" target="_blank" style="font-size:12px;">Paper</a> 120 121 122 <a href="https://muirbench.github.io/" class="btn btn-sm z-depth-0" role="button" target="_blank" style="font-size:12px;">Website</a> 123 124 125 126 <a href="https://huggingface.co/datasets/MUIRBENCH/MUIRBENCH" class="btn btn-sm z-depth-0" role="button" target="_blank" style="font-size:12px;">Dataset</a> 127 128 129 <i style="color:#e74d3c;" class="fa-solid fa-award"></i><strong><i style="color:#e74d3c;font-size:12px;"> 20k+ Downloads </i></strong> 130 131 132 </div> 133 </div> 134</div> 135</li> 136 137<br /> 138 139 140 141<li> 142<div class="pub-row"> 143 <!-- <div class="col-sm-3 abbr" style="position: relative;padding-right: 15px;padding-left: 15px;"> 144 145 </div> --> 146 <div class="col-sm-9" style="position: relative;padding-right: 5px;padding-left: 5px;"> 147 <div class="title"><a href="https://arxiv.org/abs/2410.07176">Astute RAG: Overcoming Imperfect Retrieval Augmentation and Knowledge Conflicts for Large Language Models</a></div> 148 <div class="author"><u>Fei Wang</u>, Xingchen Wan, Ruoxi Sun, Jiefeng Chen, and Sercan Ã. Arık</div> 149 <div class="periodical"><i class="fa-regular fa-bookmark"></i> <em>ACL 2025</em> 150 </div> 151 <div class="links"> 152 153 <a href="https://arxiv.org/abs/2410.07176" class="btn btn-sm z-depth-0" role="button" target="_blank" style="font-size:12px;">Paper</a> 154 155 156 157 158 159 160 </div> 161 </div> 162</div> 163</li> 164 165<br /> 166 167 168 169<li> 170<div class="pub-row"> 171 <!-- <div class="col-sm-3 abbr" style="position: relative;padding-right: 15px;padding-left: 15px;"> 172 173 </div> --> 174 <div class="col-sm-9" style="position: relative;padding-right: 5px;padding-left: 5px;"> 175 <div class="title"><a href="https://arxiv.org/abs/2403.06326">From Instructions to Constraints: Language Model Alignment with Automatic Constraint Verification</a></div> 176 <div class="author"><u>Fei Wang</u>
176, Chao Shang, Sarthak Jain, Shuai Wang, Qiang Ning, Bonan Min, Vittorio Castelli, Yassine Benajiba, and Dan Roth</div> 177 <div class="periodical"><i class="fa-regular fa-bookmark"></i> <em>Findings of NAACL 2025</em> 178 </div> 179 <div class="links"> 180 181 <a href="https://arxiv.org/abs/2403.06326" class="btn btn-sm z-depth-0" role="button" target="_blank" style="font-size:12px;">Paper</a> 182 183 184 185 186 187 188 </div> 189 </div> 190</div> 191</li> 192 193<br /> 194 195 196 197<li> 198<div class="pub-row"> 199 <!-- <div class="col-sm-3 abbr" style="position: relative;padding-right: 15px;padding-left: 15px;"> 200 201 </div> --> 202 <div class="col-sm-9" style="position: relative;padding-right: 5px;padding-left: 5px;"> 203 <div class="title"><a href="https://arxiv.org/abs/2406.11839">mDPO: Conditional Preference Optimization for Multimodal Large Language Models</a></div> 204 <div class="author"><u>Fei Wang</u>, Wenxuan Zhou, James Y. Huang, Nan Xu, Sheng Zhang, Hoifung Poon, and Muhao Chen</div> 205 <div class="periodical"><i class="fa-regular fa-bookmark"></i> <em>EMNLP 2024</em> 206 </div> 207 <div class="links"> 208 209 <a href="https://arxiv.org/abs/2406.11839" class="btn btn-sm z-depth-0" role="button" target="_blank" style="font-size:12px;">Paper</a> 210 211 212 <a href="https://feiwang96.github.io/mDPO" class="btn btn-sm z-depth-0" role="button" target="_blank" style="font-size:12px;">Website</a> 213 214 215 <a href="https://github.com/luka-group/mDPO" class="btn btn-sm z-depth-0" role="button" target="_blank" style="font-size:12px;">
