PageSourceSearch

https://ken77921.github.io/

html ken77921.github.io collected 2026-10-03 09:27:09 UTC 34,290 bytes, 57 lines download raw bytes

1<!DOCTYPE html> <html lang="en"> <head> <meta http-equiv="Content-Type" content="text/html; charset=UTF-8"> <meta charset="utf-8"> <meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no"> <meta http-equiv="X-UA-Compatible" content="IE=edge"> <title>Haw-Shiuan Chang</title> <meta name="author" content="Haw-Shiuan Chang"> <meta name="description" content="A simple, whitespace theme for academics. Based on [*folio](https://github.com/bogoli/-folio) design. "> <link rel="stylesheet" href="/assets/css/bootstrap.min.css?a4b3f509e79c54a512b890d73235ef04"> <link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/css/mdb.min.css" integrity="sha256-jpjYvU3G3N6nrrBwXJoVEYI/0zw8htfFnhT9ljN3JJw=" crossorigin="anonymous"> <link defer rel="stylesheet" href="https://unpkg.com/[email protected]/dist/bootstrap-table.min.css"> <link rel="stylesheet" href="/assets/css/academicons.min.css?f0b7046b84e425c55f3463ac249818f5"> <link rel="stylesheet" type="text/css" href="https://fonts.googleapis.com/css?family=Roboto:300,400,500,700|Roboto+Slab:100,300,400,500,700|Material+Icons"> <link rel="stylesheet" href="/assets/css/jekyll-pygments-themes-github.css?19f3075a2d19613090fe9e16b564e1fe" media="" id="highlight_theme_light"> <link rel="shortcut icon" href="data:image/svg+xml,&lt;svg%20xmlns=%22http://www.w3.org/2000/svg%22%20viewBox=%220%200%20100%20100%22&gt;&lt;text%20y=%22.9em%22%20font-size=%2290%22&gt;%F0%9F%8C%8E&lt;/text&gt;&lt;/svg&gt;"> <link rel="stylesheet" href="/assets/css/main.css?d41d8cd98f00b204e9800998ecf8427e"> <link rel="canonical" href="https://ken77921.github.io/"> <link rel="stylesheet" href="/assets/css/jekyll-pygments-themes-native.css?e74e74bf055e5729d44a7d031a5ca6a5" media="none" id="highlight_theme_dark"> 
1<script src="/assets/js/theme.js?6185d15ea1982787ad7f435576553d64"></script>
1 </head> <body class="fixed-top-nav "> <header> <nav id="navbar" class="navbar navbar-light navbar-expand-sm fixed-top"> <div class="container"> <button class="navbar-toggler collapsed ml-auto" type="button" data-toggle="collapse" data-target="#navbarNav" aria-controls="navbarNav" aria-expanded="false" aria-label="Toggle navigation"> <span class="sr-only">Toggle navigation</span> <span class="icon-bar top-bar"></span> <span class="icon-bar middle-bar"></span> <span class="icon-bar bottom-bar"></span> </button> <div class="collapse navbar-collapse text-right" id="navbarNav"> <ul class="navbar-nav ml-auto flex-nowrap"> <li class="nav-item active"> <a class="nav-link" href="/">About<span class="sr-only">(current)</span></a> </li> <li class="nav-item dropdown "> <a class="nav-link dropdown-toggle" href="#" id="navbarDropdown" role="button" data-toggle="dropdown" aria-haspopup="true" aria-expanded="false">Projects</a> <div class="dropdown-menu dropdown-menu-right" aria-labelledby="navbarDropdown"> <a class="dropdown-item " href="/projects/#header-entailment_hierarchy">Entailment Hierarchy</a> <a class="dropdown-item " href="/projects/#header-curve-extrapolation-LM">Extrapolating the Distributions of an Infinitely Large Language Model</a> <a class="dropdown-item " href="/projects/#header-multifacet-embedding-LM">Multi-facet Embeddings for Language Modeling</a> <a class="dropdown-item " href="/projects/#header-active-learning">Active Learning and Crowdsourcing</a> <a class="dropdown-item " href="/projects/#header-NLP">Natural Language Processing</a> <a class="dropdown-item " href="/projects/#header-CV">Computer Vision (Clustering and Matching)</a> </div> </li> <li class="nav-item "> <a class="nav-link" href="/publications/">Publications</a> </li> <li class="nav-item "> <a class="nav-link" href="/teaching/">Teaching</a> </li> <li class="nav-item "> <a class="nav-link" href="/software/">Software</a> </li> <li class="toggle-container"> <button id="light-toggle" title="Change theme"> <i class="fa-solid fa-moon"></i> <i class="fa-solid fa-sun"></i> </button> </li> </ul> </div> </div> </nav> <progress id="progress" value="0"> <div class="progress-container"> <span