Search | arXiv e-print repository

<!DOCTYPE html> <html lang="en"> <head> <meta charset="utf-8"/> <meta name="viewport" content="width=device-width, initial-scale=1"/>  <link rel="apple-touch-icon" sizes="180x180" href="https://static.arxiv.org/static/base/1.0.0a5/images/icons/apple-touch-icon.png"> <link rel="icon" type="image/png" sizes="32x32" href="https://static.arxiv.org/static/base/1.0.0a5/images/icons/favicon-32x32.png"> <link rel="icon" type="image/png" sizes="16x16" href="https://static.arxiv.org/static/base/1.0.0a5/images/icons/favicon-16x16.png"> <link rel="manifest" href="https://static.arxiv.org/static/base/1.0.0a5/images/icons/site.webmanifest"> <link rel="mask-icon" href="https://static.arxiv.org/static/base/1.0.0a5/images/icons/safari-pinned-tab.svg" color="#b31b1b"> <link rel="shortcut icon" href="https://static.arxiv.org/static/base/1.0.0a5/images/icons/favicon.ico"> <meta name="msapplication-TileColor" content="#b31b1b"> <meta name="msapplication-config" content="images/icons/browserconfig.xml"> <meta name="theme-color" content="#b31b1b">  <title>Search | arXiv e-print repository</title> <script defer src="https://static.arxiv.org/static/base/1.0.0a5/fontawesome-free-5.11.2-web/js/all.js"></script> <link rel="stylesheet" href="https://static.arxiv.org/static/base/1.0.0a5/css/arxivstyle.css" /> <script type="text/x-mathjax-config"> MathJax.Hub.Config({ messageStyle: "none", extensions: ["tex2jax.js"], jax: ["input/TeX", "output/HTML-CSS"], tex2jax: { inlineMath: [ ['$','$'], ["\$","\$"] ], displayMath: [ ['$$','$$'], ["\\[","\\]"] ], processEscapes: true, ignoreClass: '.*', processClass: 'mathjax.*' }, TeX: { extensions: ["AMSmath.js", "AMSsymbols.js", "noErrors.js"], noErrors: { inlineDelimiters: ["$","$"], multiLine: false, style: { "font-size": "normal", "border": "" } } }, "HTML-CSS": { availableFonts: ["TeX"] } }); </script> <script src='//static.arxiv.org/MathJax-2.7.3/MathJax.js'></script> <script src="https://static.arxiv.org/static/base/1.0.0a5/js/notification.js"></script> <link rel="stylesheet" href="https://static.arxiv.org/static/search/0.5.6/css/bulma-tooltip.min.css" /> <link rel="stylesheet" href="https://static.arxiv.org/static/search/0.5.6/css/search.css" /> <script src="https://code.jquery.com/jquery-3.2.1.slim.min.js" integrity="sha256-k2WSCIexGzOj3Euiig+TlR8gA0EmPjuc79OEeY5L45g=" crossorigin="anonymous"></script> <script src="https://static.arxiv.org/static/search/0.5.6/js/fieldset.js"></script> <style> radio#cf-customfield_11400 { display: none; } </style> </head> <body> <header><a href="#main-container" class="is-sr-only">Skip to main content</a>  <div class="attribution level is-marginless" role="banner"> <div class="level-left"> <a class="level-item" href="https://cornell.edu/"><img src="https://static.arxiv.org/static/base/1.0.0a5/images/cornell-reduced-white-SMALL.svg" alt="Cornell University" width="200" aria-label="logo" /></a> </div> <div class="level-right is-marginless"><p class="sponsors level-item is-marginless"><span id="support-ack-url">We gratefully acknowledge support from<br /> the Simons Foundation, <a href="https://info.arxiv.org/about/ourmembers.html">member institutions</a>, and all contributors. <a href="https://info.arxiv.org/about/donate.html">Donate</a></span></p></div> </div>  <div class="identity level is-marginless"> <div class="level-left"> <div class="level-item"> <a class="arxiv" href="https://arxiv.org/" aria-label="arxiv-logo"> <img src="https://static.arxiv.org/static/base/1.0.0a5/images/arxiv-logo-one-color-white.svg" aria-label="logo" alt="arxiv logo" width="85" style="width:85px;"/> </a> </div> </div> <div class="search-block level-right"> <form class="level-item mini-search" method="GET" action="https://arxiv.org/search"> <div class="field has-addons"> <div class="control"> <input class="input is-small" type="text" name="query" placeholder="Search..." aria-label="Search term or terms" /> <p class="help"><a href="https://info.arxiv.org/help">Help</a> | <a href="https://arxiv.org/search/advanced">Advanced Search</a></p> </div> <div class="control"> <div class="select is-small"> <select name="searchtype" aria-label="Field to search"> <option value="all" selected="selected">All fields</option> <option value="title">Title</option> <option value="author">Author</option> <option value="abstract">Abstract</option> <option value="comments">Comments</option> <option value="journal_ref">Journal reference</option> <option value="acm_class">ACM classification</option> <option value="msc_class">MSC classification</option> <option value="report_num">Report number</option> <option value="paper_id">arXiv identifier</option> <option value="doi">DOI</option> <option value="orcid">ORCID</option> <option value="author_id">arXiv author ID</option> <option value="help">Help pages</option> <option value="full_text">Full text</option> </select> </div> </div> <input type="hidden" name="source" value="header"> <button class="button is-small is-cul-darker">Search</button> </div> </form> </div> </div>  <div class="container"> <div class="user-tools is-size-7 has-text-right has-text-weight-bold" role="navigation" aria-label="User menu"> <a href="https://arxiv.org/login">Login</a> </div> </div> </header> <main class="container" id="main-container"> <div class="level is-marginless"> <div class="level-left"> <h1 class="title is-clearfix"> Showing 1–50 of 15,642 results for author: <span class="mathjax">Chen, Y</span> </h1> </div> <div class="level-right is-hidden-mobile">  <span class="help" style="display: inline-block;"><a href="https://github.com/arXiv/arxiv-search/releases">Search v0.5.6 released 2020-02-24</a>  </span> </div> </div> <div class="content"> <form method="GET" action="/search/" aria-role="search"> <div class="field has-addons-tablet"> <div class="control is-expanded"> <label for="query" class="hidden-label">Search term or terms</label> <input class="input is-medium" id="query" name="query" placeholder="Search term..." type="text" value="Chen, Y"> </div> <div class="select control is-medium"> <label class="is-hidden" for="searchtype">Field</label> <select class="is-medium" id="searchtype" name="searchtype"><option value="all">All fields</option><option value="title">Title</option><option selected value="author">Author(s)</option><option value="abstract">Abstract</option><option value="comments">Comments</option><option value="journal_ref">Journal reference</option><option value="acm_class">ACM classification</option><option value="msc_class">MSC classification</option><option value="report_num">Report number</option><option value="paper_id">arXiv identifier</option><option value="doi">DOI</option><option value="orcid">ORCID</option><option value="license">License (URI)</option><option value="author_id">arXiv author ID</option><option value="help">Help pages</option><option value="full_text">Full text</option></select> </div> <div class="control"> <button class="button is-link is-medium">Search</button> </div> </div> <div class="field"> <div class="control is-size-7"> <label class="radio"> <input checked id="abstracts-0" name="abstracts" type="radio" value="show"> Show abstracts </label> <label class="radio"> <input id="abstracts-1" name="abstracts" type="radio" value="hide"> Hide abstracts </label> </div> </div> <div class="is-clearfix" style="height: 2.5em"> <div class="is-pulled-right"> <a href="/search/advanced?terms-0-term=Chen%2C+Y&terms-0-field=author&size=50&order=-announced_date_first">Advanced Search</a> </div> </div> <input type="hidden" name="order" value="-announced_date_first"> <input type="hidden" name="size" value="50"> </form> <div class="level breathe-horizontal"> <div class="level-left"> <form method="GET" action="/search/"> <div style="display: none;"> <select id="searchtype" name="searchtype"><option value="all">All fields</option><option value="title">Title</option><option selected value="author">Author(s)</option><option value="abstract">Abstract</option><option value="comments">Comments</option><option value="journal_ref">Journal reference</option><option value="acm_class">ACM classification</option><option value="msc_class">MSC classification</option><option value="report_num">Report number</option><option value="paper_id">arXiv identifier</option><option value="doi">DOI</option><option value="orcid">ORCID</option><option value="license">License (URI)</option><option value="author_id">arXiv author ID</option><option value="help">Help pages</option><option value="full_text">Full text</option></select> <input id="query" name="query" type="text" value="Chen, Y"> <ul id="abstracts"><li><input checked id="abstracts-0" name="abstracts" type="radio" value="show"> <label for="abstracts-0">Show abstracts</label></li><li><input id="abstracts-1" name="abstracts" type="radio" value="hide"> <label for="abstracts-1">Hide abstracts</label></li></ul> </div> <div class="box field is-grouped is-grouped-multiline level-item"> <div class="control"> <span class="select is-small"> <select id="size" name="size"><option value="25">25</option><option selected value="50">50</option><option value="100">100</option><option value="200">200</option></select> </span> <label for="size">results per page</label>. </div> <div class="control"> <label for="order">Sort results by</label> <span class="select is-small"> <select id="order" name="order"><option selected value="-announced_date_first">Announcement date (newest first)</option><option value="announced_date_first">Announcement date (oldest first)</option><option value="-submitted_date">Submission date (newest first)</option><option value="submitted_date">Submission date (oldest first)</option><option value="">Relevance</option></select> </span> </div> <div class="control"> <button class="button is-small is-link">Go</button> </div> </div> </form> </div> </div> <nav class="pagination is-small is-centered breathe-horizontal" role="navigation" aria-label="pagination"> <a href="" class="pagination-previous is-invisible">Previous </a> <a href="/search/?searchtype=author&query=Chen%2C+Y&start=50" class="pagination-next" >Next </a> <ul class="pagination-list"> <li> <a href="/search/?searchtype=author&query=Chen%2C+Y&start=0" class="pagination-link is-current" aria-label="Goto page 1">1 </a> </li> <li> <a href="/search/?searchtype=author&query=Chen%2C+Y&start=50" class="pagination-link " aria-label="Page 2" aria-current="page">2 </a> </li> <li> <a href="/search/?searchtype=author&query=Chen%2C+Y&start=100" class="pagination-link " aria-label="Page 3" aria-current="page">3 </a> </li> <li> <a href="/search/?searchtype=author&query=Chen%2C+Y&start=150" class="pagination-link " aria-label="Page 4" aria-current="page">4 </a> </li> <li> <a href="/search/?searchtype=author&query=Chen%2C+Y&start=200" class="pagination-link " aria-label="Page 5" aria-current="page">5 </a> </li> <li><span class="pagination-ellipsis">…</span></li> </ul> </nav> <ol class="breathe-horizontal" start="1"> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.14747">arXiv:2502.14747</a> <span> [<a href="https://arxiv.org/pdf/2502.14747">pdf</a>, <a href="https://arxiv.org/format/2502.14747">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Human-Computer Interaction">cs.HC</span> </div> <div class="is-inline-block" style="margin-left: 0.5rem"> <div class="tags has-addons"> <span class="tag is-dark is-size-7">doi</span> <span class="tag is-light is-size-7"><a class="" href="https://doi.org/10.1145/3706598.3714148">10.1145/3706598.3714148 <i class="fa fa-external-link" aria-hidden="true"></i></a></span> </div> </div> </div> <p class="title is-5 mathjax"> AIdeation: Designing a Human-AI Collaborative Ideation System for Concept Designers </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Wang%2C+W">Wen-Fan Wang</a>, <a href="/search/?searchtype=author&query=Lu%2C+C">Chien-Ting Lu</a>, <a href="/search/?searchtype=author&query=Campany%C3%A0%2C+N+P">Nil Ponsa Campany脿</a>, <a href="/search/?searchtype=author&query=Chen%2C+B">Bing-Yu Chen</a>, <a href="/search/?searchtype=author&query=Chen%2C+M+Y">Mike Y. Chen</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.14747v1-abstract-short" style="display: inline;"> Concept designers in the entertainment industry create highly detailed, often imaginary environments for movies, games, and TV shows. Their early ideation phase requires intensive research, brainstorming, visual exploration, and combination of various design elements to form cohesive designs. However, existing AI tools focus on image generation from user specifications, lacking support for the uni… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14747v1-abstract-full').style.display = 'inline'; document.getElementById('2502.14747v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.14747v1-abstract-full" style="display: none;"> Concept designers in the entertainment industry create highly detailed, often imaginary environments for movies, games, and TV shows. Their early ideation phase requires intensive research, brainstorming, visual exploration, and combination of various design elements to form cohesive designs. However, existing AI tools focus on image generation from user specifications, lacking support for the unique needs and complexity of concept designers' workflows. Through a formative study with 12 professional designers, we captured their workflows and identified key requirements for AI-assisted ideation tools. Leveraging these insights, we developed AIdeation to support early ideation by brainstorming design concepts with flexible searching and recombination of reference images. A user study with 16 professional designers showed that AIdeation significantly enhanced creativity, ideation efficiency, and satisfaction (all p<.01) compared to current tools and workflows. A field study with 4 studios for 1 week provided insights into AIdeation's benefits and limitations in real-world projects. After the completion of the field study, two studios, covering films, television, and games, have continued to use AIdeation in their commercial projects to date, further validating AIdeation's improvement in ideation quality and efficiency. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14747v1-abstract-full').style.display = 'none'; document.getElementById('2502.14747v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 20 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">Accepted ACM CHI Conference on Human Factors in Computing Systems (CHI '25)</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.14739">arXiv:2502.14739</a> <span> [<a href="https://arxiv.org/pdf/2502.14739">pdf</a>, <a href="https://arxiv.org/format/2502.14739">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computation and Language">cs.CL</span> </div> </div> <p class="title is-5 mathjax"> SuperGPQA: Scaling LLM Evaluation across 285 Graduate Disciplines </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Team%2C+M">M-A-P Team</a>, <a href="/search/?searchtype=author&query=Du%2C+X">Xinrun Du</a>, <a href="/search/?searchtype=author&query=Yao%2C+Y">Yifan Yao</a>, <a href="/search/?searchtype=author&query=Ma%2C+K">Kaijing Ma</a>, <a href="/search/?searchtype=author&query=Wang%2C+B">Bingli Wang</a>, <a href="/search/?searchtype=author&query=Zheng%2C+T">Tianyu Zheng</a>, <a href="/search/?searchtype=author&query=Zhu%2C+K">Kang Zhu</a>, <a href="/search/?searchtype=author&query=Liu%2C+M">Minghao Liu</a>, <a href="/search/?searchtype=author&query=Liang%2C+Y">Yiming Liang</a>, <a href="/search/?searchtype=author&query=Jin%2C+X">Xiaolong Jin</a>, <a href="/search/?searchtype=author&query=Wei%2C+Z">Zhenlin Wei</a>, <a href="/search/?searchtype=author&query=Zheng%2C+C">Chujie Zheng</a>, <a href="/search/?searchtype=author&query=Deng%2C+K">Kaixing Deng</a>, <a href="/search/?searchtype=author&query=Guo%2C+S">Shuyue Guo</a>, <a href="/search/?searchtype=author&query=Jia%2C+S">Shian Jia</a>, <a href="/search/?searchtype=author&query=Jiang%2C+S">Sichao Jiang</a>, <a href="/search/?searchtype=author&query=Liao%2C+Y">Yiyan Liao</a>, <a href="/search/?searchtype=author&query=Li%2C+R">Rui Li</a>, <a href="/search/?searchtype=author&query=Li%2C+Q">Qinrui Li</a>, <a href="/search/?searchtype=author&query=Li%2C+S">Sirun Li</a>, <a href="/search/?searchtype=author&query=Li%2C+Y">Yizhi Li</a>, <a href="/search/?searchtype=author&query=Li%2C+Y">Yunwen Li</a>, <a href="/search/?searchtype=author&query=Ma%2C+D">Dehua Ma</a>, <a href="/search/?searchtype=author&query=Ni%2C+Y">Yuansheng Ni</a>, <a href="/search/?searchtype=author&query=Que%2C+H">Haoran Que</a> , et al. (70 additional authors not shown) </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.14739v1-abstract-short" style="display: inline;"> Large language models (LLMs) have demonstrated remarkable proficiency in mainstream academic disciplines such as mathematics, physics, and computer science. However, human knowledge encompasses over 200 specialized disciplines, far exceeding the scope of existing benchmarks. The capabilities of LLMs in many of these specialized fields-particularly in light industry, agriculture, and service-orient… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14739v1-abstract-full').style.display = 'inline'; document.getElementById('2502.14739v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.14739v1-abstract-full" style="display: none;"> Large language models (LLMs) have demonstrated remarkable proficiency in mainstream academic disciplines such as mathematics, physics, and computer science. However, human knowledge encompasses over 200 specialized disciplines, far exceeding the scope of existing benchmarks. The capabilities of LLMs in many of these specialized fields-particularly in light industry, agriculture, and service-oriented disciplines-remain inadequately evaluated. To address this gap, we present SuperGPQA, a comprehensive benchmark that evaluates graduate-level knowledge and reasoning capabilities across 285 disciplines. Our benchmark employs a novel Human-LLM collaborative filtering mechanism to eliminate trivial or ambiguous questions through iterative refinement based on both LLM responses and expert feedback. Our experimental results reveal significant room for improvement in the performance of current state-of-the-art LLMs across diverse knowledge domains (e.g., the reasoning-focused model DeepSeek-R1 achieved the highest accuracy of 61.82% on SuperGPQA), highlighting the considerable gap between current model capabilities and artificial general intelligence. Additionally, we present comprehensive insights from our management of a large-scale annotation process, involving over 80 expert annotators and an interactive Human-LLM collaborative system, offering valuable methodological guidance for future research initiatives of comparable scope. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14739v1-abstract-full').style.display = 'none'; document.getElementById('2502.14739v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 20 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.14727">arXiv:2502.14727</a> <span> [<a href="https://arxiv.org/pdf/2502.14727">pdf</a>, <a href="https://arxiv.org/format/2502.14727">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Sound">cs.SD</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Artificial Intelligence">cs.AI</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Audio and Speech Processing">eess.AS</span> </div> </div> <p class="title is-5 mathjax"> WavRAG: Audio-Integrated Retrieval Augmented Generation for Spoken Dialogue Models </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Chen%2C+Y">Yifu Chen</a>, <a href="/search/?searchtype=author&query=Ji%2C+S">Shengpeng Ji</a>, <a href="/search/?searchtype=author&query=Wang%2C+H">Haoxiao Wang</a>, <a href="/search/?searchtype=author&query=Wang%2C+Z">Ziqing Wang</a>, <a href="/search/?searchtype=author&query=Chen%2C+S">Siyu Chen</a>, <a href="/search/?searchtype=author&query=He%2C+J">Jinzheng He</a>, <a href="/search/?searchtype=author&query=Xu%2C+J">Jin Xu</a>, <a href="/search/?searchtype=author&query=Zhao%2C+Z">Zhou Zhao</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.14727v1-abstract-short" style="display: inline;"> Retrieval Augmented Generation (RAG) has gained widespread adoption owing to its capacity to empower large language models (LLMs) to integrate external knowledge. However, existing RAG frameworks are primarily designed for text-based LLMs and rely on Automatic Speech Recognition to process speech input, which discards crucial audio information, risks transcription errors, and increases computation… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14727v1-abstract-full').style.display = 'inline'; document.getElementById('2502.14727v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.14727v1-abstract-full" style="display: none;"> Retrieval Augmented Generation (RAG) has gained widespread adoption owing to its capacity to empower large language models (LLMs) to integrate external knowledge. However, existing RAG frameworks are primarily designed for text-based LLMs and rely on Automatic Speech Recognition to process speech input, which discards crucial audio information, risks transcription errors, and increases computational overhead. Therefore, we introduce WavRAG, the first retrieval augmented generation framework with native, end-to-end audio support. WavRAG offers two key features: 1) Bypassing ASR, WavRAG directly processes raw audio for both embedding and retrieval. 2) WavRAG integrates audio and text into a unified knowledge representation. Specifically, we propose the WavRetriever to facilitate the retrieval from a text-audio hybrid knowledge base, and further enhance the in-context capabilities of spoken dialogue models through the integration of chain-of-thought reasoning. In comparison to state-of-the-art ASR-Text RAG pipelines, WavRAG achieves comparable retrieval performance while delivering a 10x acceleration. Furthermore, WavRAG's unique text-audio hybrid retrieval capability extends the boundaries of RAG to the audio modality. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14727v1-abstract-full').style.display = 'none'; document.getElementById('2502.14727v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 20 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.14542">arXiv:2502.14542</a> <span> [<a href="https://arxiv.org/pdf/2502.14542">pdf</a>, <a href="https://arxiv.org/format/2502.14542">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Materials Science">cond-mat.mtrl-sci</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Computational Physics">physics.comp-ph</span> </div> </div> <p class="title is-5 mathjax"> Lattice distortion tuning resistivity invar effect in high entropy alloys </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Chen%2C+H">Hao Chen</a>, <a href="/search/?searchtype=author&query=Xu%2C+Y">Yuanji Xu</a>, <a href="/search/?searchtype=author&query=Liu%2C+L">Lihua Liu</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yue Chen</a>, <a href="/search/?searchtype=author&query=Wr%C3%B3bel%2C+J">Jan Wr贸bel</a>, <a href="/search/?searchtype=author&query=Cong%2C+D">Daoyong Cong</a>, <a href="/search/?searchtype=author&query=Tian%2C+F">Fuyang Tian</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.14542v1-abstract-short" style="display: inline;"> Materials with an ultra-low temperature coefficient of resistivity are desired for the temperature and flow sensors in high-precision electronic measuring systems. In this work, the Kubo-Greenwood formula, implemented in ab initio molecular dynamics simulations, is employed to predict the finite-temperature resistivity of multi-component alloys with severe lattice distortion. We observe a tiny cha… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14542v1-abstract-full').style.display = 'inline'; document.getElementById('2502.14542v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.14542v1-abstract-full" style="display: none;"> Materials with an ultra-low temperature coefficient of resistivity are desired for the temperature and flow sensors in high-precision electronic measuring systems. In this work, the Kubo-Greenwood formula, implemented in ab initio molecular dynamics simulations, is employed to predict the finite-temperature resistivity of multi-component alloys with severe lattice distortion. We observe a tiny change in resistivity over a wide temperature range in high-entropy alloys. The electronic resistivity invar effect in B2 Ni$_{25}$Co$_{25}$(HfTiZr)$_{50}$ Elinvar alloys results from a balance between intrinsic and residual resistivity. This effect is associated with atomic displacements from ideal lattice sites, which are caused by lattice thermal vibrations and chemical disorder-induced lattice distortions. It is further evidenced by a decrease in lattice distortion with temperature and changes in the electronic density of states. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14542v1-abstract-full').style.display = 'none'; document.getElementById('2502.14542v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 20 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.14508">arXiv:2502.14508</a> <span> [<a href="https://arxiv.org/pdf/2502.14508">pdf</a>, <a href="https://arxiv.org/ps/2502.14508">ps</a>, <a href="https://arxiv.org/format/2502.14508">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="High Energy Astrophysical Phenomena">astro-ph.HE</span> </div> <div class="is-inline-block" style="margin-left: 0.5rem"> <div class="tags has-addons"> <span class="tag is-dark is-size-7">doi</span> <span class="tag is-light is-size-7"><a class="" href="https://doi.org/10.1093/mnras/stac3238">10.1093/mnras/stac3238 <i class="fa fa-external-link" aria-hidden="true"></i></a></span> </div> </div> </div> <p class="title is-5 mathjax"> A possible overall scenario for the outburst evolution of MAXI J1820+070 revealed by Insight-HXMT </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Peng%2C+J+Q">J. Q. Peng</a>, <a href="/search/?searchtype=author&query=Zhang%2C+S">S. Zhang</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y+P">Y. P. Chen</a>, <a href="/search/?searchtype=author&query=Kong%2C+L+D">L. D. Kong</a>, <a href="/search/?searchtype=author&query=Wang%2C+P+J">P. J. Wang</a>, <a href="/search/?searchtype=author&query=Zhang%2C+S+N">S. N. Zhang</a>, <a href="/search/?searchtype=author&query=Ji%2C+L">L. Ji</a>, <a href="/search/?searchtype=author&query=Tao%2C+L">L. Tao</a>, <a href="/search/?searchtype=author&query=Qu%2C+J+L">J. L. Qu</a>, <a href="/search/?searchtype=author&query=Ge%2C+M+Y">M. Y. Ge</a>, <a href="/search/?searchtype=author&query=Shui%2C+Q+C">Q. C. Shui</a>, <a href="/search/?searchtype=author&query=Li%2C+J">J. Li</a>, <a href="/search/?searchtype=author&query=Chang%2C+Z">Z. Chang</a>, <a href="/search/?searchtype=author&query=Li%2C+Z+S">Z. S. Li</a>, <a href="/search/?searchtype=author&query=Xiao%2C+Y+X">Y. X. Xiao</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.14508v1-abstract-short" style="display: inline;"> We study the spectral and temporal properties of the black hole X-ray transient binary MAXI J1820+070 during the 2018 outburst with Insight-HXMT observations. The outburst of MAXI J1820+070 can be divided into three intervals. For the two intervals of the outburst, we find that low-energy (below 140 keV) photos lag high-energy (140-170 keV) ones, while in the decay of the outburst, high-energy pho… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14508v1-abstract-full').style.display = 'inline'; document.getElementById('2502.14508v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.14508v1-abstract-full" style="display: none;"> We study the spectral and temporal properties of the black hole X-ray transient binary MAXI J1820+070 during the 2018 outburst with Insight-HXMT observations. The outburst of MAXI J1820+070 can be divided into three intervals. For the two intervals of the outburst, we find that low-energy (below 140 keV) photos lag high-energy (140-170 keV) ones, while in the decay of the outburst, high-energy photons lag low-energy photons, both with a time scale of the order of days. Based on these results, the canonical hysteresis effect of the 'q' shape in the hardness-intensity diagram can be reformed into a roughly linear shape by taking into account the lag corrections between different energy bands. Time analysis shows that the high-frequency break of hard X-rays, derived from the power density spectrum of the first interval of the outburst is, in general, larger and more variable than that of soft X-rays. The spectral fitting shows that the coverage fraction of the hard X-rays drops sharply at the beginning of the outburst to around 0.5, then increases slightly. The coverage fraction drops to roughly zero once the source steps into a soft state and increases gradually to unity when the source returns to a low hard state. We discuss the possible overall evolution scenario of corona hinted from these discoveries. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14508v1-abstract-full').style.display = 'none'; document.getElementById('2502.14508v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 20 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">9 pages,10 figures. Published in MNRAS</span> </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Journal ref:</span> Monthly Notices of the Royal Astronomical Society, Volume 518, Issue 2, pp.2521-2528. January 2023 </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.14340">arXiv:2502.14340</a> <span> [<a href="https://arxiv.org/pdf/2502.14340">pdf</a>, <a href="https://arxiv.org/format/2502.14340">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computation and Language">cs.CL</span> </div> </div> <p class="title is-5 mathjax"> Earlier Tokens Contribute More: Learning Direct Preference Optimization From Temporal Decay Perspective </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Shao%2C+R">Ruichen Shao</a>, <a href="/search/?searchtype=author&query=Li%2C+B">Bei Li</a>, <a href="/search/?searchtype=author&query=Liu%2C+G">Gangao Liu</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yang Chen</a>, <a href="/search/?searchtype=author&query=Zhou%2C+X">Xiang Zhou</a>, <a href="/search/?searchtype=author&query=Wang%2C+J">Jingang Wang</a>, <a href="/search/?searchtype=author&query=Cai%2C+X">Xunliang Cai</a>, <a href="/search/?searchtype=author&query=Li%2C+P">Peng Li</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.14340v1-abstract-short" style="display: inline;"> Direct Preference Optimization (DPO) has gained attention as an efficient alternative to reinforcement learning from human feedback (RLHF) for aligning large language models (LLMs) with human preferences. Despite its advantages, DPO suffers from a length bias, generating responses longer than those from the reference model. Existing solutions like SimPO and SamPO address this issue but uniformly t… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14340v1-abstract-full').style.display = 'inline'; document.getElementById('2502.14340v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.14340v1-abstract-full" style="display: none;"> Direct Preference Optimization (DPO) has gained attention as an efficient alternative to reinforcement learning from human feedback (RLHF) for aligning large language models (LLMs) with human preferences. Despite its advantages, DPO suffers from a length bias, generating responses longer than those from the reference model. Existing solutions like SimPO and SamPO address this issue but uniformly treat the contribution of rewards across sequences, overlooking temporal dynamics. To this end, we propose an enhanced preference optimization method that incorporates a temporal decay factor controlled by a gamma parameter. This dynamic weighting mechanism adjusts the influence of each reward based on its position in the sequence, prioritizing earlier tokens that are more critical for alignment. By adaptively focusing on more relevant feedback, our approach mitigates overfitting to less pertinent data and remains responsive to evolving human preferences. Experimental results on several benchmarks show that our approach consistently outperforms vanilla DPO by 5.9-8.8 points on AlpacaEval 2 and 3.3-9.7 points on Arena-Hard across different model architectures and sizes. Furthermore, additional experiments on mathematical and reasoning benchmarks (MMLU, GSM8K, and MATH) confirm that our method enhances performance without compromising general capabilities. Our codebase would be available at \url{https://github.com/LotuSrc/D2PO}. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14340v1-abstract-full').style.display = 'none'; document.getElementById('2502.14340v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 20 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">Accepted by ICLR 2025</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.14287">arXiv:2502.14287</a> <span> [<a href="https://arxiv.org/pdf/2502.14287">pdf</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Soft Condensed Matter">cond-mat.soft</span> </div> </div> <p class="title is-5 mathjax"> Nearly Complete Segregation of Submerged Grains in a Rotating Drum </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Chen%2C+Y">Yu Chen</a>, <a href="/search/?searchtype=author&query=Wei%2C+D">Deheng Wei</a>, <a href="/search/?searchtype=author&query=Suo%2C+S">Si Suo</a>, <a href="/search/?searchtype=author&query=Dong%2C+M">Mingrui Dong</a>, <a href="/search/?searchtype=author&query=Gan%2C+Y">Yixiang Gan</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.14287v1-abstract-short" style="display: inline;"> Density-driven segregations, extensively studied in a simple rotating drum, are enriched with a wide range of underlying physics. Diverse symmetrical segregation patterns formed by mixing two types of dry mono-sized grains have been revealed due to variations in heavy and light grain densities, $蟻_h$ and $蟻_l$, and rotating speeds, $蠅$. We engender experimentally a nearly complete segregation, not… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14287v1-abstract-full').style.display = 'inline'; document.getElementById('2502.14287v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.14287v1-abstract-full" style="display: none;"> Density-driven segregations, extensively studied in a simple rotating drum, are enriched with a wide range of underlying physics. Diverse symmetrical segregation patterns formed by mixing two types of dry mono-sized grains have been revealed due to variations in heavy and light grain densities, $蟻_h$ and $蟻_l$, and rotating speeds, $蠅$. We engender experimentally a nearly complete segregation, not occurring in dry conditions of the same $蟻_h$, $蟻_l$, and $蠅$, in submerged states. Further, based on the experiment-validated simulations, using coupled computational fluid dynamics and discrete element method, it is found the mixing index can be well predicted over a wide parameter space in the effective density ratio, $D=(蟻_h-蟻_f)/(蟻_l-蟻_f)$ with $蟻_f$ being the fluid density. Specifically, with increasing $D$ well-mixed states transit to fully-segregated states with a rising number of vortices and severer asymmetrical patterns. When the global Reynolds number $\mathrm{Re}_g$ is enlarged, the vortex area of heavy particles shrinks for lower $D$, while the area of light particles gradually saturates; meanwhile, for higher $D$ a new vortex with a continuously expanded area can be encountered in the light particle zone. These results improve our understanding of segregation transitions especially in submerged granular systems and shed new light on various science and engineering practices. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14287v1-abstract-full').style.display = 'none'; document.getElementById('2502.14287v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 20 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">12 pages, 4 figures</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.14068">arXiv:2502.14068</a> <span> [<a href="https://arxiv.org/pdf/2502.14068">pdf</a>, <a href="https://arxiv.org/format/2502.14068">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computer Vision and Pattern Recognition">cs.CV</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Artificial Intelligence">cs.AI</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Image and Video Processing">eess.IV</span> </div> </div> <p class="title is-5 mathjax"> A Racing Dataset and Baseline Model for Track Detection in Autonomous Racing </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Ghosh%2C+S">Shreya Ghosh</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yi-Huan Chen</a>, <a href="/search/?searchtype=author&query=Huang%2C+C">Ching-Hsiang Huang</a>, <a href="/search/?searchtype=author&query=Jameel%2C+A+S+M+M">Abu Shafin Mohammad Mahdee Jameel</a>, <a href="/search/?searchtype=author&query=Ho%2C+C+C">Chien Chou Ho</a>, <a href="/search/?searchtype=author&query=Gamal%2C+A+E">Aly El Gamal</a>, <a href="/search/?searchtype=author&query=Labi%2C+S">Samuel Labi</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.14068v1-abstract-short" style="display: inline;"> A significant challenge in racing-related research is the lack of publicly available datasets containing raw images with corresponding annotations for the downstream task. In this paper, we introduce RoRaTrack, a novel dataset that contains annotated multi-camera image data from racing scenarios for track detection. The data is collected on a Dallara AV-21 at a racing circuit in Indiana, in collab… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14068v1-abstract-full').style.display = 'inline'; document.getElementById('2502.14068v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.14068v1-abstract-full" style="display: none;"> A significant challenge in racing-related research is the lack of publicly available datasets containing raw images with corresponding annotations for the downstream task. In this paper, we introduce RoRaTrack, a novel dataset that contains annotated multi-camera image data from racing scenarios for track detection. The data is collected on a Dallara AV-21 at a racing circuit in Indiana, in collaboration with the Indy Autonomous Challenge (IAC). RoRaTrack addresses common problems such as blurriness due to high speed, color inversion from the camera, and absence of lane markings on the track. Consequently, we propose RaceGAN, a baseline model based on a Generative Adversarial Network (GAN) that effectively addresses these challenges. The proposed model demonstrates superior performance compared to current state-of-the-art machine learning models in track detection. The dataset and code for this work are available at github.com/RaceGAN. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14068v1-abstract-full').style.display = 'none'; document.getElementById('2502.14068v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 19 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">Currently Under Review</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.14012">arXiv:2502.14012</a> <span> [<a href="https://arxiv.org/pdf/2502.14012">pdf</a>, <a href="https://arxiv.org/format/2502.14012">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Other Computer Science">cs.OH</span> </div> </div> <p class="title is-5 mathjax"> A double-layer placement algorithm for integrated circuit-based modules on printed circuit board </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Li%2C+H">Hangyuan Li</a>, <a href="/search/?searchtype=author&query=Yang%2C+Z">Zhaoyang Yang</a>, <a href="/search/?searchtype=author&query=Pang%2C+H">Haotian Pang</a>, <a href="/search/?searchtype=author&query=Xu%2C+N">Ning Xu</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yu Chen</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.14012v1-abstract-short" style="display: inline;"> Considering that the physical design of printed circuit board (PCB) follows the principle of modularized design, this paper proposes an automatic placement algorithm for functional modules. We first model the placement problem as a mixed-variable optimization problem, and then, developed tailored algorithms of global placement and legalization for the top-layer centralized placement subproblem and… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14012v1-abstract-full').style.display = 'inline'; document.getElementById('2502.14012v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.14012v1-abstract-full" style="display: none;"> Considering that the physical design of printed circuit board (PCB) follows the principle of modularized design, this paper proposes an automatic placement algorithm for functional modules. We first model the placement problem as a mixed-variable optimization problem, and then, developed tailored algorithms of global placement and legalization for the top-layer centralized placement subproblem and the bottom-layer pin-oriented placement subproblem. Numerical comparison demonstrates that the proposed mixed-variable optimization scheme can get optimized total wirelength of placement. Meanwhile, experimental results on several industrial PCB cases show that the developed centralized strategies can well accommodate the requirement of top-layer placement, and the pin-oriented global placement based on bin clustering contributes to optimized placement results meeting the requirement of pin-oriented design. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.14012v1-abstract-full').style.display = 'none'; document.getElementById('2502.14012v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 19 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.13979">arXiv:2502.13979</a> <span> [<a href="https://arxiv.org/pdf/2502.13979">pdf</a>, <a href="https://arxiv.org/format/2502.13979">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Risk Management">q-fin.RM</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Artificial Intelligence">cs.AI</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Machine Learning">cs.LG</span> </div> </div> <p class="title is-5 mathjax"> Utilizing Effective Dynamic Graph Learning to Shield Financial Stability from Risk Propagation </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Yu%2C+G">Guanyuan Yu</a>, <a href="/search/?searchtype=author&query=Li%2C+Q">Qing Li</a>, <a href="/search/?searchtype=author&query=Zhao%2C+Y">Yu Zhao</a>, <a href="/search/?searchtype=author&query=Wang%2C+J">Jun Wang</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">YiJun Chen</a>, <a href="/search/?searchtype=author&query=Chen%2C+S">Shaolei Chen</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.13979v1-abstract-short" style="display: inline;"> Financial risks can propagate across both tightly coupled temporal and spatial dimensions, posing significant threats to financial stability. Moreover, risks embedded in unlabeled data are often difficult to detect. To address these challenges, we introduce GraphShield, a novel approach with three key innovations: Enhanced Cross-Domain Infor mation Learning: We propose a dynamic graph learning mod… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13979v1-abstract-full').style.display = 'inline'; document.getElementById('2502.13979v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.13979v1-abstract-full" style="display: none;"> Financial risks can propagate across both tightly coupled temporal and spatial dimensions, posing significant threats to financial stability. Moreover, risks embedded in unlabeled data are often difficult to detect. To address these challenges, we introduce GraphShield, a novel approach with three key innovations: Enhanced Cross-Domain Infor mation Learning: We propose a dynamic graph learning module to improve information learning across temporal and spatial domains. Advanced Risk Recognition: By leveraging the clustering characteristics of risks, we construct a risk recognizing module to enhance the identification of hidden threats. Risk Propagation Visualization: We provide a visualization tool for quantifying and validating nodes that trigger widespread cascading risks. Extensive experiments on two real-world and two open-source datasets demonstrate the robust performance of our framework. Our approach represents a significant advancement in leveraging artificial intelligence to enhance financial stability, offering a powerful solution to mitigate the spread of risks within financial networks. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13979v1-abstract-full').style.display = 'none'; document.getElementById('2502.13979v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 18 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.13972">arXiv:2502.13972</a> <span> [<a href="https://arxiv.org/pdf/2502.13972">pdf</a>, <a href="https://arxiv.org/format/2502.13972">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Signal Processing">eess.SP</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Artificial Intelligence">cs.AI</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Machine Learning">cs.LG</span> </div> </div> <p class="title is-5 mathjax"> IncepFormerNet: A multi-scale multi-head attention network for SSVEP classification </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Huang%2C+Y">Yan Huang</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yongru Chen</a>, <a href="/search/?searchtype=author&query=Cao%2C+L">Lei Cao</a>, <a href="/search/?searchtype=author&query=Cao%2C+Y">Yongnian Cao</a>, <a href="/search/?searchtype=author&query=Yang%2C+X">Xuechun Yang</a>, <a href="/search/?searchtype=author&query=Dong%2C+Y">Yilin Dong</a>, <a href="/search/?searchtype=author&query=Liu%2C+T">Tianyu Liu</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.13972v1-abstract-short" style="display: inline;"> In recent years, deep learning (DL) models have shown outstanding performance in EEG classification tasks, particularly in Steady-State Visually Evoked Potential(SSVEP)-based Brain-Computer-Interfaces(BCI)systems. DL methods have been successfully applied to SSVEP-BCI. This study proposes a new model called IncepFormerNet, which is a hybrid of the Inception and Transformer architectures. IncepForm… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13972v1-abstract-full').style.display = 'inline'; document.getElementById('2502.13972v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.13972v1-abstract-full" style="display: none;"> In recent years, deep learning (DL) models have shown outstanding performance in EEG classification tasks, particularly in Steady-State Visually Evoked Potential(SSVEP)-based Brain-Computer-Interfaces(BCI)systems. DL methods have been successfully applied to SSVEP-BCI. This study proposes a new model called IncepFormerNet, which is a hybrid of the Inception and Transformer architectures. IncepFormerNet adeptly extracts multi-scale temporal information from time series data using parallel convolution kernels of varying sizes, accurately capturing the subtle variations and critical features within SSVEP signals.Furthermore, the model integrates the multi-head attention mechanism from the Transformer architecture, which not only provides insights into global dependencies but also significantly enhances the understanding and representation of complex patterns.Additionally, it takes advantage of filter bank techniques to extract features based on the spectral characteristics of SSVEP data. To validate the effectiveness of the proposed model, we conducted experiments on two public datasets, . The experimental results show that IncepFormerNet achieves an accuracy of 87.41 on Dataset 1 and 71.97 on Dataset 2 using a 1.0-second time window. To further verify the superiority of the proposed model, we compared it with other deep learning models, and the results indicate that our method achieves significantly higher accuracy than the others.The source codes in this work are available at: https://github.com/CECNL/SSVEP-DAN. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13972v1-abstract-full').style.display = 'none'; document.getElementById('2502.13972v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 4 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.13842">arXiv:2502.13842</a> <span> [<a href="https://arxiv.org/pdf/2502.13842">pdf</a>, <a href="https://arxiv.org/format/2502.13842">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computation and Language">cs.CL</span> </div> </div> <p class="title is-5 mathjax"> Inner Thinking Transformer: Leveraging Dynamic Depth Scaling to Foster Adaptive Internal Thinking </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Chen%2C+Y">Yilong Chen</a>, <a href="/search/?searchtype=author&query=Shang%2C+J">Junyuan Shang</a>, <a href="/search/?searchtype=author&query=Zhang%2C+Z">Zhenyu Zhang</a>, <a href="/search/?searchtype=author&query=Xie%2C+Y">Yanxi Xie</a>, <a href="/search/?searchtype=author&query=Sheng%2C+J">Jiawei Sheng</a>, <a href="/search/?searchtype=author&query=Liu%2C+T">Tingwen Liu</a>, <a href="/search/?searchtype=author&query=Wang%2C+S">Shuohuan Wang</a>, <a href="/search/?searchtype=author&query=Sun%2C+Y">Yu Sun</a>, <a href="/search/?searchtype=author&query=Wu%2C+H">Hua Wu</a>, <a href="/search/?searchtype=author&query=Wang%2C+H">Haifeng Wang</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.13842v1-abstract-short" style="display: inline;"> Large language models (LLMs) face inherent performance bottlenecks under parameter constraints, particularly in processing critical tokens that demand complex reasoning. Empirical analysis reveals challenging tokens induce abrupt gradient spikes across layers, exposing architectural stress points in standard Transformers. Building on this insight, we propose Inner Thinking Transformer (ITT), which… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13842v1-abstract-full').style.display = 'inline'; document.getElementById('2502.13842v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.13842v1-abstract-full" style="display: none;"> Large language models (LLMs) face inherent performance bottlenecks under parameter constraints, particularly in processing critical tokens that demand complex reasoning. Empirical analysis reveals challenging tokens induce abrupt gradient spikes across layers, exposing architectural stress points in standard Transformers. Building on this insight, we propose Inner Thinking Transformer (ITT), which reimagines layer computations as implicit thinking steps. ITT dynamically allocates computation through Adaptive Token Routing, iteratively refines representations via Residual Thinking Connections, and distinguishes reasoning phases using Thinking Step Encoding. ITT enables deeper processing of critical tokens without parameter expansion. Evaluations across 162M-466M parameter models show ITT achieves 96.5\% performance of a 466M Transformer using only 162M parameters, reduces training data by 43.2\%, and outperforms Transformer/Loop variants in 11 benchmarks. By enabling elastic computation allocation during inference, ITT balances performance and efficiency through architecture-aware optimization of implicit thinking pathways. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13842v1-abstract-full').style.display = 'none'; document.getElementById('2502.13842v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 19 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">15 pages, 11 figures</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.13841">arXiv:2502.13841</a> <span> [<a href="https://arxiv.org/pdf/2502.13841">pdf</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Applied Physics">physics.app-ph</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Materials Science">cond-mat.mtrl-sci</span> </div> </div> <p class="title is-5 mathjax"> Combined Light Excitation and Scanning Gate Microscopy on Heterostructure Nanowire Photovoltaic Devices </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Liu%2C+Y">Yen-Po Liu</a>, <a href="/search/?searchtype=author&query=Fast%2C+J">Jonatan Fast</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yang Chen</a>, <a href="/search/?searchtype=author&query=Zhe%2C+R">Ren Zhe</a>, <a href="/search/?searchtype=author&query=Burke%2C+A">Adam Burke</a>, <a href="/search/?searchtype=author&query=Timm%2C+R">Rainer Timm</a>, <a href="/search/?searchtype=author&query=Linke%2C+H">Heiner Linke</a>, <a href="/search/?searchtype=author&query=Mikkelsen%2C+A">Anders Mikkelsen</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.13841v1-abstract-short" style="display: inline;"> Nanoscale optoelectronic components achieve functionality via spatial variation in electronic structure induced by composition, defects, and dopants. To dynamically change the local band alignment and influence defect states, a scanning gate electrode is highly useful. However, this technique is rarely combined with photoexcitation by a controlled external light source. We explore a setup that com… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13841v1-abstract-full').style.display = 'inline'; document.getElementById('2502.13841v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.13841v1-abstract-full" style="display: none;"> Nanoscale optoelectronic components achieve functionality via spatial variation in electronic structure induced by composition, defects, and dopants. To dynamically change the local band alignment and influence defect states, a scanning gate electrode is highly useful. However, this technique is rarely combined with photoexcitation by a controlled external light source. We explore a setup that combines several types of light excitation with high resolution scanning gate and atomic force microscopy (SGM/AFM). We apply the technique to InAs nanowires with an atomic scale defined InP segment, that have attracted considerable attention for studies of hot carrier devices. Using AFM we image the topography of the nanowire device. SGM measurements without light excitation show how current profiles can be influenced by local gating near the InP segment. Modelling of the tip and nanowire can well predict the results based on the axial band structure variation and an asymmetric tip. SGM studies including light excitation are then performed using both a white light LED and laser diodes at 515 and 780nm. Both negative and positive photoconductance can be observed and the combined effect of light excitation and local gating is observed. SGM can then be used to discriminate between effects related to the wire axial compositional structure and surface states. The setup explored in the current work has significant advantages to study optoelectronics at realistic conditions and with rapid turnover. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13841v1-abstract-full').style.display = 'none'; document.getElementById('2502.13841v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 19 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.13707">arXiv:2502.13707</a> <span> [<a href="https://arxiv.org/pdf/2502.13707">pdf</a>, <a href="https://arxiv.org/format/2502.13707">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Robotics">cs.RO</span> </div> </div> <p class="title is-5 mathjax"> Human-Like Robot Impedance Regulation Skill Learning from Human-Human Demonstrations </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Li%2C+C">Chenzui Li</a>, <a href="/search/?searchtype=author&query=Wu%2C+X">Xi Wu</a>, <a href="/search/?searchtype=author&query=Liu%2C+J">Junjia Liu</a>, <a href="/search/?searchtype=author&query=Teng%2C+T">Tao Teng</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yiming Chen</a>, <a href="/search/?searchtype=author&query=Calinon%2C+S">Sylvain Calinon</a>, <a href="/search/?searchtype=author&query=Caldwell%2C+D">Darwin Caldwell</a>, <a href="/search/?searchtype=author&query=Chen%2C+F">Fei Chen</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.13707v1-abstract-short" style="display: inline;"> Humans are experts in collaborating with others physically by regulating compliance behaviors based on the perception of their partner states and the task requirements. Enabling robots to develop proficiency in human collaboration skills can facilitate more efficient human-robot collaboration (HRC). This paper introduces an innovative impedance regulation skill learning framework for achieving HRC… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13707v1-abstract-full').style.display = 'inline'; document.getElementById('2502.13707v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.13707v1-abstract-full" style="display: none;"> Humans are experts in collaborating with others physically by regulating compliance behaviors based on the perception of their partner states and the task requirements. Enabling robots to develop proficiency in human collaboration skills can facilitate more efficient human-robot collaboration (HRC). This paper introduces an innovative impedance regulation skill learning framework for achieving HRC in multiple physical collaborative tasks. The framework is designed to adjust the robot compliance to the human partner states while adhering to reference trajectories provided by human-human demonstrations. Specifically, electromyography (EMG) signals from human muscles are collected and analyzed to extract limb impedance, representing compliance behaviors during demonstrations. Human endpoint motions are captured and represented using a probabilistic learning method to create reference trajectories and corresponding impedance profiles. Meanwhile, an LSTMbased module is implemented to develop task-oriented impedance regulation policies by mapping the muscle synergistic contributions between two demonstrators. Finally, we propose a wholebody impedance controller for a human-like robot, coordinating joint outputs to achieve the desired impedance and reference trajectory during task execution. Experimental validation was conducted through a collaborative transportation task and two interactive Tai Chi pushing hands tasks, demonstrating superior performance from the perspective of interactive forces compared to a constant impedance control method. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13707v1-abstract-full').style.display = 'none'; document.getElementById('2502.13707v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 19 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">12 pages, 12 figures</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.13626">arXiv:2502.13626</a> <span> [<a href="https://arxiv.org/pdf/2502.13626">pdf</a>, <a href="https://arxiv.org/format/2502.13626">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computational Engineering, Finance, and Science">cs.CE</span> </div> </div> <p class="title is-5 mathjax"> AI-Empowered Catalyst Discovery: A Survey from Classical Machine Learning Approaches to Large Language Models </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Xu%2C+Y">Yuanyuan Xu</a>, <a href="/search/?searchtype=author&query=Wang%2C+H">Hanchen Wang</a>, <a href="/search/?searchtype=author&query=Zhang%2C+W">Wenjie Zhang</a>, <a href="/search/?searchtype=author&query=Xie%2C+L">Lexing Xie</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yin Chen</a>, <a href="/search/?searchtype=author&query=Salim%2C+F">Flora Salim</a>, <a href="/search/?searchtype=author&query=Zhang%2C+Y">Ying Zhang</a>, <a href="/search/?searchtype=author&query=Gooding%2C+J">Justin Gooding</a>, <a href="/search/?searchtype=author&query=Walsh%2C+T">Toby Walsh</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.13626v1-abstract-short" style="display: inline;"> Catalysts are essential for accelerating chemical reactions and enhancing selectivity, which is crucial for the sustainable production of energy, materials, and bioactive compounds. Catalyst discovery is fundamental yet challenging in computational chemistry and has garnered significant attention due to the promising performance of advanced Artificial Intelligence (AI) techniques. The development… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13626v1-abstract-full').style.display = 'inline'; document.getElementById('2502.13626v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.13626v1-abstract-full" style="display: none;"> Catalysts are essential for accelerating chemical reactions and enhancing selectivity, which is crucial for the sustainable production of energy, materials, and bioactive compounds. Catalyst discovery is fundamental yet challenging in computational chemistry and has garnered significant attention due to the promising performance of advanced Artificial Intelligence (AI) techniques. The development of Large Language Models (LLMs) notably accelerates progress in the discovery of both homogeneous and heterogeneous catalysts, where their chemical reactions differ significantly in material phases, temperature, dynamics, etc. However, there is currently no comprehensive survey that discusses the progress and latest developments in both areas, particularly with the application of LLM techniques. To address this gap, this paper presents a thorough and systematic survey of AI-empowered catalyst discovery, employing a unified and general categorization for homogeneous and heterogeneous catalysts. We examine the progress of AI-empowered catalyst discovery, highlighting their individual advantages and disadvantages, and discuss the challenges faced in this field. Furthermore, we suggest potential directions for future research from the perspective of computer science. Our goal is to assist researchers in computational chemistry, computer science, and related fields in easily tracking the latest advancements, providing a clear overview and roadmap of this area. We also organize and make accessible relevant resources, including article lists and datasets, in an open repository at https://github.com/LuckyGirl-XU/Awesome-Artificial-Intelligence-Empowered-Catalyst-Discovery. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13626v1-abstract-full').style.display = 'none'; document.getElementById('2502.13626v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 19 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.13540">arXiv:2502.13540</a> <span> [<a href="https://arxiv.org/pdf/2502.13540">pdf</a>, <a href="https://arxiv.org/format/2502.13540">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="High Energy Physics - Experiment">hep-ex</span> </div> </div> <p class="title is-5 mathjax"> Amplitude analysis of $蠄(3686)\to 纬K_S^0 K_S^0 $ </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=BESIII+Collaboration"> BESIII Collaboration</a>, <a href="/search/?searchtype=author&query=Ablikim%2C+M">M. Ablikim</a>, <a href="/search/?searchtype=author&query=Achasov%2C+M+N">M. N. Achasov</a>, <a href="/search/?searchtype=author&query=Adlarson%2C+P">P. Adlarson</a>, <a href="/search/?searchtype=author&query=Ai%2C+X+C">X. C. Ai</a>, <a href="/search/?searchtype=author&query=Aliberti%2C+R">R. Aliberti</a>, <a href="/search/?searchtype=author&query=Amoroso%2C+A">A. Amoroso</a>, <a href="/search/?searchtype=author&query=An%2C+Q">Q. An</a>, <a href="/search/?searchtype=author&query=Bai%2C+Y">Y. Bai</a>, <a href="/search/?searchtype=author&query=Bakina%2C+O">O. Bakina</a>, <a href="/search/?searchtype=author&query=Ban%2C+Y">Y. Ban</a>, <a href="/search/?searchtype=author&query=Bao%2C+H+-">H. -R. Bao</a>, <a href="/search/?searchtype=author&query=Batozskaya%2C+V">V. Batozskaya</a>, <a href="/search/?searchtype=author&query=Begzsuren%2C+K">K. Begzsuren</a>, <a href="/search/?searchtype=author&query=Berger%2C+N">N. Berger</a>, <a href="/search/?searchtype=author&query=Berlowski%2C+M">M. Berlowski</a>, <a href="/search/?searchtype=author&query=Bertani%2C+M">M. Bertani</a>, <a href="/search/?searchtype=author&query=Bettoni%2C+D">D. Bettoni</a>, <a href="/search/?searchtype=author&query=Bianchi%2C+F">F. Bianchi</a>, <a href="/search/?searchtype=author&query=Bianco%2C+E">E. Bianco</a>, <a href="/search/?searchtype=author&query=Bortone%2C+A">A. Bortone</a>, <a href="/search/?searchtype=author&query=Boyko%2C+I">I. Boyko</a>, <a href="/search/?searchtype=author&query=Briere%2C+R+A">R. A. Briere</a>, <a href="/search/?searchtype=author&query=Brueggemann%2C+A">A. Brueggemann</a>, <a href="/search/?searchtype=author&query=Cai%2C+H">H. Cai</a> , et al. (704 additional authors not shown) </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.13540v1-abstract-short" style="display: inline;"> Using $(2712\pm14)\times10^6$ $蠄(3686)$ events collected with the BESIII detector, we perform the first amplitude analysis of the radiative decay $蠄(3686)\to 纬K_S^0 K_S^0$ within the mass region $M_{K_S^0 K_S^0 }<2.8$ GeV/$c^2$. Employing a one-channel K-matrix approach for the description of the dynamics of the $K^0_S K^0_S$ system, the data sample is well described with four poles for the $f_0$-… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13540v1-abstract-full').style.display = 'inline'; document.getElementById('2502.13540v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.13540v1-abstract-full" style="display: none;"> Using $(2712\pm14)\times10^6$ $蠄(3686)$ events collected with the BESIII detector, we perform the first amplitude analysis of the radiative decay $蠄(3686)\to 纬K_S^0 K_S^0$ within the mass region $M_{K_S^0 K_S^0 }<2.8$ GeV/$c^2$. Employing a one-channel K-matrix approach for the description of the dynamics of the $K^0_S K^0_S$ system, the data sample is well described with four poles for the $f_0$-wave and three poles for the $f_2$-wave. The determined pole positions are consistent with those of well-established resonance states. The observed $f_0$ and $f_{2}$ states are found to be qualitatively consistent with those produced in radiative $J/蠄$ decays, indicating the similarity between the two charmonium states in their radiative decays. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13540v1-abstract-full').style.display = 'none'; document.getElementById('2502.13540v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 19 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">20 pages, 4 figures, submitted to JHEP</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.13487">arXiv:2502.13487</a> <span> [<a href="https://arxiv.org/pdf/2502.13487">pdf</a>, <a href="https://arxiv.org/format/2502.13487">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computation and Language">cs.CL</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Artificial Intelligence">cs.AI</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Computer Vision and Pattern Recognition">cs.CV</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Machine Learning">cs.LG</span> </div> </div> <p class="title is-5 mathjax"> Transferring Textual Preferences to Vision-Language Understanding through Model Merging </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Li%2C+C">Chen-An Li</a>, <a href="/search/?searchtype=author&query=Lin%2C+T">Tzu-Han Lin</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yun-Nung Chen</a>, <a href="/search/?searchtype=author&query=Lee%2C+H">Hung-yi Lee</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.13487v1-abstract-short" style="display: inline;"> Large vision-language models (LVLMs) perform outstandingly across various multimodal tasks. However, their ability to evaluate generated content remains limited, and training vision-language reward models (VLRMs) with preference data is computationally expensive. This paper explores a training-free alternative by merging text-based reward models (RMs) with LVLMs to create VLRMs. Our approach shows… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13487v1-abstract-full').style.display = 'inline'; document.getElementById('2502.13487v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.13487v1-abstract-full" style="display: none;"> Large vision-language models (LVLMs) perform outstandingly across various multimodal tasks. However, their ability to evaluate generated content remains limited, and training vision-language reward models (VLRMs) with preference data is computationally expensive. This paper explores a training-free alternative by merging text-based reward models (RMs) with LVLMs to create VLRMs. Our approach shows that integrating these models leads to improved performance over LVLMs' scoring and text-based RMs, offering an efficient method for incorporating textual preferences into LVLMs. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13487v1-abstract-full').style.display = 'none'; document.getElementById('2502.13487v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 19 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">Preprint. Under Review</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.13467">arXiv:2502.13467</a> <span> [<a href="https://arxiv.org/pdf/2502.13467">pdf</a>, <a href="https://arxiv.org/ps/2502.13467">ps</a>, <a href="https://arxiv.org/format/2502.13467">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Machine Learning">cs.LG</span> </div> </div> <p class="title is-5 mathjax"> Continuous K-Max Bandits </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Chen%2C+Y">Yu Chen</a>, <a href="/search/?searchtype=author&query=Wang%2C+S">Siwei Wang</a>, <a href="/search/?searchtype=author&query=Huang%2C+L">Longbo Huang</a>, <a href="/search/?searchtype=author&query=Chen%2C+W">Wei Chen</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.13467v1-abstract-short" style="display: inline;"> We study the $K$-Max combinatorial multi-armed bandits problem with continuous outcome distributions and weak value-index feedback: each base arm has an unknown continuous outcome distribution, and in each round the learning agent selects $K$ arms, obtains the maximum value sampled from these $K$ arms as reward and observes this reward together with the corresponding arm index as feedback. This se… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13467v1-abstract-full').style.display = 'inline'; document.getElementById('2502.13467v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.13467v1-abstract-full" style="display: none;"> We study the $K$-Max combinatorial multi-armed bandits problem with continuous outcome distributions and weak value-index feedback: each base arm has an unknown continuous outcome distribution, and in each round the learning agent selects $K$ arms, obtains the maximum value sampled from these $K$ arms as reward and observes this reward together with the corresponding arm index as feedback. This setting captures critical applications in recommendation systems, distributed computing, server scheduling, etc. The continuous $K$-Max bandits introduce unique challenges, including discretization error from continuous-to-discrete conversion, non-deterministic tie-breaking under limited feedback, and biased estimation due to partial observability. Our key contribution is the computationally efficient algorithm DCK-UCB, which combines adaptive discretization with bias-corrected confidence bounds to tackle these challenges. For general continuous distributions, we prove that DCK-UCB achieves a $\widetilde{\mathcal{O}}(T^{3/4})$ regret upper bound, establishing the first sublinear regret guarantee for this setting. Furthermore, we identify an important special case with exponential distributions under full-bandit feedback. In this case, our proposed algorithm MLE-Exp enables $\widetilde{\mathcal{O}}(\sqrt{T})$ regret upper bound through maximal log-likelihood estimation, achieving near-minimax optimality. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13467v1-abstract-full').style.display = 'none'; document.getElementById('2502.13467v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 19 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.13443">arXiv:2502.13443</a> <span> [<a href="https://arxiv.org/pdf/2502.13443">pdf</a>, <a href="https://arxiv.org/format/2502.13443">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Robotics">cs.RO</span> </div> </div> <p class="title is-5 mathjax"> Physics-Aware Robotic Palletization with Online Masking Inference </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Zhang%2C+T">Tianqi Zhang</a>, <a href="/search/?searchtype=author&query=Wu%2C+Z">Zheng Wu</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yuxin Chen</a>, <a href="/search/?searchtype=author&query=Wang%2C+Y">Yixiao Wang</a>, <a href="/search/?searchtype=author&query=Liang%2C+B">Boyuan Liang</a>, <a href="/search/?searchtype=author&query=Moura%2C+S">Scott Moura</a>, <a href="/search/?searchtype=author&query=Tomizuka%2C+M">Masayoshi Tomizuka</a>, <a href="/search/?searchtype=author&query=Ding%2C+M">Mingyu Ding</a>, <a href="/search/?searchtype=author&query=Zhan%2C+W">Wei Zhan</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.13443v1-abstract-short" style="display: inline;"> The efficient planning of stacking boxes, especially in the online setting where the sequence of item arrivals is unpredictable, remains a critical challenge in modern warehouse and logistics management. Existing solutions often address box size variations, but overlook their intrinsic and physical properties, such as density and rigidity, which are crucial for real-world applications. We use rein… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13443v1-abstract-full').style.display = 'inline'; document.getElementById('2502.13443v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.13443v1-abstract-full" style="display: none;"> The efficient planning of stacking boxes, especially in the online setting where the sequence of item arrivals is unpredictable, remains a critical challenge in modern warehouse and logistics management. Existing solutions often address box size variations, but overlook their intrinsic and physical properties, such as density and rigidity, which are crucial for real-world applications. We use reinforcement learning (RL) to solve this problem by employing action space masking to direct the RL policy toward valid actions. Unlike previous methods that rely on heuristic stability assessments which are difficult to assess in physical scenarios, our framework utilizes online learning to dynamically train the action space mask, eliminating the need for manual heuristic design. Extensive experiments demonstrate that our proposed method outperforms existing state-of-the-arts. Furthermore, we deploy our learned task planner in a real-world robotic palletizer, validating its practical applicability in operational settings. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13443v1-abstract-full').style.display = 'none'; document.getElementById('2502.13443v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 19 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">Accepted by ICRA 2025</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.13410">arXiv:2502.13410</a> <span> [<a href="https://arxiv.org/pdf/2502.13410">pdf</a>, <a href="https://arxiv.org/format/2502.13410">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computer Science and Game Theory">cs.GT</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Artificial Intelligence">cs.AI</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Theoretical Economics">econ.TH</span> </div> </div> <p class="title is-5 mathjax"> Tell Me Why: Incentivizing Explanations </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Srinivasan%2C+S">Siddarth Srinivasan</a>, <a href="/search/?searchtype=author&query=Karger%2C+E">Ezra Karger</a>, <a href="/search/?searchtype=author&query=Bakker%2C+M">Michiel Bakker</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yiling Chen</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.13410v1-abstract-short" style="display: inline;"> Common sense suggests that when individuals explain why they believe something, we can arrive at more accurate conclusions than when they simply state what they believe. Yet, there is no known mechanism that provides incentives to elicit explanations for beliefs from agents. This likely stems from the fact that standard Bayesian models make assumptions (like conditional independence of signals) th… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13410v1-abstract-full').style.display = 'inline'; document.getElementById('2502.13410v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.13410v1-abstract-full" style="display: none;"> Common sense suggests that when individuals explain why they believe something, we can arrive at more accurate conclusions than when they simply state what they believe. Yet, there is no known mechanism that provides incentives to elicit explanations for beliefs from agents. This likely stems from the fact that standard Bayesian models make assumptions (like conditional independence of signals) that preempt the need for explanations, in order to show efficient information aggregation. A natural justification for the value of explanations is that agents' beliefs tend to be drawn from overlapping sources of information, so agents' belief reports do not reveal all that needs to be known. Indeed, this work argues that rationales-explanations of an agent's private information-lead to more efficient aggregation by allowing agents to efficiently identify what information they share and what information is new. Building on this model of rationales, we present a novel 'deliberation mechanism' to elicit rationales from agents in which truthful reporting of beliefs and rationales is a perfect Bayesian equilibrium. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13410v1-abstract-full').style.display = 'none'; document.getElementById('2502.13410v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 18 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.13370">arXiv:2502.13370</a> <span> [<a href="https://arxiv.org/pdf/2502.13370">pdf</a>, <a href="https://arxiv.org/format/2502.13370">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Machine Learning">cs.LG</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Numerical Analysis">math.NA</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Quantum Physics">quant-ph</span> </div> </div> <p class="title is-5 mathjax"> Quantum Recurrent Neural Networks with Encoder-Decoder for Time-Dependent Partial Differential Equations </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Chen%2C+Y">Yuan Chen</a>, <a href="/search/?searchtype=author&query=Khaliq%2C+A">Abdul Khaliq</a>, <a href="/search/?searchtype=author&query=Furati%2C+K+M">Khaled M. Furati</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.13370v1-abstract-short" style="display: inline;"> Nonlinear time-dependent partial differential equations are essential in modeling complex phenomena across diverse fields, yet they pose significant challenges due to their computational complexity, especially in higher dimensions. This study explores Quantum Recurrent Neural Networks within an encoder-decoder framework, integrating Variational Quantum Circuits into Gated Recurrent Units and Long… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13370v1-abstract-full').style.display = 'inline'; document.getElementById('2502.13370v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.13370v1-abstract-full" style="display: none;"> Nonlinear time-dependent partial differential equations are essential in modeling complex phenomena across diverse fields, yet they pose significant challenges due to their computational complexity, especially in higher dimensions. This study explores Quantum Recurrent Neural Networks within an encoder-decoder framework, integrating Variational Quantum Circuits into Gated Recurrent Units and Long Short-Term Memory networks. Using this architecture, the model efficiently compresses high-dimensional spatiotemporal data into a compact latent space, facilitating more efficient temporal evolution. We evaluate the algorithms on the Hamilton-Jacobi-Bellman equation, Burgers' equation, the Gray-Scott reaction-diffusion system, and the three dimensional Michaelis-Menten reaction-diffusion equation. The results demonstrate the superior performance of the quantum-based algorithms in capturing nonlinear dynamics, handling high-dimensional spaces, and providing stable solutions, highlighting their potential as an innovative tool in solving challenging and complex systems. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13370v1-abstract-full').style.display = 'none'; document.getElementById('2502.13370v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 18 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.13252">arXiv:2502.13252</a> <span> [<a href="https://arxiv.org/pdf/2502.13252">pdf</a>, <a href="https://arxiv.org/format/2502.13252">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computation and Language">cs.CL</span> </div> </div> <p class="title is-5 mathjax"> Multilingual Language Model Pretraining using Machine-translated Data </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Wang%2C+J">Jiayi Wang</a>, <a href="/search/?searchtype=author&query=Lu%2C+Y">Yao Lu</a>, <a href="/search/?searchtype=author&query=Weber%2C+M">Maurice Weber</a>, <a href="/search/?searchtype=author&query=Ryabinin%2C+M">Max Ryabinin</a>, <a href="/search/?searchtype=author&query=Adelani%2C+D">David Adelani</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yihong Chen</a>, <a href="/search/?searchtype=author&query=Tang%2C+R">Raphael Tang</a>, <a href="/search/?searchtype=author&query=Stenetorp%2C+P">Pontus Stenetorp</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.13252v1-abstract-short" style="display: inline;"> High-resource languages such as English, enables the pretraining of high-quality large language models (LLMs). The same can not be said for most other languages as LLMs still underperform for non-English languages, likely due to a gap in the quality and diversity of the available multilingual pretraining corpora. In this work, we find that machine-translated texts from a single high-quality source… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13252v1-abstract-full').style.display = 'inline'; document.getElementById('2502.13252v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.13252v1-abstract-full" style="display: none;"> High-resource languages such as English, enables the pretraining of high-quality large language models (LLMs). The same can not be said for most other languages as LLMs still underperform for non-English languages, likely due to a gap in the quality and diversity of the available multilingual pretraining corpora. In this work, we find that machine-translated texts from a single high-quality source language can contribute significantly to the pretraining quality of multilingual LLMs. We translate FineWeb-Edu, a high-quality English web dataset, into nine languages, resulting in a 1.7-trillion-token dataset, which we call TransWebEdu and pretrain a 1.3B-parameter model, TransWebLLM, from scratch on this dataset. Across nine non-English reasoning tasks, we show that TransWebLLM matches or outperforms state-of-the-art multilingual models trained using closed data, such as Llama3.2, Qwen2.5, and Gemma, despite using an order of magnitude less data. We demonstrate that adding less than 5% of TransWebEdu as domain-specific pretraining data sets a new state-of-the-art in Arabic, Italian, Indonesian, Swahili, and Welsh understanding and commonsense reasoning tasks. To promote reproducibility, we release our corpus, models, and training pipeline under Open Source Initiative-approved licenses. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13252v1-abstract-full').style.display = 'none'; document.getElementById('2502.13252v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 18 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.13193">arXiv:2502.13193</a> <span> [<a href="https://arxiv.org/pdf/2502.13193">pdf</a>, <a href="https://arxiv.org/format/2502.13193">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computation and Language">cs.CL</span> </div> </div> <p class="title is-5 mathjax"> Private Text Generation by Seeding Large Language Model Prompts </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Nagesh%2C+S">Supriya Nagesh</a>, <a href="/search/?searchtype=author&query=Chen%2C+J+Y">Justin Y. Chen</a>, <a href="/search/?searchtype=author&query=Mishra%2C+N">Nina Mishra</a>, <a href="/search/?searchtype=author&query=Wagner%2C+T">Tal Wagner</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.13193v1-abstract-short" style="display: inline;"> We explore how private synthetic text can be generated by suitably prompting a large language model (LLM). This addresses a challenge for organizations like hospitals, which hold sensitive text data like patient medical records, and wish to share it in order to train machine learning models for medical tasks, while preserving patient privacy. Methods that rely on training or finetuning a model may… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13193v1-abstract-full').style.display = 'inline'; document.getElementById('2502.13193v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.13193v1-abstract-full" style="display: none;"> We explore how private synthetic text can be generated by suitably prompting a large language model (LLM). This addresses a challenge for organizations like hospitals, which hold sensitive text data like patient medical records, and wish to share it in order to train machine learning models for medical tasks, while preserving patient privacy. Methods that rely on training or finetuning a model may be out of reach, either due to API limits of third-party LLMs, or due to ethical and legal prohibitions on sharing the private data with the LLM itself. We propose Differentially Private Keyphrase Prompt Seeding (DP-KPS), a method that generates a private synthetic text corpus from a sensitive input corpus, by accessing an LLM only through privatized prompts. It is based on seeding the prompts with private samples from a distribution over phrase embeddings, thus capturing the input corpus while achieving requisite output diversity and maintaining differential privacy. We evaluate DP-KPS on downstream ML text classification tasks, and show that the corpora it generates preserve much of the predictive power of the original ones. Our findings offer hope that institutions can reap ML insights by privately sharing data with simple prompts and little compute. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13193v1-abstract-full').style.display = 'none'; document.getElementById('2502.13193v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 18 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.13189">arXiv:2502.13189</a> <span> [<a href="https://arxiv.org/pdf/2502.13189">pdf</a>, <a href="https://arxiv.org/format/2502.13189">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Machine Learning">cs.LG</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Artificial Intelligence">cs.AI</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Computation and Language">cs.CL</span> </div> </div> <p class="title is-5 mathjax"> MoBA: Mixture of Block Attention for Long-Context LLMs </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Lu%2C+E">Enzhe Lu</a>, <a href="/search/?searchtype=author&query=Jiang%2C+Z">Zhejun Jiang</a>, <a href="/search/?searchtype=author&query=Liu%2C+J">Jingyuan Liu</a>, <a href="/search/?searchtype=author&query=Du%2C+Y">Yulun Du</a>, <a href="/search/?searchtype=author&query=Jiang%2C+T">Tao Jiang</a>, <a href="/search/?searchtype=author&query=Hong%2C+C">Chao Hong</a>, <a href="/search/?searchtype=author&query=Liu%2C+S">Shaowei Liu</a>, <a href="/search/?searchtype=author&query=He%2C+W">Weiran He</a>, <a href="/search/?searchtype=author&query=Yuan%2C+E">Enming Yuan</a>, <a href="/search/?searchtype=author&query=Wang%2C+Y">Yuzhi Wang</a>, <a href="/search/?searchtype=author&query=Huang%2C+Z">Zhiqi Huang</a>, <a href="/search/?searchtype=author&query=Yuan%2C+H">Huan Yuan</a>, <a href="/search/?searchtype=author&query=Xu%2C+S">Suting Xu</a>, <a href="/search/?searchtype=author&query=Xu%2C+X">Xinran Xu</a>, <a href="/search/?searchtype=author&query=Lai%2C+G">Guokun Lai</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yanru Chen</a>, <a href="/search/?searchtype=author&query=Zheng%2C+H">Huabin Zheng</a>, <a href="/search/?searchtype=author&query=Yan%2C+J">Junjie Yan</a>, <a href="/search/?searchtype=author&query=Su%2C+J">Jianlin Su</a>, <a href="/search/?searchtype=author&query=Wu%2C+Y">Yuxin Wu</a>, <a href="/search/?searchtype=author&query=Zhang%2C+N+Y">Neo Y. Zhang</a>, <a href="/search/?searchtype=author&query=Yang%2C+Z">Zhilin Yang</a>, <a href="/search/?searchtype=author&query=Zhou%2C+X">Xinyu Zhou</a>, <a href="/search/?searchtype=author&query=Zhang%2C+M">Mingxing Zhang</a>, <a href="/search/?searchtype=author&query=Qiu%2C+J">Jiezhong Qiu</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.13189v1-abstract-short" style="display: inline;"> Scaling the effective context length is essential for advancing large language models (LLMs) toward artificial general intelligence (AGI). However, the quadratic increase in computational complexity inherent in traditional attention mechanisms presents a prohibitive overhead. Existing approaches either impose strongly biased structures, such as sink or window attention which are task-specific, or… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13189v1-abstract-full').style.display = 'inline'; document.getElementById('2502.13189v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.13189v1-abstract-full" style="display: none;"> Scaling the effective context length is essential for advancing large language models (LLMs) toward artificial general intelligence (AGI). However, the quadratic increase in computational complexity inherent in traditional attention mechanisms presents a prohibitive overhead. Existing approaches either impose strongly biased structures, such as sink or window attention which are task-specific, or radically modify the attention mechanism into linear approximations, whose performance in complex reasoning tasks remains inadequately explored. In this work, we propose a solution that adheres to the ``less structure'' principle, allowing the model to determine where to attend autonomously, rather than introducing predefined biases. We introduce Mixture of Block Attention (MoBA), an innovative approach that applies the principles of Mixture of Experts (MoE) to the attention mechanism. This novel architecture demonstrates superior performance on long-context tasks while offering a key advantage: the ability to seamlessly transition between full and sparse attention, enhancing efficiency without the risk of compromising performance. MoBA has already been deployed to support Kimi's long-context requests and demonstrates significant advancements in efficient attention computation for LLMs. Our code is available at https://github.com/MoonshotAI/MoBA. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13189v1-abstract-full').style.display = 'none'; document.getElementById('2502.13189v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 18 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">15 pages</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.13053">arXiv:2502.13053</a> <span> [<a href="https://arxiv.org/pdf/2502.13053">pdf</a>, <a href="https://arxiv.org/format/2502.13053">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computation and Language">cs.CL</span> </div> </div> <p class="title is-5 mathjax"> AEIA-MN: Evaluating the Robustness of Multimodal LLM-Powered Mobile Agents Against Active Environmental Injection Attacks </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Chen%2C+Y">Yurun Chen</a>, <a href="/search/?searchtype=author&query=Hu%2C+X">Xueyu Hu</a>, <a href="/search/?searchtype=author&query=Yin%2C+K">Keting Yin</a>, <a href="/search/?searchtype=author&query=Li%2C+J">Juncheng Li</a>, <a href="/search/?searchtype=author&query=Zhang%2C+S">Shengyu Zhang</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.13053v1-abstract-short" style="display: inline;"> As researchers continuously optimize AI agents to perform tasks more effectively within operating systems, they often neglect to address the critical need for enabling these agents to identify "impostors" within the system. Through an analysis of the agents' operating environment, we identified a potential threat: attackers can disguise their attack methods as environmental elements, injecting act… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13053v1-abstract-full').style.display = 'inline'; document.getElementById('2502.13053v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.13053v1-abstract-full" style="display: none;"> As researchers continuously optimize AI agents to perform tasks more effectively within operating systems, they often neglect to address the critical need for enabling these agents to identify "impostors" within the system. Through an analysis of the agents' operating environment, we identified a potential threat: attackers can disguise their attack methods as environmental elements, injecting active disturbances into the agents' execution process, thereby disrupting their decision-making. We define this type of attack as Active Environment Injection Attack (AEIA). Based on this, we propose AEIA-MN, an active environment injection attack scheme that exploits interaction vulnerabilities in the mobile operating system to evaluate the robustness of MLLM-based agents against such threats. Experimental results show that even advanced MLLMs are highly vulnerable to this attack, achieving a maximum attack success rate of 93% in the AndroidWorld benchmark. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.13053v1-abstract-full').style.display = 'none'; document.getElementById('2502.13053v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 18 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.12987">arXiv:2502.12987</a> <span> [<a href="https://arxiv.org/pdf/2502.12987">pdf</a>, <a href="https://arxiv.org/format/2502.12987">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Machine Learning">cs.LG</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Atmospheric and Oceanic Physics">physics.ao-ph</span> </div> </div> <p class="title is-5 mathjax"> Ensemble Kalman filter in latent space using a variational autoencoder pair </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Pasmans%2C+I">Ivo Pasmans</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yumeng Chen</a>, <a href="/search/?searchtype=author&query=Finn%2C+T+S">Tobias Sebastian Finn</a>, <a href="/search/?searchtype=author&query=Bocquet%2C+M">Marc Bocquet</a>, <a href="/search/?searchtype=author&query=Carrassi%2C+A">Alberto Carrassi</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.12987v1-abstract-short" style="display: inline;"> Popular (ensemble) Kalman filter data assimilation (DA) approaches assume that the errors in both the a priori estimate of the state and those in the observations are Gaussian. For constrained variables, e.g. sea ice concentration or stress, such an assumption does not hold. The variational autoencoder (VAE) is a machine learning (ML) technique that allows to map an arbitrary distribution to/from… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12987v1-abstract-full').style.display = 'inline'; document.getElementById('2502.12987v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.12987v1-abstract-full" style="display: none;"> Popular (ensemble) Kalman filter data assimilation (DA) approaches assume that the errors in both the a priori estimate of the state and those in the observations are Gaussian. For constrained variables, e.g. sea ice concentration or stress, such an assumption does not hold. The variational autoencoder (VAE) is a machine learning (ML) technique that allows to map an arbitrary distribution to/from a latent space in which the distribution is supposedly closer to a Gaussian. We propose a novel hybrid DA-ML approach in which VAEs are incorporated in the DA procedure. Specifically, we introduce a variant of the popular ensemble transform Kalman filter (ETKF) in which the analysis is applied in the latent space of a single VAE or a pair of VAEs. In twin experiments with a simple circular model, whereby the circle represents an underlying submanifold to be respected, we find that the use of a VAE ensures that a posteri ensemble members lie close to the manifold containing the truth. Furthermore, online updating of the VAE is necessary and achievable when this manifold varies in time, i.e. when it is non-stationary. We demonstrate that introducing an additional second latent space for the observational innovations improves robustness against detrimental effects of non-Gaussianity and bias in the observational errors but it slightly lessens the performance if observational errors are strictly Gaussian. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12987v1-abstract-full').style.display = 'none'; document.getElementById('2502.12987v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 18 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.12963">arXiv:2502.12963</a> <span> [<a href="https://arxiv.org/pdf/2502.12963">pdf</a>, <a href="https://arxiv.org/format/2502.12963">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Robotics">cs.RO</span> </div> </div> <p class="title is-5 mathjax"> D3-ARM: High-Dynamic, Dexterous and Fully Decoupled Cable-driven Robotic Arm </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Luo%2C+H">Hong Luo</a>, <a href="/search/?searchtype=author&query=Xu%2C+J">Jianle Xu</a>, <a href="/search/?searchtype=author&query=Li%2C+S">Shoujie Li</a>, <a href="/search/?searchtype=author&query=Liang%2C+H">Huayue Liang</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yanbo Chen</a>, <a href="/search/?searchtype=author&query=Xia%2C+C">Chongkun Xia</a>, <a href="/search/?searchtype=author&query=Wang%2C+X">Xueqian Wang</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.12963v1-abstract-short" style="display: inline;"> Cable transmission enables motors of robotic arm to operate lightweight and low-inertia joints remotely in various environments, but it also creates issues with motion coupling and cable routing that can reduce arm's control precision and performance. In this paper, we present a novel motion decoupling mechanism with low-friction to align the cables and efficiently transmit the motor's power. By a… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12963v1-abstract-full').style.display = 'inline'; document.getElementById('2502.12963v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.12963v1-abstract-full" style="display: none;"> Cable transmission enables motors of robotic arm to operate lightweight and low-inertia joints remotely in various environments, but it also creates issues with motion coupling and cable routing that can reduce arm's control precision and performance. In this paper, we present a novel motion decoupling mechanism with low-friction to align the cables and efficiently transmit the motor's power. By arranging these mechanisms at the joints, we fabricate a fully decoupled and lightweight cable-driven robotic arm called D3-Arm with all the electrical components be placed at the base. Its 776 mm length moving part boasts six degrees of freedom (DOF) and only 1.6 kg weights. To address the issue of cable slack, a cable-pretension mechanism is integrated to enhance the stability of long-distance cable transmission. Through a series of comprehensive tests, D3-Arm demonstrated 1.29 mm average positioning error and 2.0 kg payload capacity, proving the practicality of the proposed decoupling mechanisms in cable-driven robotic arm. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12963v1-abstract-full').style.display = 'none'; document.getElementById('2502.12963v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 18 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.12893">arXiv:2502.12893</a> <span> [<a href="https://arxiv.org/pdf/2502.12893">pdf</a>, <a href="https://arxiv.org/format/2502.12893">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computation and Language">cs.CL</span> </div> </div> <p class="title is-5 mathjax"> H-CoT: Hijacking the Chain-of-Thought Safety Reasoning Mechanism to Jailbreak Large Reasoning Models, Including OpenAI o1/o3, DeepSeek-R1, and Gemini 2.0 Flash Thinking </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Kuo%2C+M">Martin Kuo</a>, <a href="/search/?searchtype=author&query=Zhang%2C+J">Jianyi Zhang</a>, <a href="/search/?searchtype=author&query=Ding%2C+A">Aolin Ding</a>, <a href="/search/?searchtype=author&query=Wang%2C+Q">Qinsi Wang</a>, <a href="/search/?searchtype=author&query=DiValentin%2C+L">Louis DiValentin</a>, <a href="/search/?searchtype=author&query=Bao%2C+Y">Yujia Bao</a>, <a href="/search/?searchtype=author&query=Wei%2C+W">Wei Wei</a>, <a href="/search/?searchtype=author&query=Juan%2C+D">Da-Cheng Juan</a>, <a href="/search/?searchtype=author&query=Li%2C+H">Hai Li</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yiran Chen</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.12893v1-abstract-short" style="display: inline;"> Large Reasoning Models (LRMs) have recently extended their powerful reasoning capabilities to safety checks-using chain-of-thought reasoning to decide whether a request should be answered. While this new approach offers a promising route for balancing model utility and safety, its robustness remains underexplored. To address this gap, we introduce Malicious-Educator, a benchmark that disguises ext… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12893v1-abstract-full').style.display = 'inline'; document.getElementById('2502.12893v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.12893v1-abstract-full" style="display: none;"> Large Reasoning Models (LRMs) have recently extended their powerful reasoning capabilities to safety checks-using chain-of-thought reasoning to decide whether a request should be answered. While this new approach offers a promising route for balancing model utility and safety, its robustness remains underexplored. To address this gap, we introduce Malicious-Educator, a benchmark that disguises extremely dangerous or malicious requests beneath seemingly legitimate educational prompts. Our experiments reveal severe security flaws in popular commercial-grade LRMs, including OpenAI o1/o3, DeepSeek-R1, and Gemini 2.0 Flash Thinking. For instance, although OpenAI's o1 model initially maintains a high refusal rate of about 98%, subsequent model updates significantly compromise its safety; and attackers can easily extract criminal strategies from DeepSeek-R1 and Gemini 2.0 Flash Thinking without any additional tricks. To further highlight these vulnerabilities, we propose Hijacking Chain-of-Thought (H-CoT), a universal and transferable attack method that leverages the model's own displayed intermediate reasoning to jailbreak its safety reasoning mechanism. Under H-CoT, refusal rates sharply decline-dropping from 98% to below 2%-and, in some instances, even transform initially cautious tones into ones that are willing to provide harmful content. We hope these findings underscore the urgent need for more robust safety mechanisms to preserve the benefits of advanced reasoning capabilities without compromising ethical standards. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12893v1-abstract-full').style.display = 'none'; document.getElementById('2502.12893v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 18 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.12678">arXiv:2502.12678</a> <span> [<a href="https://arxiv.org/pdf/2502.12678">pdf</a>, <a href="https://arxiv.org/format/2502.12678">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Machine Learning">cs.LG</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Artificial Intelligence">cs.AI</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Computation and Language">cs.CL</span> </div> </div> <p class="title is-5 mathjax"> Multi-Step Alignment as Markov Games: An Optimistic Online Gradient Descent Approach with Convergence Guarantees </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Wu%2C+Y">Yongtao Wu</a>, <a href="/search/?searchtype=author&query=Viano%2C+L">Luca Viano</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yihang Chen</a>, <a href="/search/?searchtype=author&query=Zhu%2C+Z">Zhenyu Zhu</a>, <a href="/search/?searchtype=author&query=Antonakopoulos%2C+K">Kimon Antonakopoulos</a>, <a href="/search/?searchtype=author&query=Gu%2C+Q">Quanquan Gu</a>, <a href="/search/?searchtype=author&query=Cevher%2C+V">Volkan Cevher</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.12678v1-abstract-short" style="display: inline;"> Reinforcement Learning from Human Feedback (RLHF) has been highly successful in aligning large language models with human preferences. While prevalent methods like DPO have demonstrated strong performance, they frame interactions with the language model as a bandit problem, which limits their applicability in real-world scenarios where multi-turn conversations are common. Additionally, DPO relies… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12678v1-abstract-full').style.display = 'inline'; document.getElementById('2502.12678v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.12678v1-abstract-full" style="display: none;"> Reinforcement Learning from Human Feedback (RLHF) has been highly successful in aligning large language models with human preferences. While prevalent methods like DPO have demonstrated strong performance, they frame interactions with the language model as a bandit problem, which limits their applicability in real-world scenarios where multi-turn conversations are common. Additionally, DPO relies on the Bradley-Terry model assumption, which does not adequately capture the non-transitive nature of human preferences. In this paper, we address these challenges by modeling the alignment problem as a two-player constant-sum Markov game, where each player seeks to maximize their winning rate against the other across all steps of the conversation. Our approach Multi-step Preference Optimization (MPO) is built upon the natural actor-critic framework~\citep{peters2008natural}. We further develop OMPO based on the optimistic online gradient descent algorithm~\citep{rakhlin2013online,joulani17a}. Theoretically, we provide a rigorous analysis for both algorithms on convergence and show that OMPO requires $\mathcal{O}(蔚^{-1})$ policy updates to converge to an $蔚$-approximate Nash equilibrium. We also validate the effectiveness of our method on multi-turn conversations dataset and math reasoning dataset. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12678v1-abstract-full').style.display = 'none'; document.getElementById('2502.12678v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 18 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">Accepted as oral presentation in NeurIPS LanGame Workshop, revised from ICLR submission</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.12584">arXiv:2502.12584</a> <span> [<a href="https://arxiv.org/pdf/2502.12584">pdf</a>, <a href="https://arxiv.org/format/2502.12584">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Machine Learning">cs.LG</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Artificial Intelligence">cs.AI</span> </div> </div> <p class="title is-5 mathjax"> Enhancing Semi-supervised Learning with Noisy Zero-shot Pseudolabels </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Chung%2C+J">Jichan Chung</a>, <a href="/search/?searchtype=author&query=Chen%2C+I+Y">Irene Y. Chen</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.12584v1-abstract-short" style="display: inline;"> Semi-supervised learning (SSL) leverages limited labeled data alongside abundant unlabeled data to address labeling costs in machine learning. While recent foundation models enable zero-shot inference, attempts to integrate these capabilities into SSL through pseudo-labeling have shown mixed results due to unreliable zero-shot predictions. We present ZMT (Zero-Shot Multi-Task Learning), a framewor… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12584v1-abstract-full').style.display = 'inline'; document.getElementById('2502.12584v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.12584v1-abstract-full" style="display: none;"> Semi-supervised learning (SSL) leverages limited labeled data alongside abundant unlabeled data to address labeling costs in machine learning. While recent foundation models enable zero-shot inference, attempts to integrate these capabilities into SSL through pseudo-labeling have shown mixed results due to unreliable zero-shot predictions. We present ZMT (Zero-Shot Multi-Task Learning), a framework that jointly optimizes zero-shot pseudo-labels and unsupervised representation learning objectives from contemporary SSL approaches. Our method introduces a multi-task learning-based mechanism that incorporates pseudo-labels while ensuring robustness to varying pseudo-label quality. Experiments across 8 datasets in vision, language, and audio domains demonstrate that ZMT reduces error by up to 56% compared to traditional SSL methods, with particularly compelling results when pseudo-labels are noisy and unreliable. ZMT represents a significant step toward making semi-supervised learning more effective and accessible in resource-constrained environments. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12584v1-abstract-full').style.display = 'none'; document.getElementById('2502.12584v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 18 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">Under review for ICML 2025</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.12532">arXiv:2502.12532</a> <span> [<a href="https://arxiv.org/pdf/2502.12532">pdf</a>, <a href="https://arxiv.org/format/2502.12532">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Artificial Intelligence">cs.AI</span> </div> </div> <p class="title is-5 mathjax"> CityEQA: A Hierarchical LLM Agent on Embodied Question Answering Benchmark in City Space </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Zhao%2C+Y">Yong Zhao</a>, <a href="/search/?searchtype=author&query=Xu%2C+K">Kai Xu</a>, <a href="/search/?searchtype=author&query=Zhu%2C+Z">Zhengqiu Zhu</a>, <a href="/search/?searchtype=author&query=Hu%2C+Y">Yue Hu</a>, <a href="/search/?searchtype=author&query=Zheng%2C+Z">Zhiheng Zheng</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yingfeng Chen</a>, <a href="/search/?searchtype=author&query=Ji%2C+Y">Yatai Ji</a>, <a href="/search/?searchtype=author&query=Gao%2C+C">Chen Gao</a>, <a href="/search/?searchtype=author&query=Li%2C+Y">Yong Li</a>, <a href="/search/?searchtype=author&query=Huang%2C+J">Jincai Huang</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.12532v2-abstract-short" style="display: inline;"> Embodied Question Answering (EQA) has primarily focused on indoor environments, leaving the complexities of urban settings - spanning environment, action, and perception - largely unexplored. To bridge this gap, we introduce CityEQA, a new task where an embodied agent answers open-vocabulary questions through active exploration in dynamic city spaces. To support this task, we present CityEQA-EC, t… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12532v2-abstract-full').style.display = 'inline'; document.getElementById('2502.12532v2-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.12532v2-abstract-full" style="display: none;"> Embodied Question Answering (EQA) has primarily focused on indoor environments, leaving the complexities of urban settings - spanning environment, action, and perception - largely unexplored. To bridge this gap, we introduce CityEQA, a new task where an embodied agent answers open-vocabulary questions through active exploration in dynamic city spaces. To support this task, we present CityEQA-EC, the first benchmark dataset featuring 1,412 human-annotated tasks across six categories, grounded in a realistic 3D urban simulator. Moreover, we propose Planner-Manager-Actor (PMA), a novel agent tailored for CityEQA. PMA enables long-horizon planning and hierarchical task execution: the Planner breaks down the question answering into sub-tasks, the Manager maintains an object-centric cognitive map for spatial reasoning during the process control, and the specialized Actors handle navigation, exploration, and collection sub-tasks. Experiments demonstrate that PMA achieves 60.7% of human-level answering accuracy, significantly outperforming frontier-based baselines. While promising, the performance gap compared to humans highlights the need for enhanced visual reasoning in CityEQA. This work paves the way for future advancements in urban spatial intelligence. Dataset and code are available at https://github.com/BiluYong/CityEQA.git. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12532v2-abstract-full').style.display = 'none'; document.getElementById('2502.12532v2-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 20 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">v1</span> submitted 17 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.12483">arXiv:2502.12483</a> <span> [<a href="https://arxiv.org/pdf/2502.12483">pdf</a>, <a href="https://arxiv.org/format/2502.12483">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computation and Language">cs.CL</span> </div> </div> <p class="title is-5 mathjax"> The Knowledge Microscope: Features as Better Analytical Lenses than Neurons </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Chen%2C+Y">Yuheng Chen</a>, <a href="/search/?searchtype=author&query=Cao%2C+P">Pengfei Cao</a>, <a href="/search/?searchtype=author&query=Liu%2C+K">Kang Liu</a>, <a href="/search/?searchtype=author&query=Zhao%2C+J">Jun Zhao</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.12483v1-abstract-short" style="display: inline;"> Previous studies primarily utilize MLP neurons as units of analysis for understanding the mechanisms of factual knowledge in Language Models (LMs); however, neurons suffer from polysemanticity, leading to limited knowledge expression and poor interpretability. In this paper, we first conduct preliminary experiments to validate that Sparse Autoencoders (SAE) can effectively decompose neurons into f… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12483v1-abstract-full').style.display = 'inline'; document.getElementById('2502.12483v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.12483v1-abstract-full" style="display: none;"> Previous studies primarily utilize MLP neurons as units of analysis for understanding the mechanisms of factual knowledge in Language Models (LMs); however, neurons suffer from polysemanticity, leading to limited knowledge expression and poor interpretability. In this paper, we first conduct preliminary experiments to validate that Sparse Autoencoders (SAE) can effectively decompose neurons into features, which serve as alternative analytical units. With this established, our core findings reveal three key advantages of features over neurons: (1) Features exhibit stronger influence on knowledge expression and superior interpretability. (2) Features demonstrate enhanced monosemanticity, showing distinct activation patterns between related and unrelated facts. (3) Features achieve better privacy protection than neurons, demonstrated through our proposed FeatureEdit method, which significantly outperforms existing neuron-based approaches in erasing privacy-sensitive information from LMs.Code and dataset will be available. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12483v1-abstract-full').style.display = 'none'; document.getElementById('2502.12483v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 17 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">ARR February UnderReview</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.12413">arXiv:2502.12413</a> <span> [<a href="https://arxiv.org/pdf/2502.12413">pdf</a>, <a href="https://arxiv.org/format/2502.12413">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Machine Learning">cs.LG</span> </div> </div> <p class="title is-5 mathjax"> DivIL: Unveiling and Addressing Over-Invariance for Out-of- Distribution Generalization </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Wang%2C+J">Jiaqi Wang</a>, <a href="/search/?searchtype=author&query=Zhou%2C+Y">Yuhang Zhou</a>, <a href="/search/?searchtype=author&query=Zhang%2C+Z">Zhixiong Zhang</a>, <a href="/search/?searchtype=author&query=Chen%2C+Q">Qiguang Chen</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yongqiang Chen</a>, <a href="/search/?searchtype=author&query=Cheng%2C+J">James Cheng</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.12413v1-abstract-short" style="display: inline;"> Out-of-distribution generalization is a common problem that expects the model to perform well in the different distributions even far from the train data. A popular approach to addressing this issue is invariant learning (IL), in which the model is compiled to focus on invariant features instead of spurious features by adding strong constraints during training. However, there are some potential pi… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12413v1-abstract-full').style.display = 'inline'; document.getElementById('2502.12413v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.12413v1-abstract-full" style="display: none;"> Out-of-distribution generalization is a common problem that expects the model to perform well in the different distributions even far from the train data. A popular approach to addressing this issue is invariant learning (IL), in which the model is compiled to focus on invariant features instead of spurious features by adding strong constraints during training. However, there are some potential pitfalls of strong invariant constraints. Due to the limited number of diverse environments and over-regularization in the feature space, it may lead to a loss of important details in the invariant features while alleviating the spurious correlations, namely the over-invariance, which can also degrade the generalization performance. We theoretically define the over-invariance and observe that this issue occurs in various classic IL methods. To alleviate this issue, we propose a simple approach Diverse Invariant Learning (DivIL) by adding the unsupervised contrastive learning and the random masking mechanism compensatory for the invariant constraints, which can be applied to various IL methods. Furthermore, we conduct experiments across multiple modalities across 12 datasets and 6 classic models, verifying our over-invariance insight and the effectiveness of our DivIL framework. Our code is available at https://github.com/kokolerk/DivIL. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12413v1-abstract-full').style.display = 'none'; document.getElementById('2502.12413v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 17 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.12355">arXiv:2502.12355</a> <span> [<a href="https://arxiv.org/pdf/2502.12355">pdf</a>, <a href="https://arxiv.org/format/2502.12355">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Robotics">cs.RO</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Machine Learning">cs.LG</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Systems and Control">eess.SY</span> </div> </div> <p class="title is-5 mathjax"> Hovering Flight of Soft-Actuated Insect-Scale Micro Aerial Vehicles using Deep Reinforcement Learning </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Hsiao%2C+Y">Yi-Hsuan Hsiao</a>, <a href="/search/?searchtype=author&query=Chen%2C+W">Wei-Tung Chen</a>, <a href="/search/?searchtype=author&query=Chang%2C+Y">Yun-Sheng Chang</a>, <a href="/search/?searchtype=author&query=Agrawal%2C+P">Pulkit Agrawal</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">YuFeng Chen</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.12355v1-abstract-short" style="display: inline;"> Soft-actuated insect-scale micro aerial vehicles (IMAVs) pose unique challenges for designing robust and computationally efficient controllers. At the millimeter scale, fast robot dynamics ($\sim$ms), together with system delay, model uncertainty, and external disturbances significantly affect flight performances. Here, we design a deep reinforcement learning (RL) controller that addresses system… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12355v1-abstract-full').style.display = 'inline'; document.getElementById('2502.12355v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.12355v1-abstract-full" style="display: none;"> Soft-actuated insect-scale micro aerial vehicles (IMAVs) pose unique challenges for designing robust and computationally efficient controllers. At the millimeter scale, fast robot dynamics ($\sim$ms), together with system delay, model uncertainty, and external disturbances significantly affect flight performances. Here, we design a deep reinforcement learning (RL) controller that addresses system delay and uncertainties. To initialize this neural network (NN) controller, we propose a modified behavior cloning (BC) approach with state-action re-matching to account for delay and domain-randomized expert demonstration to tackle uncertainty. Then we apply proximal policy optimization (PPO) to fine-tune the policy during RL, enhancing performance and smoothing commands. In simulations, our modified BC substantially increases the mean reward compared to baseline BC; and RL with PPO improves flight quality and reduces command fluctuations. We deploy this controller on two different insect-scale aerial robots that weigh 720 mg and 850 mg, respectively. The robots demonstrate multiple successful zero-shot hovering flights, with the longest lasting 50 seconds and root-mean-square errors of 1.34 cm in lateral direction and 0.05 cm in altitude, marking the first end-to-end deep RL-based flight on soft-driven IMAVs. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12355v1-abstract-full').style.display = 'none'; document.getElementById('2502.12355v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 17 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">7 pages, 7 figures, accepted to 2025 IEEE International Conference on Soft Robotics (RoboSoft)</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.12231">arXiv:2502.12231</a> <span> [<a href="https://arxiv.org/pdf/2502.12231">pdf</a>, <a href="https://arxiv.org/format/2502.12231">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computer Vision and Pattern Recognition">cs.CV</span> </div> </div> <p class="title is-5 mathjax"> PUGS: Zero-shot Physical Understanding with Gaussian Splatting </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Shuai%2C+Y">Yinghao Shuai</a>, <a href="/search/?searchtype=author&query=Yu%2C+R">Ran Yu</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yuantao Chen</a>, <a href="/search/?searchtype=author&query=Jiang%2C+Z">Zijian Jiang</a>, <a href="/search/?searchtype=author&query=Song%2C+X">Xiaowei Song</a>, <a href="/search/?searchtype=author&query=Wang%2C+N">Nan Wang</a>, <a href="/search/?searchtype=author&query=Zheng%2C+J">Jv Zheng</a>, <a href="/search/?searchtype=author&query=Ma%2C+J">Jianzhu Ma</a>, <a href="/search/?searchtype=author&query=Yang%2C+M">Meng Yang</a>, <a href="/search/?searchtype=author&query=Wang%2C+Z">Zhicheng Wang</a>, <a href="/search/?searchtype=author&query=Ding%2C+W">Wenbo Ding</a>, <a href="/search/?searchtype=author&query=Zhao%2C+H">Hao Zhao</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.12231v1-abstract-short" style="display: inline;"> Current robotic systems can understand the categories and poses of objects well. But understanding physical properties like mass, friction, and hardness, in the wild, remains challenging. We propose a new method that reconstructs 3D objects using the Gaussian splatting representation and predicts various physical properties in a zero-shot manner. We propose two techniques during the reconstruction… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12231v1-abstract-full').style.display = 'inline'; document.getElementById('2502.12231v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.12231v1-abstract-full" style="display: none;"> Current robotic systems can understand the categories and poses of objects well. But understanding physical properties like mass, friction, and hardness, in the wild, remains challenging. We propose a new method that reconstructs 3D objects using the Gaussian splatting representation and predicts various physical properties in a zero-shot manner. We propose two techniques during the reconstruction phase: a geometry-aware regularization loss function to improve the shape quality and a region-aware feature contrastive loss function to promote region affinity. Two other new techniques are designed during inference: a feature-based property propagation module and a volume integration module tailored for the Gaussian representation. Our framework is named as zero-shot physical understanding with Gaussian splatting, or PUGS. PUGS achieves new state-of-the-art results on the standard benchmark of ABO-500 mass prediction. We provide extensive quantitative ablations and qualitative visualization to demonstrate the mechanism of our designs. We show the proposed methodology can help address challenging real-world grasping tasks. Our codes, data, and models are available at https://github.com/EverNorif/PUGS <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12231v1-abstract-full').style.display = 'none'; document.getElementById('2502.12231v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 17 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">ICRA 2025, Project page: https://evernorif.github.io/PUGS/</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.12206">arXiv:2502.12206</a> <span> [<a href="https://arxiv.org/pdf/2502.12206">pdf</a>, <a href="https://arxiv.org/format/2502.12206">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Artificial Intelligence">cs.AI</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Computation and Language">cs.CL</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Machine Learning">cs.LG</span> </div> </div> <p class="title is-5 mathjax"> Evaluating the Paperclip Maximizer: Are RL-Based Language Models More Likely to Pursue Instrumental Goals? </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=He%2C+Y">Yufei He</a>, <a href="/search/?searchtype=author&query=Li%2C+Y">Yuexin Li</a>, <a href="/search/?searchtype=author&query=Wu%2C+J">Jiaying Wu</a>, <a href="/search/?searchtype=author&query=Sui%2C+Y">Yuan Sui</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yulin Chen</a>, <a href="/search/?searchtype=author&query=Hooi%2C+B">Bryan Hooi</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.12206v1-abstract-short" style="display: inline;"> As large language models (LLMs) continue to evolve, ensuring their alignment with human goals and values remains a pressing challenge. A key concern is \textit{instrumental convergence}, where an AI system, in optimizing for a given objective, develops unintended intermediate goals that override the ultimate objective and deviate from human-intended goals. This issue is particularly relevant in re… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12206v1-abstract-full').style.display = 'inline'; document.getElementById('2502.12206v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.12206v1-abstract-full" style="display: none;"> As large language models (LLMs) continue to evolve, ensuring their alignment with human goals and values remains a pressing challenge. A key concern is \textit{instrumental convergence}, where an AI system, in optimizing for a given objective, develops unintended intermediate goals that override the ultimate objective and deviate from human-intended goals. This issue is particularly relevant in reinforcement learning (RL)-trained models, which can generate creative but unintended strategies to maximize rewards. In this paper, we explore instrumental convergence in LLMs by comparing models trained with direct RL optimization (e.g., the o1 model) to those trained with reinforcement learning from human feedback (RLHF). We hypothesize that RL-driven models exhibit a stronger tendency for instrumental convergence due to their optimization of goal-directed behavior in ways that may misalign with human intentions. To assess this, we introduce InstrumentalEval, a benchmark for evaluating instrumental convergence in RL-trained LLMs. Initial experiments reveal cases where a model tasked with making money unexpectedly pursues instrumental objectives, such as self-replication, implying signs of instrumental convergence. Our findings contribute to a deeper understanding of alignment challenges in AI systems and the risks posed by unintended model behaviors. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12206v1-abstract-full').style.display = 'none'; document.getElementById('2502.12206v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 16 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.12176">arXiv:2502.12176</a> <span> [<a href="https://arxiv.org/pdf/2502.12176">pdf</a>, <a href="https://arxiv.org/format/2502.12176">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Machine Learning">cs.LG</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Artificial Intelligence">cs.AI</span> </div> </div> <p class="title is-5 mathjax"> Ten Challenging Problems in Federated Foundation Models </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Fan%2C+T">Tao Fan</a>, <a href="/search/?searchtype=author&query=Gu%2C+H">Hanlin Gu</a>, <a href="/search/?searchtype=author&query=Cao%2C+X">Xuemei Cao</a>, <a href="/search/?searchtype=author&query=Chan%2C+C+S">Chee Seng Chan</a>, <a href="/search/?searchtype=author&query=Chen%2C+Q">Qian Chen</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yiqiang Chen</a>, <a href="/search/?searchtype=author&query=Feng%2C+Y">Yihui Feng</a>, <a href="/search/?searchtype=author&query=Gu%2C+Y">Yang Gu</a>, <a href="/search/?searchtype=author&query=Geng%2C+J">Jiaxiang Geng</a>, <a href="/search/?searchtype=author&query=Luo%2C+B">Bing Luo</a>, <a href="/search/?searchtype=author&query=Liu%2C+S">Shuoling Liu</a>, <a href="/search/?searchtype=author&query=Ong%2C+W+K">Win Kent Ong</a>, <a href="/search/?searchtype=author&query=Ren%2C+C">Chao Ren</a>, <a href="/search/?searchtype=author&query=Shao%2C+J">Jiaqi Shao</a>, <a href="/search/?searchtype=author&query=Sun%2C+C">Chuan Sun</a>, <a href="/search/?searchtype=author&query=Tang%2C+X">Xiaoli Tang</a>, <a href="/search/?searchtype=author&query=Tae%2C+H+X">Hong Xi Tae</a>, <a href="/search/?searchtype=author&query=Tong%2C+Y">Yongxin Tong</a>, <a href="/search/?searchtype=author&query=Wei%2C+S">Shuyue Wei</a>, <a href="/search/?searchtype=author&query=Wu%2C+F">Fan Wu</a>, <a href="/search/?searchtype=author&query=Xi%2C+W">Wei Xi</a>, <a href="/search/?searchtype=author&query=Xu%2C+M">Mingcong Xu</a>, <a href="/search/?searchtype=author&query=Yang%2C+H">He Yang</a>, <a href="/search/?searchtype=author&query=Yang%2C+X">Xin Yang</a>, <a href="/search/?searchtype=author&query=Yan%2C+J">Jiangpeng Yan</a> , et al. (8 additional authors not shown) </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.12176v1-abstract-short" style="display: inline;"> Federated Foundation Models (FedFMs) represent a distributed learning paradigm that fuses general competences of foundation models as well as privacy-preserving capabilities of federated learning. This combination allows the large foundation models and the small local domain models at the remote clients to learn from each other in a teacher-student learning setting. This paper provides a comprehen… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12176v1-abstract-full').style.display = 'inline'; document.getElementById('2502.12176v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.12176v1-abstract-full" style="display: none;"> Federated Foundation Models (FedFMs) represent a distributed learning paradigm that fuses general competences of foundation models as well as privacy-preserving capabilities of federated learning. This combination allows the large foundation models and the small local domain models at the remote clients to learn from each other in a teacher-student learning setting. This paper provides a comprehensive summary of the ten challenging problems inherent in FedFMs, encompassing foundational theory, utilization of private data, continual learning, unlearning, Non-IID and graph data, bidirectional knowledge transfer, incentive mechanism design, game mechanism design, model watermarking, and efficiency. The ten challenging problems manifest in five pivotal aspects: ``Foundational Theory," which aims to establish a coherent and unifying theoretical framework for FedFMs. ``Data," addressing the difficulties in leveraging domain-specific knowledge from private data while maintaining privacy; ``Heterogeneity," examining variations in data, model, and computational resources across clients; ``Security and Privacy," focusing on defenses against malicious attacks and model theft; and ``Efficiency," highlighting the need for improvements in training, communication, and parameter efficiency. For each problem, we offer a clear mathematical definition on the objective function, analyze existing methods, and discuss the key challenges and potential solutions. This in-depth exploration aims to advance the theoretical foundations of FedFMs, guide practical implementations, and inspire future research to overcome these obstacles, thereby enabling the robust, efficient, and privacy-preserving FedFMs in various real-world applications. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12176v1-abstract-full').style.display = 'none'; document.getElementById('2502.12176v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 13 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.12135">arXiv:2502.12135</a> <span> [<a href="https://arxiv.org/pdf/2502.12135">pdf</a>, <a href="https://arxiv.org/format/2502.12135">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computer Vision and Pattern Recognition">cs.CV</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Graphics">cs.GR</span> </div> </div> <p class="title is-5 mathjax"> MagicArticulate: Make Your 3D Models Articulation-Ready </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Song%2C+C">Chaoyue Song</a>, <a href="/search/?searchtype=author&query=Zhang%2C+J">Jianfeng Zhang</a>, <a href="/search/?searchtype=author&query=Li%2C+X">Xiu Li</a>, <a href="/search/?searchtype=author&query=Yang%2C+F">Fan Yang</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yiwen Chen</a>, <a href="/search/?searchtype=author&query=Xu%2C+Z">Zhongcong Xu</a>, <a href="/search/?searchtype=author&query=Liew%2C+J+H">Jun Hao Liew</a>, <a href="/search/?searchtype=author&query=Guo%2C+X">Xiaoyang Guo</a>, <a href="/search/?searchtype=author&query=Liu%2C+F">Fayao Liu</a>, <a href="/search/?searchtype=author&query=Feng%2C+J">Jiashi Feng</a>, <a href="/search/?searchtype=author&query=Lin%2C+G">Guosheng Lin</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.12135v2-abstract-short" style="display: inline;"> With the explosive growth of 3D content creation, there is an increasing demand for automatically converting static 3D models into articulation-ready versions that support realistic animation. Traditional approaches rely heavily on manual annotation, which is both time-consuming and labor-intensive. Moreover, the lack of large-scale benchmarks has hindered the development of learning-based solutio… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12135v2-abstract-full').style.display = 'inline'; document.getElementById('2502.12135v2-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.12135v2-abstract-full" style="display: none;"> With the explosive growth of 3D content creation, there is an increasing demand for automatically converting static 3D models into articulation-ready versions that support realistic animation. Traditional approaches rely heavily on manual annotation, which is both time-consuming and labor-intensive. Moreover, the lack of large-scale benchmarks has hindered the development of learning-based solutions. In this work, we present MagicArticulate, an effective framework that automatically transforms static 3D models into articulation-ready assets. Our key contributions are threefold. First, we introduce Articulation-XL, a large-scale benchmark containing over 33k 3D models with high-quality articulation annotations, carefully curated from Objaverse-XL. Second, we propose a novel skeleton generation method that formulates the task as a sequence modeling problem, leveraging an auto-regressive transformer to naturally handle varying numbers of bones or joints within skeletons and their inherent dependencies across different 3D models. Third, we predict skinning weights using a functional diffusion process that incorporates volumetric geodesic distance priors between vertices and joints. Extensive experiments demonstrate that MagicArticulate significantly outperforms existing methods across diverse object categories, achieving high-quality articulation that enables realistic animation. Project page: https://chaoyuesong.github.io/MagicArticulate. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.12135v2-abstract-full').style.display = 'none'; document.getElementById('2502.12135v2-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 18 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">v1</span> submitted 17 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">Project: https://chaoyuesong.github.io/MagicArticulate</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.11974">arXiv:2502.11974</a> <span> [<a href="https://arxiv.org/pdf/2502.11974">pdf</a>, <a href="https://arxiv.org/format/2502.11974">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computer Vision and Pattern Recognition">cs.CV</span> </div> </div> <p class="title is-5 mathjax"> Image Inversion: A Survey from GANs to Diffusion and Beyond </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Chen%2C+Y">Yinan Chen</a>, <a href="/search/?searchtype=author&query=Zhang%2C+J">Jiangning Zhang</a>, <a href="/search/?searchtype=author&query=Bi%2C+Y">Yali Bi</a>, <a href="/search/?searchtype=author&query=Hu%2C+X">Xiaobin Hu</a>, <a href="/search/?searchtype=author&query=Hu%2C+T">Teng Hu</a>, <a href="/search/?searchtype=author&query=Xue%2C+Z">Zhucun Xue</a>, <a href="/search/?searchtype=author&query=Yi%2C+R">Ran Yi</a>, <a href="/search/?searchtype=author&query=Liu%2C+Y">Yong Liu</a>, <a href="/search/?searchtype=author&query=Tai%2C+Y">Ying Tai</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.11974v1-abstract-short" style="display: inline;"> Image inversion is a fundamental task in generative models, aiming to map images back to their latent representations to enable downstream applications such as editing, restoration, and style transfer. This paper provides a comprehensive review of the latest advancements in image inversion techniques, focusing on two main paradigms: Generative Adversarial Network (GAN) inversion and diffusion mode… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11974v1-abstract-full').style.display = 'inline'; document.getElementById('2502.11974v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.11974v1-abstract-full" style="display: none;"> Image inversion is a fundamental task in generative models, aiming to map images back to their latent representations to enable downstream applications such as editing, restoration, and style transfer. This paper provides a comprehensive review of the latest advancements in image inversion techniques, focusing on two main paradigms: Generative Adversarial Network (GAN) inversion and diffusion model inversion. We categorize these techniques based on their optimization methods. For GAN inversion, we systematically classify existing methods into encoder-based approaches, latent optimization approaches, and hybrid approaches, analyzing their theoretical foundations, technical innovations, and practical trade-offs. For diffusion model inversion, we explore training-free strategies, fine-tuning methods, and the design of additional trainable modules, highlighting their unique advantages and limitations. Additionally, we discuss several popular downstream applications and emerging applications beyond image tasks, identifying current challenges and future research directions. By synthesizing the latest developments, this paper aims to provide researchers and practitioners with a valuable reference resource, promoting further advancements in the field of image inversion. We keep track of the latest works at https://github.com/RyanChenYN/ImageInversion <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11974v1-abstract-full').style.display = 'none'; document.getElementById('2502.11974v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 17 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">10 pages, 2 figures</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.11766">arXiv:2502.11766</a> <span> [<a href="https://arxiv.org/pdf/2502.11766">pdf</a>, <a href="https://arxiv.org/format/2502.11766">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computation and Language">cs.CL</span> </div> </div> <p class="title is-5 mathjax"> Warmup-Distill: Bridge the Distribution Mismatch between Teacher and Student before Knowledge Distillation </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Sun%2C+Z">Zengkui Sun</a>, <a href="/search/?searchtype=author&query=Liu%2C+Y">Yijin Liu</a>, <a href="/search/?searchtype=author&query=Meng%2C+F">Fandong Meng</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yufeng Chen</a>, <a href="/search/?searchtype=author&query=Xu%2C+J">Jinan Xu</a>, <a href="/search/?searchtype=author&query=Zhou%2C+J">Jie Zhou</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.11766v1-abstract-short" style="display: inline;"> The widespread deployment of Large Language Models (LLMs) is hindered by the high computational demands, making knowledge distillation (KD) crucial for developing compact smaller ones. However, the conventional KD methods endure the distribution mismatch issue between the teacher and student models, leading to the poor performance of distillation. For instance, the widely-used KL-based methods suf… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11766v1-abstract-full').style.display = 'inline'; document.getElementById('2502.11766v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.11766v1-abstract-full" style="display: none;"> The widespread deployment of Large Language Models (LLMs) is hindered by the high computational demands, making knowledge distillation (KD) crucial for developing compact smaller ones. However, the conventional KD methods endure the distribution mismatch issue between the teacher and student models, leading to the poor performance of distillation. For instance, the widely-used KL-based methods suffer the mode-averaging and mode-collapsing problems, since the mismatched probabitliy distribution between both models. Previous studies mainly optimize this issue via different distance calculations towards the distribution of both models. Unfortunately, the distribution mismatch issue still exists in the early stage of the distillation. Hence, to reduce the impact of distribution mismatch, we propose a simple yet efficient method, named Warmup-Distill, which aligns the distillation of the student to that of the teacher in advance of distillation. Specifically, we first detect the distribution of the student model in practical scenarios with its internal knowledge, and then modify the knowledge with low probability via the teacher as the checker. Consequently, Warmup-Distill aligns the internal student's knowledge to that of the teacher, which expands the distribution of the student with the teacher's, and assists the student model to learn better in the subsequent distillation. Experiments on the seven benchmarks demonstrate that Warmup-Distill could provide a warmup student more suitable for distillation, which outperforms the vanilla student by as least +0.4 averaged score among all benchmarks. Noteably, with the assistance of Warmup-Distill, the distillation on the math task could yield a further improvement, at most +1.9% accuracy. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11766v1-abstract-full').style.display = 'none'; document.getElementById('2502.11766v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 17 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">11 Pages, 4 figures, Code at https://github.com/Acerkoo/WarmupDistill</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.11659">arXiv:2502.11659</a> <span> [<a href="https://arxiv.org/pdf/2502.11659">pdf</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Human-Computer Interaction">cs.HC</span> </div> </div> <p class="title is-5 mathjax"> An Innovative Brain-Computer Interface Interaction System Based on the Large Language Model </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Jin%2C+J">Jing Jin</a>, <a href="/search/?searchtype=author&query=Zhang%2C+Y">Yutao Zhang</a>, <a href="/search/?searchtype=author&query=Xu%2C+R">Ruitian Xu</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yixin Chen</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.11659v2-abstract-short" style="display: inline;"> Recent advancements in large language models (LLMs) provide a more effective pathway for upgrading brain-computer interface (BCI) technology in terms of user interaction. The widespread adoption of BCIs in daily application scenarios is still limited by factors such as their single functionality, restricted paradigm design, weak multilingual support, and low levels of intelligence. In this paper,… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11659v2-abstract-full').style.display = 'inline'; document.getElementById('2502.11659v2-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.11659v2-abstract-full" style="display: none;"> Recent advancements in large language models (LLMs) provide a more effective pathway for upgrading brain-computer interface (BCI) technology in terms of user interaction. The widespread adoption of BCIs in daily application scenarios is still limited by factors such as their single functionality, restricted paradigm design, weak multilingual support, and low levels of intelligence. In this paper, we propose an innovative BCI system that deeply integrates a steady-state visual evoked potential (SSVEP) speller with an LLM application programming interface (API). It allows natural language input through the SSVEP speller and dynamically calls large models to generate SSVEP paradigms. The command prompt, blinking frequency, and layout position are adjustable to meet the user's control requirements in various scenarios. More than ten languages are compatible with the multilingual support of LLM. A variety of task scenarios, such as home appliance control, robotic arm operation, and unmanned aerial vehicle (UAV) management are provided. The task interfaces of the system can be personalized according to the user's habits, usage scenarios, and equipment characteristics. By combining the SSVEP speller with an LLM, the system solves numerous challenges faced by current BCI systems and makes breakthroughs in functionality, intelligence, and multilingual support. The introduction of LLM not only enhances user experience but also expands the potential applications of BCI technology in real-world environments. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11659v2-abstract-full').style.display = 'none'; document.getElementById('2502.11659v2-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 18 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">v1</span> submitted 17 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">10 pages,3 figures</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.11610">arXiv:2502.11610</a> <span> [<a href="https://arxiv.org/pdf/2502.11610">pdf</a>, <a href="https://arxiv.org/format/2502.11610">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Information Retrieval">cs.IR</span> </div> </div> <p class="title is-5 mathjax"> Accuracy Assessment of OpenAlex and Clarivate Scholar ID with an LLM-Assisted Benchmark </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Zhao%2C+R">Renyu Zhao</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yunxin Chen</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.11610v1-abstract-short" style="display: inline;"> In quantitative SciSci (science of science) studies, accurately identifying individual scholars is paramount for scientific data analysis. However, the variability in how names are represented-due to commonality, abbreviations, and different spelling conventions-complicates this task. While identifier systems like ORCID are being developed, many scholars remain unregistered, and numerous publicati… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11610v1-abstract-full').style.display = 'inline'; document.getElementById('2502.11610v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.11610v1-abstract-full" style="display: none;"> In quantitative SciSci (science of science) studies, accurately identifying individual scholars is paramount for scientific data analysis. However, the variability in how names are represented-due to commonality, abbreviations, and different spelling conventions-complicates this task. While identifier systems like ORCID are being developed, many scholars remain unregistered, and numerous publications are not included. Scholarly databases such as Clarivate and OpenAlex have introduced their own ID systems as preliminary name disambiguation solutions. This study evaluates the effectiveness of these systems across different groups to determine their suitability for various application scenarios. We sampled authors from the top quartile (Q1) of Web of Science (WOS) journals based on country, discipline, and number of corresponding author papers. For each group, we selected 100 scholars and meticulously annotated all their papers using a Search-enhanced Large Language Model method. Using these annotations, we identified the corresponding IDs in OpenAlex and Clarivate, extracted all associated papers, filtered for Q1 WOS journals, and calculated precision and recall by comparing against the annotated dataset. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11610v1-abstract-full').style.display = 'none'; document.getElementById('2502.11610v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 17 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.11440">arXiv:2502.11440</a> <span> [<a href="https://arxiv.org/pdf/2502.11440">pdf</a>, <a href="https://arxiv.org/format/2502.11440">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computer Vision and Pattern Recognition">cs.CV</span> </div> </div> <p class="title is-5 mathjax"> Medical Image Registration Meets Vision Foundation Model: Prototype Learning and Contour Awareness </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Xu%2C+H">Hao Xu</a>, <a href="/search/?searchtype=author&query=Xue%2C+T">Tengfei Xue</a>, <a href="/search/?searchtype=author&query=Fan%2C+J">Jianan Fan</a>, <a href="/search/?searchtype=author&query=Liu%2C+D">Dongnan Liu</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yuqian Chen</a>, <a href="/search/?searchtype=author&query=Zhang%2C+F">Fan Zhang</a>, <a href="/search/?searchtype=author&query=Westin%2C+C">Carl-Fredrik Westin</a>, <a href="/search/?searchtype=author&query=Kikinis%2C+R">Ron Kikinis</a>, <a href="/search/?searchtype=author&query=O%27Donnell%2C+L+J">Lauren J. O'Donnell</a>, <a href="/search/?searchtype=author&query=Cai%2C+W">Weidong Cai</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.11440v1-abstract-short" style="display: inline;"> Medical image registration is a fundamental task in medical image analysis, aiming to establish spatial correspondences between paired images. However, existing unsupervised deformable registration methods rely solely on intensity-based similarity metrics, lacking explicit anatomical knowledge, which limits their accuracy and robustness. Vision foundation models, such as the Segment Anything Model… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11440v1-abstract-full').style.display = 'inline'; document.getElementById('2502.11440v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.11440v1-abstract-full" style="display: none;"> Medical image registration is a fundamental task in medical image analysis, aiming to establish spatial correspondences between paired images. However, existing unsupervised deformable registration methods rely solely on intensity-based similarity metrics, lacking explicit anatomical knowledge, which limits their accuracy and robustness. Vision foundation models, such as the Segment Anything Model (SAM), can generate high-quality segmentation masks that provide explicit anatomical structure knowledge, addressing the limitations of traditional methods that depend only on intensity similarity. Based on this, we propose a novel SAM-assisted registration framework incorporating prototype learning and contour awareness. The framework includes: (1) Explicit anatomical information injection, where SAM-generated segmentation masks are used as auxiliary inputs throughout training and testing to ensure the consistency of anatomical information; (2) Prototype learning, which leverages segmentation masks to extract prototype features and aligns prototypes to optimize semantic correspondences between images; and (3) Contour-aware loss, a contour-aware loss is designed that leverages the edges of segmentation masks to improve the model's performance in fine-grained deformation fields. Extensive experiments demonstrate that the proposed framework significantly outperforms existing methods across multiple datasets, particularly in challenging scenarios with complex anatomical structures and ambiguous boundaries. Our code is available at https://github.com/HaoXu0507/IPMI25-SAM-Assisted-Registration. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11440v1-abstract-full').style.display = 'none'; document.getElementById('2502.11440v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 16 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">Accepted by Information Processing in Medical Imaging (IPMI) 2025</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.11430">arXiv:2502.11430</a> <span> [<a href="https://arxiv.org/pdf/2502.11430">pdf</a>, <a href="https://arxiv.org/format/2502.11430">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Human-Computer Interaction">cs.HC</span> </div> </div> <p class="title is-5 mathjax"> "An Image of Ourselves in Our Minds": How College-educated Online Dating Users Construct Profiles for Effective Self Presentation </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Zhang%2C+F">Fan Zhang</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yun Chen</a>, <a href="/search/?searchtype=author&query=Zeng%2C+X">Xiaoke Zeng</a>, <a href="/search/?searchtype=author&query=Wang%2C+T">Tianqi Wang</a>, <a href="/search/?searchtype=author&query=Ling%2C+L">Long Ling</a>, <a href="/search/?searchtype=author&query=LC%2C+R">RAY LC</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.11430v1-abstract-short" style="display: inline;"> Online dating is frequently used by individuals looking for potential relationships and intimate connections. Central to dating apps is the creation and refinement of a dating profile, which represents the way individuals desire to present themselves to potential mates, while hiding information they do not care to share. To investigate the way frequent users of dating apps construct their online p… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11430v1-abstract-full').style.display = 'inline'; document.getElementById('2502.11430v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.11430v1-abstract-full" style="display: none;"> Online dating is frequently used by individuals looking for potential relationships and intimate connections. Central to dating apps is the creation and refinement of a dating profile, which represents the way individuals desire to present themselves to potential mates, while hiding information they do not care to share. To investigate the way frequent users of dating apps construct their online profiles and perceive the effectiveness of strategies taken in making profiles, we conducted semi-structured interviews with 20 experienced users who are Chinese college-educated young adults and uncovered the processes and rationales by which they make profiles for online dating, particularly in selecting images for inclusion. We found that participants used idealized photos that exaggerated their positive personality traits, sometimes traits that they do not possess but perceive others to desire, and sometimes even traits they wish they had possessed. Users also strategically used photos that show personality and habits without showing themselves, and often hid certain identifying information to reduce privacy risks. This analysis signals potential factors that are key in building online dating profiles, providing design implications for systems that limit the use of inaccurate information while still promoting self-expression in relationship platforms. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11430v1-abstract-full').style.display = 'none'; document.getElementById('2502.11430v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 16 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">30 pages, 6 figures, to be published in CSCW 2025</span> </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">ACM Class:</span> J.4 </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.11405">arXiv:2502.11405</a> <span> [<a href="https://arxiv.org/pdf/2502.11405">pdf</a>, <a href="https://arxiv.org/format/2502.11405">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computation and Language">cs.CL</span> </div> </div> <p class="title is-5 mathjax"> LayAlign: Enhancing Multilingual Reasoning in Large Language Models via Layer-Wise Adaptive Fusion and Alignment Strategy </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Ruan%2C+Z">Zhiwen Ruan</a>, <a href="/search/?searchtype=author&query=Li%2C+Y">Yixia Li</a>, <a href="/search/?searchtype=author&query=Zhu%2C+H">He Zhu</a>, <a href="/search/?searchtype=author&query=Wang%2C+L">Longyue Wang</a>, <a href="/search/?searchtype=author&query=Luo%2C+W">Weihua Luo</a>, <a href="/search/?searchtype=author&query=Zhang%2C+K">Kaifu Zhang</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yun Chen</a>, <a href="/search/?searchtype=author&query=Chen%2C+G">Guanhua Chen</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.11405v1-abstract-short" style="display: inline;"> Despite being pretrained on multilingual corpora, large language models (LLMs) exhibit suboptimal performance on low-resource languages. Recent approaches have leveraged multilingual encoders alongside LLMs by introducing trainable parameters connecting the two models. However, these methods typically focus on the encoder's output, overlooking valuable information from other layers. We propose \an… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11405v1-abstract-full').style.display = 'inline'; document.getElementById('2502.11405v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.11405v1-abstract-full" style="display: none;"> Despite being pretrained on multilingual corpora, large language models (LLMs) exhibit suboptimal performance on low-resource languages. Recent approaches have leveraged multilingual encoders alongside LLMs by introducing trainable parameters connecting the two models. However, these methods typically focus on the encoder's output, overlooking valuable information from other layers. We propose \aname (\mname), a framework that integrates representations from all encoder layers, coupled with the \attaname mechanism to enable layer-wise interaction between the LLM and the multilingual encoder. Extensive experiments on multilingual reasoning tasks, along with analyses of learned representations, show that our approach consistently outperforms existing baselines. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11405v1-abstract-full').style.display = 'none'; document.getElementById('2502.11405v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 16 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">In Findings of NAACL 2025(The 2025 Annual Conference of the Nations of the Americas Chapter of the ACL)</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.11332">arXiv:2502.11332</a> <span> [<a href="https://arxiv.org/pdf/2502.11332">pdf</a>, <a href="https://arxiv.org/format/2502.11332">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Methodology">stat.ME</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Applications">stat.AP</span> </div> </div> <p class="title is-5 mathjax"> Stochastic Block Covariance Matrix Estimation </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Chen%2C+Y">Yunran Chen</a>, <a href="/search/?searchtype=author&query=Tokdar%2C+S+T">Surya T Tokdar</a>, <a href="/search/?searchtype=author&query=Groh%2C+J+M">Jennifer M Groh</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.11332v1-abstract-short" style="display: inline;"> Motivated by a neuroscience application we study the problem of statistical estimation of a high-dimensional covariance matrix with a block structure. The block model embeds a structural assumption: the population of items (neurons) can be divided into latent sub-populations with shared associative covariation within blocks and shared associative or dis-associative covariation across blocks. Unlik… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11332v1-abstract-full').style.display = 'inline'; document.getElementById('2502.11332v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.11332v1-abstract-full" style="display: none;"> Motivated by a neuroscience application we study the problem of statistical estimation of a high-dimensional covariance matrix with a block structure. The block model embeds a structural assumption: the population of items (neurons) can be divided into latent sub-populations with shared associative covariation within blocks and shared associative or dis-associative covariation across blocks. Unlike the block diagonal assumption, our block structure incorporates positive or negative pairwise correlation between blocks. In addition to offering reasonable modeling choices in neuroscience and economics, the block covariance matrix assumption is interesting purely from the perspective of statistical estimation theory: (a) it offers in-built dimension reduction and (b) it resembles a regularized factor model without the need of choosing the number of factors. We discuss a hierarchical Bayesian estimation method to simultaneously recover the latent blocks and estimate the overall covariance matrix. We show with numerical experiments that a hierarchical structure and a shrinkage prior are essential to accurate recovery when several blocks are present. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11332v1-abstract-full').style.display = 'none'; document.getElementById('2502.11332v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 16 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.11308">arXiv:2502.11308</a> <span> [<a href="https://arxiv.org/pdf/2502.11308">pdf</a>, <a href="https://arxiv.org/format/2502.11308">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Cryptography and Security">cs.CR</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Artificial Intelligence">cs.AI</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Computation and Language">cs.CL</span> </div> </div> <p class="title is-5 mathjax"> ALGEN: Few-shot Inversion Attacks on Textual Embeddings using Alignment and Generation </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Chen%2C+Y">Yiyi Chen</a>, <a href="/search/?searchtype=author&query=Xu%2C+Q">Qiongkai Xu</a>, <a href="/search/?searchtype=author&query=Bjerva%2C+J">Johannes Bjerva</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.11308v2-abstract-short" style="display: inline;"> With the growing popularity of Large Language Models (LLMs) and vector databases, private textual data is increasingly processed and stored as numerical embeddings. However, recent studies have proven that such embeddings are vulnerable to inversion attacks, where original text is reconstructed to reveal sensitive information. Previous research has largely assumed access to millions of sentences t… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11308v2-abstract-full').style.display = 'inline'; document.getElementById('2502.11308v2-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.11308v2-abstract-full" style="display: none;"> With the growing popularity of Large Language Models (LLMs) and vector databases, private textual data is increasingly processed and stored as numerical embeddings. However, recent studies have proven that such embeddings are vulnerable to inversion attacks, where original text is reconstructed to reveal sensitive information. Previous research has largely assumed access to millions of sentences to train attack models, e.g., through data leakage or nearly unrestricted API access. With our method, a single data point is sufficient for a partially successful inversion attack. With as little as 1k data samples, performance reaches an optimum across a range of black-box encoders, without training on leaked data. We present a Few-shot Textual Embedding Inversion Attack using ALignment and GENeration (ALGEN), by aligning victim embeddings to the attack space and using a generative model to reconstruct text. We find that ALGEN attacks can be effectively transferred across domains and languages, revealing key information. We further examine a variety of defense mechanisms against ALGEN, and find that none are effective, highlighting the vulnerabilities posed by inversion attacks. By significantly lowering the cost of inversion and proving that embedding spaces can be aligned through one-step optimization, we establish a new textual embedding inversion paradigm with broader applications for embedding alignment in NLP. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11308v2-abstract-full').style.display = 'none'; document.getElementById('2502.11308v2-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 18 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">v1</span> submitted 16 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">18 pages, 13 tables, 6 figures</span> </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">ACM Class:</span> I.2; J.6 </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.11238">arXiv:2502.11238</a> <span> [<a href="https://arxiv.org/pdf/2502.11238">pdf</a>, <a href="https://arxiv.org/ps/2502.11238">ps</a>, <a href="https://arxiv.org/format/2502.11238">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Machine Learning">cs.LG</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Information Theory">cs.IT</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Optimization and Control">math.OC</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Machine Learning">stat.ML</span> </div> </div> <p class="title is-5 mathjax"> Span-Agnostic Optimal Sample Complexity and Oracle Inequalities for Average-Reward RL </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Zurek%2C+M">Matthew Zurek</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yudong Chen</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.11238v1-abstract-short" style="display: inline;"> We study the sample complexity of finding an $\varepsilon$-optimal policy in average-reward Markov Decision Processes (MDPs) with a generative model. The minimax optimal span-based complexity of $\widetilde{O}(SAH/\varepsilon^2)$, where $H$ is the span of the optimal bias function, has only been achievable with prior knowledge of the value of $H$. Prior-knowledge-free algorithms have been the obje… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11238v1-abstract-full').style.display = 'inline'; document.getElementById('2502.11238v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.11238v1-abstract-full" style="display: none;"> We study the sample complexity of finding an $\varepsilon$-optimal policy in average-reward Markov Decision Processes (MDPs) with a generative model. The minimax optimal span-based complexity of $\widetilde{O}(SAH/\varepsilon^2)$, where $H$ is the span of the optimal bias function, has only been achievable with prior knowledge of the value of $H$. Prior-knowledge-free algorithms have been the objective of intensive research, but several natural approaches provably fail to achieve this goal. We resolve this problem, developing the first algorithms matching the optimal span-based complexity without $H$ knowledge, both when the dataset size is fixed and when the suboptimality level $\varepsilon$ is fixed. Our main technique combines the discounted reduction approach with a method for automatically tuning the effective horizon based on empirical confidence intervals or lower bounds on performance, which we term horizon calibration. We also develop an empirical span penalization approach, inspired by sample variance penalization, which satisfies an oracle inequality performance guarantee. In particular this algorithm can outperform the minimax complexity in benign settings such as when there exist near-optimal policies with span much smaller than $H$. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11238v1-abstract-full').style.display = 'none'; document.getElementById('2502.11238v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 16 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.11104">arXiv:2502.11104</a> <span> [<a href="https://arxiv.org/pdf/2502.11104">pdf</a>, <a href="https://arxiv.org/format/2502.11104">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Computation and Language">cs.CL</span> </div> </div> <p class="title is-5 mathjax"> Enhancing Cross-Tokenizer Knowledge Distillation with Contextual Dynamical Mapping </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Chen%2C+Y">Yijie Chen</a>, <a href="/search/?searchtype=author&query=Liu%2C+Y">Yijin Liu</a>, <a href="/search/?searchtype=author&query=Meng%2C+F">Fandong Meng</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yufeng Chen</a>, <a href="/search/?searchtype=author&query=Xu%2C+J">Jinan Xu</a>, <a href="/search/?searchtype=author&query=Zhou%2C+J">Jie Zhou</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.11104v1-abstract-short" style="display: inline;"> Knowledge Distillation (KD) has emerged as a prominent technique for model compression. However, conventional KD approaches primarily focus on homogeneous architectures with identical tokenizers, constraining their applicability in cross-architecture scenarios. As for the cross-tokenizer KD, the differences in the tokenizers give rise to two fundamental challenges: (1) sequence misalignment caused… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11104v1-abstract-full').style.display = 'inline'; document.getElementById('2502.11104v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.11104v1-abstract-full" style="display: none;"> Knowledge Distillation (KD) has emerged as a prominent technique for model compression. However, conventional KD approaches primarily focus on homogeneous architectures with identical tokenizers, constraining their applicability in cross-architecture scenarios. As for the cross-tokenizer KD, the differences in the tokenizers give rise to two fundamental challenges: (1) sequence misalignment caused by divergent tokenization strategies, and (2) mismatched vocabulary size and composition. While existing probability-matching methods attempt to address these issues, their efficacy remains limited due to suboptimal alignment in both the sequence and vocabulary aspects. To overcome these limitations, we propose Contextual Dynamic Mapping (CDM), a novel cross-tokenizer distillation framework that employs contextual information to enhance sequence alignment precision and dynamically improves vocabulary mapping. We evaluated the effectiveness of our approach across five advanced and widely-used model families (i.e, LLama3, Phi3, Gemma2, OPT and Qwen2), which were configured into three distinct teacher-student pairs. Our method shows significant advantages over existing cross-tokenizer distillation baselines across diverse benchmarks, including instruction-following, code generation and math. Notably, our analysis reveals that combining conventional same-tokenizer distillation and cross-tokenizer distillation through CDM yields further performance improvements. The code is available at https://github.com/pppa2019/ContexualDynamicMapping <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11104v1-abstract-full').style.display = 'none'; document.getElementById('2502.11104v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 16 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">The code is available at https://github.com/pppa2019/ContexualDynamicMapping</span> </p> </li> <li class="arxiv-result"> <div class="is-marginless"> <p class="list-title is-inline-block"><a href="https://arxiv.org/abs/2502.11102">arXiv:2502.11102</a> <span> [<a href="https://arxiv.org/pdf/2502.11102">pdf</a>, <a href="https://arxiv.org/format/2502.11102">other</a>] </span> </p> <div class="tags is-inline-block"> <span class="tag is-small is-link tooltip is-tooltip-top" data-tooltip="Artificial Intelligence">cs.AI</span> <span class="tag is-small is-grey tooltip is-tooltip-top" data-tooltip="Machine Learning">cs.LG</span> </div> </div> <p class="title is-5 mathjax"> OptMATH: A Scalable Bidirectional Data Synthesis Framework for Optimization Modeling </p> <p class="authors"> <span class="search-hit">Authors:</span> <a href="/search/?searchtype=author&query=Lu%2C+H">Hongliang Lu</a>, <a href="/search/?searchtype=author&query=Xie%2C+Z">Zhonglin Xie</a>, <a href="/search/?searchtype=author&query=Wu%2C+Y">Yaoyu Wu</a>, <a href="/search/?searchtype=author&query=Ren%2C+C">Can Ren</a>, <a href="/search/?searchtype=author&query=Chen%2C+Y">Yuxuan Chen</a>, <a href="/search/?searchtype=author&query=Wen%2C+Z">Zaiwen Wen</a> </p> <p class="abstract mathjax"> <span class="has-text-black-bis has-text-weight-semibold">Abstract</span>: <span class="abstract-short has-text-grey-dark mathjax" id="2502.11102v1-abstract-short" style="display: inline;"> Despite the rapid development of large language models (LLMs), a fundamental challenge persists: the lack of high-quality optimization modeling datasets hampers LLMs' robust modeling of practical optimization problems from natural language descriptions (NL). This data scarcity also contributes to the generalization difficulties experienced by learning-based methods. To address these challenges, we… <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11102v1-abstract-full').style.display = 'inline'; document.getElementById('2502.11102v1-abstract-short').style.display = 'none';">▽ More</a> </span> <span class="abstract-full has-text-grey-dark mathjax" id="2502.11102v1-abstract-full" style="display: none;"> Despite the rapid development of large language models (LLMs), a fundamental challenge persists: the lack of high-quality optimization modeling datasets hampers LLMs' robust modeling of practical optimization problems from natural language descriptions (NL). This data scarcity also contributes to the generalization difficulties experienced by learning-based methods. To address these challenges, we propose a scalable framework for synthesizing a high-quality dataset, named OptMATH. Starting from curated seed data with mathematical formulations (MF), this framework automatically generates problem data (PD) with controllable complexity. Then, a back-translation step is employed to obtain NL. To verify the correspondence between the NL and the PD, a forward modeling step followed by rejection sampling is used. The accepted pairs constitute the training part of OptMATH. Then a collection of rejected pairs is identified and further filtered. This collection serves as a new benchmark for optimization modeling, containing difficult instances whose lengths are much longer than these of NL4OPT and MAMO. Through extensive experiments, we demonstrate that models of various sizes (0.5B-32B parameters) trained on OptMATH achieve superior results on multiple modeling benchmarks, thereby validating the effectiveness and scalability of our approach. <a class="is-size-7" style="white-space: nowrap;" onclick="document.getElementById('2502.11102v1-abstract-full').style.display = 'none'; document.getElementById('2502.11102v1-abstract-short').style.display = 'inline';">△ Less</a> </span> </p> <p class="is-size-7"><span class="has-text-black-bis has-text-weight-semibold">Submitted</span> 16 February, 2025; <span class="has-text-black-bis has-text-weight-semibold">originally announced</span> February 2025. </p> <p class="comments is-size-7"> <span class="has-text-black-bis has-text-weight-semibold">Comments:</span> <span class="has-text-grey-dark mathjax">This paper has 36 pages, 18 figures, and two co-first authors: Hongliang Lu and Zhonglin Xie</span> </p> </li> </ol> <nav class="pagination is-small is-centered breathe-horizontal" role="navigation" aria-label="pagination"> <a href="" class="pagination-previous is-invisible">Previous </a> <a href="/search/?searchtype=author&query=Chen%2C+Y&start=50" class="pagination-next" >Next </a> <ul class="pagination-list"> <li> <a href="/search/?searchtype=author&query=Chen%2C+Y&start=0" class="pagination-link is-current" aria-label="Goto page 1">1 </a> </li> <li> <a href="/search/?searchtype=author&query=Chen%2C+Y&start=50" class="pagination-link " aria-label="Page 2" aria-current="page">2 </a> </li> <li> <a href="/search/?searchtype=author&query=Chen%2C+Y&start=100" class="pagination-link " aria-label="Page 3" aria-current="page">3 </a> </li> <li> <a href="/search/?searchtype=author&query=Chen%2C+Y&start=150" class="pagination-link " aria-label="Page 4" aria-current="page">4 </a> </li> <li> <a href="/search/?searchtype=author&query=Chen%2C+Y&start=200" class="pagination-link " aria-label="Page 5" aria-current="page">5 </a> </li> <li><span class="pagination-ellipsis">…</span></li> </ul> </nav> <div class="is-hidden-tablet">  <span class="help" style="display: inline-block;"><a href="https://github.com/arXiv/arxiv-search/releases">Search v0.5.6 released 2020-02-24</a>  </span> </div> </div> </main> <footer> <div class="columns is-desktop" role="navigation" aria-label="Secondary">  <div class="column"> <div class="columns"> <div class="column"> <ul class="nav-spaced"> <li><a href="https://info.arxiv.org/about">About</a></li> <li><a href="https://info.arxiv.org/help">Help</a></li> </ul> </div> <div class="column"> <ul class="nav-spaced"> <li> <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 512 512" class="icon filter-black" role="presentation"><title>contact arXiv</title><desc>Click here to contact arXiv</desc><path d="M502.3 190.8c3.9-3.1 9.7-.2 9.7 4.7V400c0 26.5-21.5 48-48 48H48c-26.5 0-48-21.5-48-48V195.6c0-5 5.7-7.8 9.7-4.7 22.4 17.4 52.1 39.5 154.1 113.6 21.1 15.4 56.7 47.8 92.2 47.6 35.7.3 72-32.8 92.3-47.6 102-74.1 131.6-96.3 154-113.7zM256 320c23.2.4 56.6-29.2 73.4-41.4 132.7-96.3 142.8-104.7 173.4-128.7 5.8-4.5 9.2-11.5 9.2-18.9v-19c0-26.5-21.5-48-48-48H48C21.5 64 0 85.5 0 112v19c0 7.4 3.4 14.3 9.2 18.9 30.6 23.9 40.7 32.4 173.4 128.7 16.8 12.2 50.2 41.8 73.4 41.4z"/></svg> <a href="https://info.arxiv.org/help/contact.html"> Contact</a> </li> <li> <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 512 512" class="icon filter-black" role="presentation"><title>subscribe to arXiv mailings</title><desc>Click here to subscribe</desc><path d="M476 3.2L12.5 270.6c-18.1 10.4-15.8 35.6 2.2 43.2L121 358.4l287.3-253.2c5.5-4.9 13.3 2.6 8.6 8.3L176 407v80.5c0 23.6 28.5 32.9 42.5 15.8L282 426l124.6 52.2c14.2 6 30.4-2.9 33-18.2l72-432C515 7.8 493.3-6.8 476 3.2z"/></svg> <a href="https://info.arxiv.org/help/subscribe"> Subscribe</a> </li> </ul> </div> </div> </div>   <div class="column"> <div class="columns"> <div class="column"> <ul class="nav-spaced"> <li><a href="https://info.arxiv.org/help/license/index.html">Copyright</a></li> <li><a href="https://info.arxiv.org/help/policies/privacy_policy.html">Privacy Policy</a></li> </ul> </div> <div class="column sorry-app-links"> <ul class="nav-spaced"> <li><a href="https://info.arxiv.org/help/web_accessibility.html">Web Accessibility Assistance</a></li> <li> <p class="help"> <a class="a11y-main-link" href="https://status.arxiv.org" target="_blank">arXiv Operational Status <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 256 512" class="icon filter-dark_grey" role="presentation"><path d="M224.3 273l-136 136c-9.4 9.4-24.6 9.4-33.9 0l-22.6-22.6c-9.4-9.4-9.4-24.6 0-33.9l96.4-96.4-96.4-96.4c-9.4-9.4-9.4-24.6 0-33.9L54.3 103c9.4-9.4 24.6-9.4 33.9 0l136 136c9.5 9.4 9.5 24.6.1 34z"/></svg></a><br> Get status notifications via <a class="is-link" href="https://subscribe.sorryapp.com/24846f03/email/new" target="_blank"><svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 512 512" class="icon filter-black" role="presentation"><path d="M502.3 190.8c3.9-3.1 9.7-.2 9.7 4.7V400c0 26.5-21.5 48-48 48H48c-26.5 0-48-21.5-48-48V195.6c0-5 5.7-7.8 9.7-4.7 22.4 17.4 52.1 39.5 154.1 113.6 21.1 15.4 56.7 47.8 92.2 47.6 35.7.3 72-32.8 92.3-47.6 102-74.1 131.6-96.3 154-113.7zM256 320c23.2.4 56.6-29.2 73.4-41.4 132.7-96.3 142.8-104.7 173.4-128.7 5.8-4.5 9.2-11.5 9.2-18.9v-19c0-26.5-21.5-48-48-48H48C21.5 64 0 85.5 0 112v19c0 7.4 3.4 14.3 9.2 18.9 30.6 23.9 40.7 32.4 173.4 128.7 16.8 12.2 50.2 41.8 73.4 41.4z"/></svg>email</a> or <a class="is-link" href="https://subscribe.sorryapp.com/24846f03/slack/new" target="_blank"><svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 448 512" class="icon filter-black" role="presentation"><path d="M94.12 315.1c0 25.9-21.16 47.06-47.06 47.06S0 341 0 315.1c0-25.9 21.16-47.06 47.06-47.06h47.06v47.06zm23.72 0c0-25.9 21.16-47.06 47.06-47.06s47.06 21.16 47.06 47.06v117.84c0 25.9-21.16 47.06-47.06 47.06s-47.06-21.16-47.06-47.06V315.1zm47.06-188.98c-25.9 0-47.06-21.16-47.06-47.06S139 32 164.9 32s47.06 21.16 47.06 47.06v47.06H164.9zm0 23.72c25.9 0 47.06 21.16 47.06 47.06s-21.16 47.06-47.06 47.06H47.06C21.16 243.96 0 222.8 0 196.9s21.16-47.06 47.06-47.06H164.9zm188.98 47.06c0-25.9 21.16-47.06 47.06-47.06 25.9 0 47.06 21.16 47.06 47.06s-21.16 47.06-47.06 47.06h-47.06V196.9zm-23.72 0c0 25.9-21.16 47.06-47.06 47.06-25.9 0-47.06-21.16-47.06-47.06V79.06c0-25.9 21.16-47.06 47.06-47.06 25.9 0 47.06 21.16 47.06 47.06V196.9zM283.1 385.88c25.9 0 47.06 21.16 47.06 47.06 0 25.9-21.16 47.06-47.06 47.06-25.9 0-47.06-21.16-47.06-47.06v-47.06h47.06zm0-23.72c-25.9 0-47.06-21.16-47.06-47.06 0-25.9 21.16-47.06 47.06-47.06h117.84c25.9 0 47.06 21.16 47.06 47.06 0 25.9-21.16 47.06-47.06 47.06H283.1z"/></svg>slack</a> </p> </li> </ul> </div> </div> </div>  </div> </footer> <script src="https://static.arxiv.org/static/base/1.0.0a5/js/member_acknowledgement.js"></script> </body> </html>

CINXE.COM

Search | arXiv e-print repository