Andy Liu

I am a third-year PhD student at the Language Technologies Institute within Carnegie Mellon University’s School of Computer Science, advised by Mona Diab and Daniel Fried. My research is supported by the NSF Graduate Fellowship.

I’m broadly interested in value alignment and multi-agent interaction, especially in the context of language models. Some specific subquestions I’m currently thinking about include:

How can we robustly specify and shape LLMs’ values and character?
What are the best ways to train agents to collaborate with human task partners?
How can we effectively oversee real-world interactions between many LLM agents?

Please reach out (email is best) if you’re interested in my work or in collaborating!

Outside of research, I enjoy running, bouldering, and writing.

'; function previousTooltipHtml(lines) { if (!lines || !lines.length) return null; var rows = lines .map(escapeHtml) .map(function (line) { return ''; }) .join(''); return 'Previously:' + rows; } document.addEventListener('DOMContentLoaded', function () { if (typeof jQuery === 'undefined') return; var $ = jQuery; const contentDiv = document.querySelector('.clearfix'); if (!contentDiv) return; const paragraphs = contentDiv.querySelectorAll('p'); if (paragraphs.length === 0) return; const lastParagraph = paragraphs[paragraphs.length - 1]; const bookTitle = "Accelerando"; const bookAuthor = "Charles Stross"; const previousBookTitlesRaw = ["A Theory of Justice","Abundance","Too Like the Lightning"] || []; const hasLastfm = true; const previousArtists = ["glaive","Baby Keem","LucianBlomkamp"] || []; const artistName = "Denzel Curry"; const readingLead = " I am currently reading "; const previousBookTitles = dedupePreservingOrder(previousBookTitlesRaw).filter(function (t) { return t && t !== bookTitle; }); const bookTip = previousTooltipHtml(previousBookTitles); const $bookPhrase = $(''); if (bookTip) { $bookPhrase.attr('tabindex', '0'); $bookPhrase.addClass('reading-hover'); $bookPhrase.tooltip({ html: true, title: bookTip, template: readingTooltipTemplate, container: 'body', }); } $bookPhrase.append(document.createTextNode(bookTitle)); if (bookAuthor) { $bookPhrase.append(document.createTextNode(' by ' + bookAuthor)); } let $artist = null; if (hasLastfm && artistName) { const artistLines = dedupePreservingOrder(previousArtists).filter(function (a) { return a !== artistName; }); const artistTip = previousTooltipHtml(artistLines); $artist = $('').text(artistName); if (artistTip) { $artist.addClass('reading-hover'); $artist.tooltip({ html: true, title: artistTip, template: readingTooltipTemplate, container: 'body', }); } } $(lastParagraph).append(document.createTextNode(readingLead)); $(lastParagraph).append($bookPhrase); if ($artist) { $(lastParagraph).append(document.createTextNode(' and listening to ')); $(lastParagraph).append($artist); } $(lastParagraph).append(document.createTextNode('.')); }); })();

news

Jan 31, 2026	Generative Value Conflicts Reveal LLM Priorities was accepted to ICLR 2026. See you in Brazil!
Aug 25, 2025	I’ve joined Meta FAIR in Seattle as a research intern, working with Ruta Desai on training collaborative agents.
Jun 13, 2025	I was awarded a 2025 NSF Graduate Research Fellowship.
Dec 19, 2024	Our paper on Dynamic Coalition Structure Detection in Natural-Language-based Interactions was accepted to AAMAS 2025. See you in Detroit!
Aug 11, 2024	I presented my first PhD paper, Evaluating Large Language Model Biases in Persona-Steered Generation, at ACL 2024 in Bangkok.