class="progress-bar"></span> </div> </progress> <div class="container mt-5"> <div class="post"> <header class="post-header"> <h1 class="post-title"> Haw-Shiuan Chang </h1> <p class="desc">張浩軒</p> </header> <article> <div class="profile float-right"> <figure> <picture> <source class="responsive-img-srcset" srcset=" /assets/img/profile-hschang-6-480.webp 480w, /assets/img/profile-hschang-6-800.webp 800w, /assets/img/profile-hschang-6-1400.webp 1400w, " sizes="(min-width: 1000px) 291.0px, (min-width: 576px) 30vw, 95vw" type="image/webp"></source> <img src="/assets/img/profile-hschang-6.jpeg?f19d0796ba662a760cd9cfe15981e5eb" class="img-fluid z-depth-1 rounded-circle" width="100%" height="auto" alt="profile-hschang-6.jpeg" onerror="this.onerror=null; $('.responsive-img-srcset').remove();"> </picture> </figure> </div> <div class="clearfix"> <p>I am an Assistant Professor in the College of Information Science at the University of Arizona. My research bridges machine learning (ML), natural language processing (NLP), and information retrieval (IR), focusing on <b>fundamentally narrowing the gap between large language models (LLMs) and human intelligence</b> without relying on scaling laws such as increasing model size or training data size. Toward that end, I am particularly interested in <br> (1) enhancing the generation factuality, diversity, and novelty of LLMs by encouraging the usage of more causal inferences and fewer guesses, <br> (2) discovering LLMs’ limitations in their current architectures, decoding algorithms, evaluation, and training data, and <br> (3) overcoming the limitations through the techniques inspired by human behavior.</p> <p>Previously, I was a postdoctoral research associate at UMass Amherst Center for Intelligent Information Retrieval (CIIR), advised by Professor <a href="https://groups.cs.umass.edu/zamani/" rel="external nofollow noopener" target="_blank">Hamed Zamani</a>, and a postdoctoral scientist at Amazon AGI Foundations and worked with Professor <a href="https://vnpeng.net/" rel="external nofollow noopener" target="_blank">Violet Peng</a>, Professor <a href="https://www.cs.unc.edu/~mbansal/" rel="external nofollow noopener" target="_blank">Mohit Bansal</a>, and Dr. Tagyoung Chung. I got my PhD at the University of Massachusetts Amherst, advised by Professor <a href="https://people.cs.umass.edu/~mccallum/" rel="external nofollow noopener" target="_blank">Andrew McCallum</a>. Prior to PhD, I worked with Professor <a href="http://vllab.ee.ntu.edu.tw/members.html" rel="external nofollow noopener" target="_blank">Yu-Chiang Frank Wang</a> and Dr. Kuan-Ta Chen at <a href="https://www.sinica.edu.tw/en" rel="external nofollow noopener" target="_blank">Academia Sinica, Taiwan</a>. I received my BS in the EECS Undergraduate Honors Program from National Yang Ming Chiao Tung University (NYCU), Taiwan.</p> <p style="color: red; font-weight: bold;">In 2027 Fall, I will hire one PhD student who is interested in overcoming fundamental limitations of LLMs (e.g., creativity and factuality), discovering knowledge structure from text, using LLMs to improve education or scientific document analyses, and/or LLMs in general.</p> <p style="color: red; font-weight: bold;">If you are interested in working me in your PhD journey, please fill out this <a href="https://docs.google.com/forms/d/e/1FAIpQLSf3EVCIkyi95y7uBLb1LoEUTbD1YuFUuoSjHkn6dUIjt-lBRw/viewform" rel="external nofollow noopener" target="_blank">form</a>. I will check the form before the application deadline and contact you if we are potentially a good match.</p> <p style="color: red; font-weight: bold;">If you are interested in working with me as an unpaid research intern, please send me your resume/CV to my arizona email address ([email protected]). I will contact you if I have some projects that fit your background/interest. </p> </div> <h2><a href="/publications/" style="color: inherit;">Selected Publications</a></h2> <div class="publications"> <ol class="bibliography"> <li> <div class="row"> <div class="col-sm-2 abbr"><abbr class="badge">Review</abbr></div>
