index.html

<!DOCTYPE HTML>
<!--
    Read Only by HTML5 UP
    html5up.net | @ajlkn
    Free for personal and commercial use under the CCA 3.0 license (html5up.net/license)
-->
<html>
  <head>
    <title>Benjamin Plaut</title>
    <meta charset="utf-8" />
    <meta name="viewport" content="width=device-width, initial-scale=1, user-scalable=no" />
    <link rel="icon" href="favicon.ico" />
    <link rel="stylesheet" href="assets/main.css" />
    <style type="text/css">
      @import url('https://fonts.googleapis.com/css2?family=Merriweather:wght@300;400;700&display=swap');
      body {
        font-family: "Merriweather", serif;
        color: #444;
      }
      a {
        color: #1E81B0;
      }

      #nav ul li a {
      font-size: var(--base-font-size);
      padding: 2px 4px;
      }


      #logo {
        font-size: 1.8em;
      }

      #header > header {
      padding: 2em 2em 1em 2em;
      }

      #header .image.avatar {
      width: 200px;
      height: 200px;
      margin-bottom: 1.6em;
      }
      #header .image.avatar img {
        width: 100%;
        object-fit: cover;
        border-radius: 50%;
      }

      #main .container p,
      #main .container li {
        font-size: var(--base-font-size);
        line-height: var(--base-line-height);
      }
      #main .container h2 {
        font-size: 1.6em;
      }
      #main .container h3 {
        font-size: 1.2em;
        margin-top: 1em;
      }

      .experience-item {
        margin-bottom: 8px;
        padding-bottom: 4px;
        border-bottom: 1px solid #ccc;
      }
      .experience-item:last-child {
        border-bottom: none;
      }
      .experience-item .job-title {
        display: block;
        font-weight: bold;
        font-size: var(--base-font-size);
        line-height: var(--base-line-height);
        margin-bottom: 2px;
      }
      .experience-item .organization {
        display: block;
        font-size: var(--base-font-size);
        line-height: var(--base-line-height);
        margin-bottom: 2px;
      }
      .experience-item .timeframe {
        display: block;
        font-size: var(--base-font-size);
        line-height: var(--base-line-height);
        color: #555;
        margin-bottom: 2px;
      }

      #other .image.main img {
        width: 60%;
        display: block;
        margin: 0 auto;
      }
      #main > section > .container {
      /* grow with viewport, but stay readable */
      width: min(100%, clamp(40rem, 70vw, 64rem));
      /* preserve the existing side gutter behavior */
      max-width: calc(100% - 4.5em);
      }
    </style>
  </head>
  <body class="is-preload">

    <!-- Header -->
    <section id="header">
      <header>
        <span class="image avatar"><img src="me.jpg" alt="Benjamin Plaut" /></span>
        <h1 id="logo">Benjamin Plaut</h1>
        <p>Member of Technical Staff at FAR.AI</p>
	<p style="margin-top:0;">AI Safety</p>
	<p style="font-style:normal;font-size:var(--base-font-size);margin-top:0.5em;">
        <a href="mailto:bjplaut@gmail.com">bjplaut@gmail.com</a> | <a href="CV.pdf">CV</a> 
		| <a href="https://scholar.google.com/citations?user=6ndk_nAAAAAJ">Scholar</a>
</p>
      </header>
      <nav id="nav">
        <ul>
          <li><a href="#about">About me</a></li>
          <li><a href="#publications">Publications</a></li>
          <li><a href="#experience">Work experience</a></li>
  		  <li><a href="#mentees">Mentees</a></li>			
          <li><a href="#other">Other</a></li>
        </ul>
      </nav>
      <footer>
        <ul class="icons">
          <li>
            <a href="mailto:bjplaut@gmail.com" class="icon solid fa-envelope">
              <span class="label">Email</span>
            </a>
          </li>
          <li>
            <a href="CV.pdf" class="icon solid fa-file">
              <span class="label">CV</span>
            </a>
          </li>
        </ul>
      </footer>
    </section>

    <!-- Wrapper -->
    <div id="wrapper">

      <!-- Main -->
      <div id="main">

        <!-- Research Interests Section -->
        <section id="about">
          <div class="container">
            <h2>About me</h2>
            <p>
              I am a Member of Technical Staff at <a href="https://far.ai">FAR.AI</a>. I was previously a postdoc at the 
              <a href="https://humancompatible.ai/">Center for Human-Compatible AI</a> (CHAI) at <a href="https://www.berkeley.edu/">UC Berkeley</a>,
              mentored by <a href="http://people.eecs.berkeley.edu/~russell/">Stuart Russell</a>.
              In 2021, I received my PhD from the 
              <a href="http://www-cs.stanford.edu/">Computer Science Department</a> at 
              <a href="https://www.stanford.edu/">Stanford University</a>, advised by 
              <a href="http://web.stanford.edu/~ashishg/">Ashish Goel</a> and supported by an NSF Graduate Research Fellowship.
              I also spent two years doing science and product work at Lyft.
            </p>