215Code</a> 216 217 218 219 220 </div> 221 </div> 222</div> 223</li> 224 225<br /> 226 227 228 229<li> 230<div class="pub-row"> 231 <!-- <div class="col-sm-3 abbr" style="position: relative;padding-right: 15px;padding-left: 15px;"> 232 233 </div> --> 234 <div class="col-sm-9" style="position: relative;padding-right: 5px;padding-left: 5px;"> 235 <div class="title"><a href="https://arxiv.org/abs/2305.14710">Instructions as Backdoors: Backdoor Vulnerabilities of Instruction Tuning for Large Language Models</a></div> 236 <div class="author">Jiashu Xu, Mingyu Derek Ma, <u>Fei Wang</u>, Chaowei Xiao, and Muhao Chen</div> 237 <div class="periodical"><i class="fa-regular fa-bookmark"></i> <em>NAACL 2024</em> 238 </div> 239 <div class="links"> 240 241 <a href="https://arxiv.org/abs/2305.14710" class="btn btn-sm z-depth-0" role="button" target="_blank" style="font-size:12px;">Paper</a> 242 243 244 245 246 247 248 </div> 249 </div> 250</div> 251</li> 252 253<br /> 254 255 256 257<li> 258<div class="pub-row"> 259 <!-- <div class="col-sm-3 abbr" style="position: relative;padding-right: 15px;padding-left: 15px;"> 260 261 </div> --> 262 <div class="col-sm-9" style="position: relative;padding-right: 5px;padding-left: 5px;"> 263 <div class="title"><a href="https://arxiv.org/abs/2205.03972">Robust (Controlled) Table-to-Text Generation with Structure-Aware Equivariance Learning</a></div> 264 <div class="author"><u>Fei Wang</u>, Zhewei Xu, Pedro Szekely, and Muhao Chen</div> 265 <div class="periodical"><i class="fa-regular fa-bookmark"></i> <em>NAACL 2022</em> 266 </div> 267 <div class="links"> 268 269 <a href="https://arxiv.org/abs/2205.03972" class="btn btn-sm z-depth-0" role="button" target="_blank" style="font-size:12px;">Paper</a> 270 271 272 273 <a href="https://github.com/luka-group/Lattice" class="btn btn-sm z-depth-0" role="button" target="_blank" style="font-size:12px;">
273Code</a> 274 275 276 277 278 </div> 279 </div> 280</div> 281</li> 282 283<br /> 284 285 286 287</ol> 288</div> 289 290<p><br /></p> 291 292<h2 id="services">Services</h2> 293 294<ul> 295 <li>Area Chair for NeurIPS (2025), ACL (2024-), EMNLP (2024-), NAACL (2024-)</li> 296 <li>Action Editor for ACL Rolling Review (2024-)</li> 297 <li>Reviewer for ICLR (2022, 2024, 2025), ICML (2023â2025), NeurIPS (2022â2025), ACL (2021, 2023, 2025), EMNLP (2021â2023), NAACL (2021), COLM (2024), ICCV (2025), AAAI (2023â2025), EACL (2023), ACL Rolling Review (2021â2024), TPAMI (2025), JAIR (2023, 2024), TMLR (2024), TIST (2024), TOIS (2022)</li> 298</ul> 299 300 301 302 <br> 303 304 305 <p><small>Powered by Jekyll and <a href="https://github.com/yaoyao-liu/minimal-light" target="_blank" rel="noopener">Minimal Light</a> theme.</small></p> 306 307 308 </section> 309 <footer> 310 311 </footer> 312 </div> 313
313<script src="/assets/js/scale.fix.js"></script>
313 314 315 </body> 316</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.