1 <div id="balakrishnan2026comprehend" class="col-sm-8"> <div class="title">COMPREHEND: Ultra-Fine-Grained Benchmark Understanding by Entailment Hierarchy</div> <div class="author"> Sivaraaman Balakrishnan*, Prakriti Shankar Shetty*, Jeremiah Hong, Hamed Zamani, and <em>Haw-Shiuan Chang*</em> </div> <div class="periodical"> <em>In Under Review</em>, 2026 </div> <div class="periodical"> </div> <div class="links"> <a href="https://openreview.net/pdf/79ec8d9753400abf35737f25b7a55f31483dc116.pdf" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="external nofollow noopener">Paper</a> <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> </div> <div class="bibtex hidden"> <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@inproceedings</span><span class="p">{</span><span class="nl">balakrishnan2026comprehend</span><span class="p">,</span>
2  <span class="na">title</span> <span class="p">=</span> <span class="s">{COMPREHEND: Ultra-Fine-Grained Benchmark Understanding by Entailment Hierarchy}</span><span class="p">,</span>
3  <span class="na">author</span> <span class="p">=</span> <span class="s">{Balakrishnan*, Sivaraaman and Shetty*, Prakriti Shankar and Hong, Jeremiah and Zamani, Hamed and Chang*, Haw-Shiuan}</span><span class="p">,</span>
4  <span class="na">booktitle</span> <span class="p">=</span> <span class="s">{Under Review}</span><span class="p">,</span>
5  <span class="na">paper</span> <span class="p">=</span> <span class="s">{https://openreview.net/pdf/79ec8d9753400abf35737f25b7a55f31483dc116.pdf}</span><span class="p">,</span>
6  <span class="na">year</span> <span class="p">=</span> <span class="s">{2026}</span>
7<span class="p">}</span></code></pre></figure> </div> </div> </div> </li> <li> <div class="row"> <div class="col-sm-2 abbr"><abbr class="badge">EMNLP</abbr></div> <div id="bhuiya2026prompt2box" class="col-sm-8"> <div class="title">PROMPT2BOX: Uncovering Entailment Structure among LLM Prompts</div> <div class="author"> Neeladri Bhuiya, Shib Sankar Dasgupta, Andrew McCallum, and <em>Haw-Shiuan Chang</em> </div> <div class="periodical"> <em>In Conference on Empirical Methods in Natural Language Processing (EMNLP)</em>, 2026 </div> <div class="periodical"> </div> <div class="links"> <a href="https://arxiv.org/abs/2603.21438" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="external nofollow noopener">Paper</a> <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> </div> <div class="bibtex hidden"> <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@inproceedings</span><span class="p">{</span><span class="nl">bhuiya2026prompt2box</span><span class="p">,</span>
8  <span class="na">title</span> <span class="p">=</span> <span class="s">{PROMPT2BOX: Uncovering Entailment Structure among LLM Prompts}</span><span class="p">,</span>
9  <span class="na">author</span> <span class="p">=</span> <span class="s">{Bhuiya, Neeladri and Dasgupta, Shib Sankar and McCallum, Andrew and Chang, Haw-Shiuan}</span><span class="p">,</span>
10  <span class="na">booktitle</span> <span class="p">=</span> <span class="s">{Conference on Empirical Methods in Natural Language Processing (EMNLP)}</span><span class="p">,</span>
11  <span class="na">paper</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2603.21438}</span><span class="p">,</span>
12  <span class="na">year</span> <span class="p">=</span> <span class="s">{2026}</span>
13<span class="p">}</span></code></pre></figure> </div> </div> </div> </li> <li> <div class="row"> <div class="col-sm-2 abbr"><abbr class="badge">TACL</abbr></div> <div id="chang2025real" class="col-sm-8"> <div class="title">REAL Sampling: Boosting Factuality and Diversity of Open-Ended Generation by Extrapolating the Entropy of an Infinitely Large LM</div> <div class="author"> <em>Haw-Shiuan Chang</em>, Nanyun Peng, Mohit Bansal, Anil Ramakrishna, and Tagyoung Chung</div> <div class="periodical"> <em>Transactions of the Association for Computational Linguistics (TACL)</em>, 2025 </div> <div class="periodical"> </div> <div class="links"> <a href="https://arxiv.org/abs/2406.07735" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="external nofollow noopener">Paper</a> <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> <a href="https://github.com/amazon-science/llm-asymptotic-decoding/" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Code</a> </div> <div class="bibtex hidden"> <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">chang2025real</span><span class="p">,</span>