            <p>
              I want to use my career to do good in the world. During my PhD, I designed markets and algorithms that achieve fair outcomes even when people are selfish.
              As AI capabilities advanced, I became increasingly concerned about risks from AI and shifted my research to AI safety.
				I'm concerned about a wide range of risks from AI, including but not limited to loss of control, facilitating malicious actors (especially for cyberattacks and CBRN attacks),
	       concentration of power, exacerbation of societal inequalities, and economic disruption. My research is primarily relevant to the first two.
            </p>
            <h3>Research interests</h3>
            <p>
              My current focus is evaluation awareness: when models recognize that they’re being tested. Current safety assessments rely heavily on behavioral evaluations, 
				and if models behave differently when they know they’re being tested, these assessments could be systematically misleading. My goal is to determine the extent
				of this issue and its implications for AI safety. I'm more focused on measuring the impact of eval awareness on behavior, rather than the 
				prevalence of eval awareness. 
			</p>
			<p>
				My previous work centered on generalization: how a model handles unfamiliar inputs. I think that many types of safety failures can be framed as misgeneralization.
              I especially focused on training models to recognize when they're in unfamiliar situations and behave cautiously if so (e.g., ask for help). I studied this topic
			  in the context of LLMs, deep RL in video games, and theory. See this <a href="generalization_and_caution_agenda.pdf">meta-paper</a> for an overview of my generalization work.
				I also did some work on optimization pressure in post-training.
            </p>
		</div>
	</section>


        <!-- Publications Section -->
<section id="publications">
  <div class="container">
    <h2>Publications</h2>
    <p style="margin-bottom:0;">Lead author(s): * | Senior author(s): &dagger;</p>

    <!-- AI Safety Subsection -->
    <h3>AI Safety</h3>
    <ol style="margin-bottom:0;">
      <li>
	<a href="https://www.jmlr.org/papers/v27/25-2248.html">Safe Learning Under Irreversible Dynamics via Asking for Help</a><br>
	<strong>Benjamin Plaut*</strong>, Juan Li&eacute;vano-Karim, Hanlin Zhu, Stuart Russell&dagger;. <br>
	      Journal of Machine Learning Research (JMLR), 2026.
      </li>				
      <li>
	<a href="https://arxiv.org/pdf/2510.14884">Learning When Not to Learn: Risk-Sensitive Abstention in Bandits with Unbounded Rewards</a><br>
	Sarah Liaw*, <strong>Benjamin Plaut*</strong>. <br>
	      International Conference on Artificial Intelligence and Statistics (AISTATS), 2026.
      </li>				
      <li>
	<a href="https://openreview.net/forum?id=YOE0TRK8oU">YRC-Bench: A Benchmark for Learning to Coordinate with Experts</a><br>
	Mohamad H. Danesh*, Nguyen X. Khanh, Tu Trinh, <strong>Benjamin Plaut</strong>. <br>
	      Transactions on Machine Learning Research (TMLR), 2025.
      </li>
      <li>
	<a href="https://openreview.net/pdf?id=ALwpW0ygzI">Check Yourself Before You Wreck Yourself: Selectively Quitting Improves LLM Agent Safety</a><br>
	Vamshi Krishna Bonagiri*, Ponnurangam Kumaraguru, Nguyen X. Khanh, <strong>Benjamin Plaut&dagger;</strong>. <br>
	      Neural Information Processing Systems (NeurIPS) 2025 Workshops on Reliable and Regulatable ML.
      </li>	
      <li>
        <a href="https://arxiv.org/pdf/2402.08062.pdf">Avoiding Catastrophe in Online Learning by Asking for Help</a><br>
        <strong>Benjamin Plaut*</strong>, Hanlin Zhu, Stuart Russell&dagger;. <br>
        International Conference on Machine Learning (ICML), 2025.
      </li>	 		
      <li>
        <a href="https://openreview.net/pdf?id=E6LOh5vz5x">Probabilities of Chat LLMs Are Miscalibrated but Still Predict Correctness on Multiple-Choice Q&amp;A</a><br>
        <strong>Benjamin Plaut*</strong>, Nguyen X. Khanh, Tu Trinh. <br>
        Transactions on Machine Learning Research (TMLR), 2025.
      </li>	    
      <li>
        <a href="https://arxiv.org/pdf/2410.21052">Getting By Goal Misgeneralization With a Little Help From a Mentor</a><br>
        Tu Trinh*, Mohamad Danesh, Nguyen X. Khanh, <strong>Benjamin Plaut&dagger;</strong>. <br>
        Neural Information Processing Systems (NeurIPS) 2024 Workshop on Safe and Trustworthy Agents.
      </li>
	  <li>
		<a href="https://arxiv.org/pdf/2603.02229">Safety Training May Persist Through Helpfulness Optimization in LLM Agents</a><br>
		<strong>Benjamin Plaut</strong>.<br>
		Preprint.
	  </li>