14  <span class="na">title</span> <span class="p">=</span> <span class="s">{REAL Sampling: Boosting Factuality and Diversity of Open-Ended Generation by Extrapolating the Entropy of an Infinitely Large LM}</span><span class="p">,</span>
15  <span class="na">author</span> <span class="p">=</span> <span class="s">{Chang, Haw-Shiuan and Peng, Nanyun and Bansal, Mohit and Ramakrishna, Anil and Chung, Tagyoung}</span><span class="p">,</span>
16  <span class="na">
16journal</span> <span class="p">=</span> <span class="s">{Transactions of the Association for Computational Linguistics (TACL)}</span><span class="p">,</span>
17  <span class="na">year</span> <span class="p">=</span> <span class="s">{2025}</span><span class="p">,</span>
18  <span class="na">paper</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2406.07735}</span>
19<span class="p">}</span></code></pre></figure> </div> </div> </div> </li> <li> <div class="row"> <div class="col-sm-2 abbr"><abbr class="badge">EMNLP Oral</abbr></div> <div id="chang2024explaining" class="col-sm-8"> <div class="title">Explaining and Improving Contrastive Decoding by Extrapolating the Probabilities of a Huge and Hypothetical LM</div> <div class="author"> <em>Haw-Shiuan Chang</em>, Nanyun Peng, Mohit Bansal, Anil Ramakrishna, and Tagyoung Chung</div> <div class="periodical"> <em>In Conference on Empirical Methods in Natural Language Processing (EMNLP) (Oral) (🏆 Best Paper Nomination from a Reviewer)</em>, 2024 </div> <div class="periodical"> </div> <div class="links"> <a href="https://arxiv.org/abs/2411.01610" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="external nofollow noopener">Paper</a> <a href="https://go.screenpal.com/watch/cZ6tFEnc2Ta" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="external nofollow noopener">Talk</a> <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> <a href="https://github.com/amazon-science/llm-asymptotic-decoding/" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Code</a> <a href="/assets/pdf/EMNLP_2024_APD_poster.pdf" class="btn btn-sm z-depth-0" role="button">Poster</a> <a href="/assets/pdf/EMNLP_2024_APD_release.key" class="btn btn-sm z-depth-0" role="button">Slides</a> </div> <div class="bibtex hidden"> <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@inproceedings</span><span class="p">{</span><span class="nl">chang2024explaining</span><span class="p">,</span>
20  <span class="na">title</span> <span class="p">=</span> <span class="s">{Explaining and Improving Contrastive Decoding by Extrapolating the Probabilities of a Huge and Hypothetical LM}</span><span class="p">,</span>
21  <span class="na">author</span> <span class="p">=</span> <span class="s">{Chang, Haw-Shiuan and Peng, Nanyun and Bansal, Mohit and Ramakrishna, Anil and Chung, Tagyoung}</span><span class="p">,</span>
22  <span class="na">booktitle</span> <span class="p">=</span> <span class="s">{Conference on Empirical Methods in Natural Language Processing (EMNLP) (Oral) (🏆 Best Paper Nomination from a Reviewer)}</span><span class="p">,</span>
23  <span class="na">paper</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2411.01610}</span><span class="p">,</span>
24  <span class="na">year</span> <span class="p">=</span> <span class="s">{2024}</span>
25<span class="p">}</span></code></pre></figure> </div> <div class="abstract hidden"> <div style="text-align: center;"> <figure> <iframe src="https://go.screenpal.com/watch/cZ6tFEnc2Ta" class="img-fluid rounded z-depth-1" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowfullscreen width="auto" height="auto"></iframe> </figure> </div> </div> </div> </div> </li> <li> <div class="row"> <div class="col-sm-2 abbr"><abbr class="badge">WSDM</abbr></div> <div id="chang2024copy" class="col-sm-8"> <div class="title">To Copy, or not to Copy; That is a Critical Issue of the Output Softmax Layer in Neural Sequential Recommenders</div> <div class="author"> <em>Haw-Shiuan Chang</em>, Nikhil Agarwal, and Andrew McCallum</div> <div class="periodical"> <em>In Proceedings of The 17th ACM International Conference on Web Search and Data Mining</em>, 2024 </div> <div class="periodical"> </div> <div class="links"> <a href="https://arxiv.org/abs/2310.14079" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="external nofollow noopener">Paper</a> <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> <a href="https://github.com/iesl/softmax_CPR_recommend" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Code</a> <a href="/assets/pdf/WSDM_vertical_poster.pdf" class="btn btn-sm z-depth-0" role="button">Poster</a> </div> <div class="bibtex hidden"> <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@inproceedings</span><span class="p">{</span><span class="nl">chang2024copy</span><span class="p">,</span>
26  <span class="na">title</span> <span class="p">=</span> <span class="s">{To Copy, or not to Copy; That is a Critical Issue of the Output Softmax Layer in Neural Sequential Recommenders}</span><span class="p">,</span>
27  <span class="na">author</span> <span class="p">=</span> <span class="s">{Chang, Haw-Shiuan and Agarwal, Nikhil and McCallum, Andrew}</span><span class="p">,</span>
28  <span class="na">booktitle</span> <span class="p">=</span> <span class="s">{Proceedings of The 17th ACM International Conference on Web Search and Data Mining}</span><span class="p">,</span>
29  <span class="na">year</span> <span class="p">=</span> <span class="s">{2024}</span><span class="p">,</span>
30  <span class="na">paper</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2310.14079}</span><span class="p">,</span>
31<span class="p">}</span></code></pre></figure> </div> </div> </div> </li> <li> <div class="row"> <div class="col-sm-2 abbr"><abbr class="badge">ACL Findings</abbr></div> <div id="chang2023revisiting" class="col-sm-8"> <div class="title">Revisiting the Architectures like Pointer Networks to Efficiently Improve the Next Word Distribution, Summarization Factuality, and Beyond</div> <div class="author"> <em>Haw-Shiuan Chang*</em>, Zonghai Yao*, Alolika Gon, Hong Yu, and Andrew McCallum</div> <div class="periodical"> <em>In Findings of the Association for Computational Linguistics: ACL 2023 (Findings of ACL)</em>, 2023 </div> <div class="periodical"> </div> <div class="links"> <a href="http://arxiv.org/abs/2305.12289" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="external nofollow noopener">Paper</a> <a href="https://screenpal.com/watch/c0111cVB2t5" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="external nofollow noopener">Talk</a> <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> <a href="https://github.com/iesl/Softmax-CPR" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Code</a> <a href="/assets/pdf/ACL_2023_pointer_poster.pdf" class="btn btn-sm z-depth-0" role="button">Poster</a> <a href="/assets/pdf/ACL_2023_softmax-cpr.pdf" class="btn btn-sm z-depth-0" role="button">Slides</a> </div> <div class="bibtex hidden"> <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@inproceedings</span><span class="p">{</span><span class="nl">chang2023revisiting</span><span class="p">,</span>
32  <span class="na">title</span> <span class="p">=</span> <span class="s">{Revisiting the Architectures like Pointer Networks to Efficiently Improve the Next Word Distribution, Summarization Factuality, and Beyond}</span><span class="p">,</span>
33  <span class="na">author</span> <span class="p">=</span> <span class="s">{Chang*, Haw-Shiuan and Yao*, Zonghai and Gon, Alolika and Yu, Hong and McCallum, Andrew}</span><span class="p">,</span>
34  <span class="na">booktitle</span> <span class="p">=</span> <span class="s">{Findings of the Association for Computational Linguistics: ACL 2023 (Findings of ACL)}</span><span class="p">,</span>
35  <span class="na">year</span> <span class="p">=</span> <span class="s">{2023}</span><span class="p">,</span>
36  <span class="na">paper</span> <span class="p">=</span> <span class="s">{http://arxiv.org/abs/2305.12289}</span><span class="p">,</span>