    </ol>

    <!-- Resource Allocation Subsection -->
    <h3>Resource Allocation Algorithms</h3>
    <ol style="margin-bottom:0;">
      <li>
        <a href="BenPlautThesisFinal.pdf">Algorithms for Fair Public and Private Resource Allocation</a><br>
        <strong>Benjamin Plaut</strong>.
        PhD dissertation, 2021.
      </li>
      <li>
        <a href="https://arxiv.org/pdf/2009.09351.pdf">Counteracting Inequality in Markets via Convex Pricing</a><br>
        Ashish Goel&dagger;, <strong>Benjamin Plaut*</strong>. 
        Conference on Web and Internet Economics (WINE) 2020.
      </li>
      <li>
        <a href="https://arxiv.org/pdf/2009.09336.pdf">Almost Envy-free Repeated Matching in Two-sided Markets</a><br>
        Sreenivas Gollapudi&dagger;, Kostas Kollias, <strong>Benjamin Plaut*</strong>. 
        Conference on Web and Internet Economics (WINE) 2020.
      </li>
      <li>
        <a href="https://arxiv.org/pdf/1904.03322.pdf">Optimal Nash Equilibria for Bandwidth Allocation</a><br>
        <strong>Benjamin Plaut*</strong>. 
        Conference on Web and Internet Economics (WINE) 2020.
      </li>
      <li>
        <a href="equality_of_power.pdf">Equality of Power and Fair Public Decision-making</a><br>
        Nicole Immorlica&dagger;, <strong>Benjamin Plaut*</strong>, E. Glen Weyl&dagger;. 
        Conference on Web and Internet Economics (WINE) 2019.
      </li>
      <li>
        <a href="https://arxiv.org/pdf/1807.05293.pdf">Markets Beyond Nash Welfare for Leontief Utilities</a><br>
        Ashish Goel&dagger;, Reyna Hulett, <strong>Benjamin Plaut*</strong>. 
        Conference on Web and Internet Economics (WINE) 2019.
      </li>
      <li>
        <a href="CC.pdf">Communication Complexity of Discrete Fair Division</a><br>
        <strong>Benjamin Plaut*</strong>, Tim Roughgarden&dagger;. 
        Symposium on Discrete Algorithms (SODA), 2019; SIAM Journal on Computing (SICOMP), 2020.
      </li>
      <li>
        <a href="https://arxiv.org/pdf/1807.10836.pdf">Markets for Public Decision-making</a><br>
        Nikhil Garg, Ashish Goel&dagger;, <strong>Benjamin Plaut*</strong>. 
        Conference on Web and Internet Economics (WINE) 2018.
      </li>
      <li>
        <a href="EFX.pdf">Almost Envy-Freeness with General Valuations</a><br>
        <strong>Benjamin Plaut*</strong>, Tim Roughgarden&dagger;. 
        Symposium on Discrete Algorithms (SODA), 2018; SIAM Journal on Discrete Mathematics (SIDMA), 2020.
      </li>
      <li>
        <a href="seniorThesis.pdf">Algorithms for Social Good: Kidney Exchange</a><br>
        <strong>Benjamin Plaut</strong>. 
        Undergraduate honors thesis, 2016. Won the Allen Newell Award for Excellence in Undergraduate Research (best thesis in Computer Science).
      </li>
      <li>
        <a href="http://arxiv.org/abs/1606.00117">Hardness of the Pricing Problem in Barter Exchanges</a><br>
        <strong>Benjamin Plaut*</strong>, John P. Dickerson, Tuomas Sandholm&dagger;. 
        Preprint.
      </li>
      <li>
        <a href="https://arxiv.org/pdf/1606.01623v2.pdf">Position-Indexed Formulations for Kidney Exchange</a><br>
        John P. Dickerson, David Manlove&dagger;, <strong>Benjamin Plaut</strong>, Tuomas Sandholm&dagger;, and John Trimble*. 
        Economics and Computation (EC), 2016.
      </li>
      <li>
        <a href="http://jpdickerson.com/pubs/plaut16fast.pdf">Fast Optimal Clearing of Capped-Chain Barter Exchanges</a><br>
        <strong>Benjamin Plaut*</strong>, John P. Dickerson, Tuomas Sandholm&dagger;. 
        AAAI Conference on Artificial Intelligence, 2016.
      </li>
    </ol>