37<span class="p">}</span></code></pre></figure> </div> <div class="abstract hidden"> <div style="text-align: center;"> <figure> <iframe src="https://screenpal.com/watch/c0111cVB2t5" class="img-fluid rounded z-depth-1" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowfullscreen width="auto" height="auto"></iframe> </figure> </div> </div> </div> </div> </li> <li> <div class="row"> <div class="col-sm-2 abbr"><abbr class="badge">ACL</abbr></div> <div id="chang2023multi-cls" class="col-sm-8"> <div class="title">Multi-CLS BERT: An Efficient Alternative to Traditional Ensembling</div> <div class="author"> <em>Haw-Shiuan Chang*</em>, Ruei-Yao Sun*, Kathryn Ricci*, and Andrew McCallum</div> <div class="periodical"> <em>In Annual Meeting of the Association for Computational Linguistics (ACL)</em>, 2023 </div> <div class="periodical"> </div> <div class="links"> <a href="https://arxiv.org/abs/2210.05043" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="external nofollow noopener">Paper</a> <a href="https://screenpal.com/watch/c01hjIVBXqN" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="external nofollow noopener">Talk</a> <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a>
37 <a href="https://github.com/iesl/multicls/" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Code</a> <a href="/assets/pdf/ACL_2023_multi-cls_poster.pdf" class="btn btn-sm z-depth-0" role="button">Poster</a> <a href="/assets/pdf/ACL2023_multi-cls_slides.pdf" class="btn btn-sm z-depth-0" role="button">Slides</a> </div> <div class="bibtex hidden"> <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@inproceedings</span><span class="p">{</span><span class="nl">chang2023multi-cls</span><span class="p">,</span>
38  <span class="na">title</span> <span class="p">=</span> <span class="s">{Multi-CLS BERT: An Efficient Alternative to Traditional Ensembling}</span><span class="p">,</span>
39  <span class="na">author</span> <span class="p">=</span> <span class="s">{Chang*, Haw-Shiuan and Sun*, Ruei-Yao and Ricci*, Kathryn and McCallum, Andrew}</span><span class="p">,</span>
40  <span class="na">booktitle</span> <span class="p">=</span> <span class="s">{Annual Meeting of the Association for Computational Linguistics (ACL)}</span><span class="p">,</span>
41  <span class="na">year</span> <span class="p">=</span> <span class="s">{2023}</span><span class="p">,</span>
42  <span class="na">paper</span> <span class="p">=</span> <span class="s">{https://arxiv.org/abs/2210.05043}</span><span class="p">,</span>
43<span class="p">}</span></code></pre></figure> </div> <div class="abstract hidden"> <div style="text-align: center;"> <figure> <iframe src="https://screenpal.com/watch/c01hjIVBXqN" class="img-fluid rounded z-depth-1" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowfullscreen width="auto" height="auto"></iframe> </figure> </div> </div> </div> </div> </li> <li> <div class="row"> <div class="col-sm-2 abbr"><abbr class="badge">ACL</abbr></div> <div id="chang2022softmax" class="col-sm-8"> <div class="title">Softmax Bottleneck Makes Language Models Unable to Represent Multi-mode Word Distributions</div> <div class="author"> <em>Haw-Shiuan Chang</em>, and Andrew McCallum</div> <div class="periodical"> <em>In Annual Meeting of the Association for Computational Linguistics (ACL)</em>, 2022 </div> <div class="periodical"> </div> <div class="links"> <a href="https://aclanthology.org/2022.acl-long.554.pdf" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="external nofollow noopener">Paper</a> <a href="https://screencast-o-matic.com/watch/c3fresVYo4M" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="external nofollow noopener">Talk</a> <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> <a href="https://aclanthology.org/attachments/2022.acl-long.554.software.zip" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Code</a> <a href="/assets/pdf/ACL2022_poster.pdf" class="btn btn-sm z-depth-0" role="button">Poster</a> <a href="/assets/pdf/ACL2022_slides.pdf" class="btn btn-sm z-depth-0" role="button">Slides</a> </div> <div class="bibtex hidden"> <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@inproceedings</span><span class="p">{</span><span class="nl">chang2022softmax</span><span class="p">,</span>
44  <span class="na">title</span> <span class="p">=</span> <span class="s">{Softmax Bottleneck Makes Language Models Unable to Represent Multi-mode Word Distributions}</span><span class="p">,</span>
45  <span class="na">author</span> <span class="p">=</span> <span class="s">{Chang, Haw-Shiuan and McCallum, Andrew}</span><span class="p">,</span>
46  <span class="na">booktitle</span> <span class="p">=</span> <span class="s">{Annual Meeting of the Association for Computational Linguistics (ACL)}</span><span class="p">,</span>
47  <span class="na">year</span> <span class="p">=</span> <span class="s">{2022}</span><span class="p">,</span>
48  <span class="na">paper</span> <span class="p">=</span> <span class="s">{https://aclanthology.org/2022.acl-long.554.pdf}</span><span class="p">,</span>
49<span class="p">}</span></code></pre></figure> </div> <div class="abstract hidden"> <div style="text-align: center;"> <figure> <iframe src="https://screencast-o-matic.com/watch/c3fresVYo4M" class="img-fluid rounded z-depth-1" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowfullscreen width="auto" height="auto"></iframe> </figure> </div> </div> </div> </div> </li> <li> <div class="row"> <div class="col-sm-2 abbr"><abbr class="badge">ML</abbr></div> <div id="chang2019ovecoming" class="col-sm-8"> <div class="title">Using Error Decay Prediction to Overcome Practical Issues of Deep Active Learning for Named Entity Recognition</div> <div class="author"> <em>Haw-Shiuan Chang</em>, Shankar Vembu, Sunil Mohan, Rheeya Uppaal, and Andrew McCallum</div> <div class="periodical"> <em>Machine Learning</em>, 2020 </div> <div class="periodical"> </div> <div class="links"> <a href="http://arxiv.org/abs/1911.07335" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="external nofollow noopener">Paper</a> <a href="https://slideslive.com/38933012/using-error-decay-prediction-to-overcome-practical-issues-of-deep-active-learning-for-named-entity-recognition" class="btn btn-sm z-depth-0" role="button" target="_blank" rel="external nofollow noopener">Talk</a> <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> <a href="https://docs.google.com/presentation/d/1h3bI1dS8-vS5ItcfGOHN3J7Wi7dnWWBf0gA4-yBXrk4/edit?usp=sharing" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Slides</a> </div> <div class="bibtex hidden"> <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">chang2019ovecoming</span><span class="p">,</span>
50  <span class="na">author</span> <span class="p">=</span> <span class="s">{Chang, Haw-Shiuan and Vembu, Shankar and Mohan, Sunil and Uppaal, Rheeya and McCallum, Andrew}</span><span class="p">,</span>
51  <span class="na">
51journal</span> <span class="p">=</span> <span class="s">{Machine Learning}</span><span class="p">,</span>
52  <span class="na">paper</span> <span class="p">=</span> <span class="s">{http://arxiv.org/abs/1911.07335}</span><span class="p">,</span>
53  <span class="na">title</span> <span class="p">=</span> <span class="s">{Using Error Decay Prediction to Overcome Practical Issues of Deep Active Learning for Named Entity Recognition}</span><span class="p">,</span>
54  <span class="na">year</span> <span class="p">=</span> <span class="s">{2020}</span><span class="p">,</span>
55  <span class="na">doi</span> <span class="p">=</span> <span class="s">{10.1007/s10994-020-05897-1}</span><span class="p">,</span>
56  <span class="na">publisher</span> <span class="p">=</span> <span class="s">{Springer}</span><span class="p">,</span>