    <!-- Miscellaneous Subsection -->
    <h3>Physical Chemistry</h3>
    <ol style="margin-bottom:0;">
      <li>
        <a href="http://pubs.acs.org/doi/abs/10.1021/acs.jpcb.6b10362">Direct Observation of Folding Energy Landscape of RNA Hairpin at Mechanical Loading Rates</a><br>
        Huizhong Xu*, <strong>Benjamin Plaut</strong>, Xiran Zhu, Maverick Chen, Udit Mavinkurve, Anindita Maiti, Guangtao Song, Krishna Murari, and Maumita Mandal&dagger;. 
        <em>The Journal of Physical Chemistry</em>, 2017.
      </li>
    </ol>
  </div>
</section>

        <!-- Work Experience Section -->
        <section id="experience">
          <div class="container">
            <h2>Work experience</h2>
            <div class="experience-item">
              <span class="job-title">Member of Technical Staff</span>
              <span class="organization">FAR.AI</span>
              <span class="timeframe">August 2026 – present</span>
            </div>
            <div class="experience-item">
              <span class="job-title">Postdoctoral Researcher</span>
              <span class="organization">UC Berkeley</span>
              <span class="timeframe">September 2023 – August 2026</span>
            </div>
            <div class="experience-item">
              <span class="job-title">Data Scientist</span>
              <span class="organization">Lyft</span>
              <span class="timeframe">June 2021 – May 2023</span>
            </div>
            <div class="experience-item">
              <span class="job-title">Research Intern</span>
              <span class="organization">Google</span>
              <span class="timeframe">Summer 2019, Summer 2020</span>
            </div>
          </div>
        </section>

        <!-- Mentees -->
        <section id="mentees">
        <div class="container">
        	<h2>Research mentees</h2>	
            <ul>
              <li>
                Sarah Liaw (2025), now a PhD student at Harvard
              </li>
	            <li>
				Pavel Czempin (2025), now a PhD student at USC
		            </li>
              <li>
				Vamshi Krishna Bonagiri (2025), now a PhD student at MBZUAI
              </li>
			  <li>
				Juan Li&eacute;vano-Karim (2024), now a Teaching Professor at Universidad de los Andes
			  </li>
			  <li>
				Tu Trinh (2024), now a Machine Learning Research Engineer at Scale AI  
			  </li>
            </ul>
          </div>
        </section>			
		  
        <!-- Personal Section -->
        <section id="other">
          <div class="container">
            <h2>Other things about me</h2>
            <ul>
              <li>
                I speak English (native) and Spanish (advanced proficiency).
              </li>
	            <li>
		      I wrote an algorithmic art-generator called <a href="https://github.com/bplaut/RAMbrandt">RAMbrandt</a> as a personal project in college.
		            </li>
              <li>
                I've composed and produced music. Check out my <a href="https://open.spotify.com/artist/3uPD6HhfJygQOVj9Em030P">Spotify</a>!
              </li>
            </ul>
          </div>
        </section>

      </div>

      <!-- Footer -->
      <section id="footer">
        <div class="container">
          <ul class="copyright">
            <li>&copy; Benjamin Plaut. Website last updated August 2026.</li>
          </ul>
        </div>
      </section>
    </div>
    <!-- Scripts -->
    <script src="assets/jquery.min.js"></script>
    <script src="assets/jquery.scrollex.min.js"></script>
    <script src="assets/jquery.scrolly.min.js"></script>
    <script src="assets/browser.min.js"></script>
    <script src="assets/breakpoints.min.js"></script>
    <script src="assets/util.js"></script>
    <script src="assets/main.js"></script>
    <!-- Add this block to force a scroll event after smooth scrolling -->
    <script>
      $(document).ready(function() {
        $('#nav a').off('click').scrolly({
          speed: 1000,
          callback: function() {
            $(window).trigger('scroll');
          }
        });
      });
    </script>
  </body>
</html>