57<span class="p">}</span></code></pre></figure> </div> <div class="abstract hidden"> <div style="text-align: center;"> <figure> <iframe src="https://slideslive.com/38933012/using-error-decay-prediction-to-overcome-practical-issues-of-deep-active-learning-for-named-entity-recognition" class="img-fluid rounded z-depth-1" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowfullscreen width="auto" height="auto"></iframe> </figure> </div> </div> </div> </div> </li> </ol> </div> <div class="social"> <div class="contact-icons"> <a href="mailto:%68%73%63%68%61%6E%67@%63%73.%75%6D%61%73%73.%65%64%75" title="email"><i class="fa-solid fa-envelope"></i></a> <a href="https://www.semanticscholar.org/author/144827671" title="Semantic Scholar" rel="external nofollow noopener" target="_blank"><i class="ai ai-semantic-scholar"></i></a> <a href="https://scholar.google.com/citations?user=fdGXVd4AAAAJ" title="Google Scholar" rel="external nofollow noopener" target="_blank"><i class="ai ai-google-scholar"></i></a> <a href="https://twitter.com/Haw_Shiuan" title="X" rel="external nofollow noopener" target="_blank"><i class="fa-brands fa-x-twitter"></i></a> </div> <div class="contact-note"> </div> </div> </article> </div> </div> <footer class="fixed-bottom"> <div class="container mt-0"> © Copyright 2026 Haw-Shiuan Chang. Powered by <a href="https://jekyllrb.com/" target="_blank" rel="external nofollow noopener">Jekyll</a> with <a href="https://github.com/alshedivat/al-folio" rel="external nofollow noopener" target="_blank">al-folio</a> theme. Hosted by <a href="https://pages.github.com/" target="_blank" rel="external nofollow noopener">GitHub Pages</a>. </div> </footer> 
57<script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/jquery.min.js" integrity="sha256-/xUj+3OJU5yExlq6GSYGSHk7tPXikynS7ogEvDej/m4=" crossorigin="anonymous"></script>
57 
57<script src="/assets/js/bootstrap.bundle.min.js"></script>
57 
57<script src="https://cdn.jsdelivr.net/npm/[email protected]/js/mdb.min.js" integrity="sha256-NdbiivsvWt7VYCt6hYNT3h/th9vSTL4EDWeGs5SN3DA=" crossorigin="anonymous"></script>
57 
57<script defer src="https://cdn.jsdelivr.net/npm/[email protected]/dist/masonry.pkgd.min.js" integrity="sha256-Nn1q/fx0H7SNLZMQ5Hw5JLaTRZp0yILA/FRexe19VdI=" crossorigin="anonymous"></script>
57 
57<script defer src="https://cdn.jsdelivr.net/npm/imagesloaded@4/imagesloaded.pkgd.min.js"></script>
57 
57<script defer src="/assets/js/masonry.js" type="text/javascript"></script>
57 
57<script defer src="https://cdn.jsdelivr.net/npm/[email protected]/dist/medium-zoom.min.js" integrity="sha256-7PhEpEWEW0XXQ0k6kQrPKwuoIomz8R8IYyuU1Qew4P8=" crossorigin="anonymous"></script>
57 
57<script defer src="/assets/js/zoom.js?7b30caa5023af4af8408a472dc4e1ebb"></script>
57 
57<script defer src="https://unpkg.com/[email protected]/dist/bootstrap-table.min.js"></script>
57 
57<script src="/assets/js/no_defer.js?d633890033921b33e0ceb13d22340a9c"></script>
57 
57<script defer src="/assets/js/common.js?acdb9690d7641b2f8d40529018c71a01"></script>
57 
57<script defer src="/assets/js/copy_code.js?9b43d6e67ddc7c0855b1478ee4c48c2d" type="text/javascript"></script>
57 
57<script async src="https://d1bxh8uas1mnw7.cloudfront.net/assets/embed.js"></script>
57 
57<script async src="https://badge.dimensions.ai/badge.js"></script>
57 
57<script type="text/javascript">window.MathJax={tex:{tags:"ams"}};</script>
57 
57<script defer type="text/javascript" id="MathJax-script" src="https://cdn.jsdelivr.net/npm/[email protected]/es5/tex-mml-chtml.js"></script>
57 
57<script type="text/javascript">function progressBarSetup(){"max"in document.createElement("progress")?(initializeProgressElement(),$(document).on("scroll",function(){progressBar.attr({value:getCurrentScrollPosition()})}),$(window).on("resize",initializeProgressElement)):(resizeProgressBar(),$(document).on("scroll",resizeProgressBar),$(window).on("resize",resizeProgressBar))}function getCurrentScrollPosition(){return $(window).scrollTop()}function initializeProgressElement(){let e=$("#navbar").outerHeight(!0);$("body").css({"padding-top":e}),$("progress-container").css({"padding-top":e}),progressBar.css({top:e}),progressBar.attr({max:getDistanceToScroll(),value:getCurrentScrollPosition()})}function getDistanceToScroll(){return $(document).height()-$(window).height()}function resizeProgressBar(){progressBar.css({width:getWidthPercentage()+"%"})}function getWidthPercentage(){return getCurrentScrollPosition()/getDistanceToScroll()*100}const progressBar=$("#progress");window.onload=function(){setTimeout(progressBarSetup,50)};</script>
57 </header> </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.