{"id":16416,"date":"2025-06-02T09:07:36","date_gmt":"2025-06-02T13:07:36","guid":{"rendered":"https:\/\/www.bu.edu\/cds-faculty\/?p=16416"},"modified":"2025-06-02T09:12:46","modified_gmt":"2025-06-02T13:12:46","slug":"pacchiano-new-method-adaptive-llms","status":"publish","type":"post","link":"https:\/\/www.bu.edu\/cds-faculty\/2025\/06\/02\/pacchiano-new-method-adaptive-llms\/","title":{"rendered":"Teaching AI to Personalize: Aldo Pacchiano Introduces a New Method for Adaptive Large Language Models"},"content":{"rendered":"

As large language models (LLMs) like ChatGPT become embedded in everyday life\u2014from drafting emails to debugging code\u2014the expectation that they understand how different people want to be helped is growing. Yet despite their sophistication, most of these models still treat all users more or less the same. This is where Aldo Pacchiano<\/a>, Assistant Professor at Boston University\u2019s Faculty of Computing and Data Sciences (CDS), saw a problem worth solving. \u201cCurrent models optimize for what most people tend to like,\u201d he explains, \u201cbut they\u2019re not actually learning what you like.\u201d<\/p>\n

\"Aldo
Aldo Pacchiano, Assistant Professor of Computing & Data Sciences at Boston University<\/figcaption><\/figure>\n

In a new paper<\/a> titled \u201cLanguage Model Personalization via Reward Factorization,\u201d Pacchiano and his co-authors introduce Personalization via Reward Factorization (PReF): a framework that personalizes LLM responses to individual users without requiring extensive retraining or massive datasets. Instead of averaging preferences across millions of users, PReF gradually learns which traits the user values most, like humor, brevity, or formality. With just a handful of these comparisons, it builds a personalized profile that guides the model\u2019s future responses to better match the individual\u2019s style. This approach offers a more flexible and efficient alternative to default systems like GPT-4o and moves toward AI that could better understand and serve each individual, not just the average user.<\/p>\n

To test how well this lightweight personalization could work, Pacchiano and his team first trained PReF on synthetic users, essentially role-played personalities created by prompting a language model to favor specific traits. Each simulated user chose between dozens of response pairs, gradually revealing their underlying preferences. \u201cIt\u2019s a trick,\u201d he explains, \u201cbut it works. You tell the model, \u2018Pretend you\u2019re a user who likes things this way,\u2019 and that generates reliable training data.\u201d Once PReF could consistently adapt to these synthetic users, the team moved on to a more difficult benchmark: real people. Using the PRISM dataset, which includes thousands of user profiles and prompt\u2013response interactions, they tested whether the model could still match individuals\u2019 preferences even when the traits weren\u2019t cleanly labeled or pre-defined. The results were promising: with just 10 to 20 response comparisons, PReF could generate replies that real users preferred over GPT-4o\u2019s standard outputs.<\/p>\n

While general-purpose tools like ChatGPT allow users to adjust tone or style by tweaking their prompts, Pacchiano believes this places too much burden on the user. In contrast, he sees PReF as especially useful in applications where users either don\u2019t know how to steer the model or won\u2019t interact with it frequently enough to learn how. \u201cIf you\u2019re deploying a customer support bot, or even an internal tool like a Slack assistant, the user might only interact with it once,\u201d he notes. \u201cBut in that short time, the system still needs to feel helpful, efficient, and intuitive.\u201d In these settings, PReF\u2019s ability to adapt automatically, without explicit instruction, offers a major advantage. Whether the user prefers long explanations or direct answers, formal language or casual phrasing, the model can learn and adjust accordingly, even with minimal input. This opens the door to more responsive AI agents in education, workplace tools, healthcare communication, and other domains where personalization can directly improve trust, usability, and satisfaction.<\/p>\n

What this work ultimately raises is a deeper question: if AI is going to be everywhere, should it learn to meet us more than halfway? For Pacchiano, personalization is a shift in how we think about human-AI interaction. His next steps involve extending this adaptability beyond style and tone, toward reasoning and decision-making in real-world settings. He imagines systems that don\u2019t just tailor their language, but learn how to move through space, solve unfamiliar problems, and adapt with limited feedback. That may sound far from a simple pairwise comparison, but the principle is the same. If AI can start by listening more closely to how we prefer to be answered, perhaps it can also learn how we prefer to be helped.<\/p>\n

– Neeza Singh (CDS’25), CDS Research Communications Intern<\/span><\/p>\n

 <\/p>\n","protected":false},"excerpt":{"rendered":"

Check out the recently published paper from CDS Assistant Professor Aldo Pacchiano’s lab based on improved personalizations of LLMs.<\/p>\n","protected":false},"author":18839,"featured_media":16631,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[251,317,230,244],"tags":[34,379,377,279,22,709,803],"_links":{"self":[{"href":"https:\/\/www.bu.edu\/cds-faculty\/wp-json\/wp\/v2\/posts\/16416"}],"collection":[{"href":"https:\/\/www.bu.edu\/cds-faculty\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.bu.edu\/cds-faculty\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.bu.edu\/cds-faculty\/wp-json\/wp\/v2\/users\/18839"}],"replies":[{"embeddable":true,"href":"https:\/\/www.bu.edu\/cds-faculty\/wp-json\/wp\/v2\/comments?post=16416"}],"version-history":[{"count":4,"href":"https:\/\/www.bu.edu\/cds-faculty\/wp-json\/wp\/v2\/posts\/16416\/revisions"}],"predecessor-version":[{"id":16632,"href":"https:\/\/www.bu.edu\/cds-faculty\/wp-json\/wp\/v2\/posts\/16416\/revisions\/16632"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.bu.edu\/cds-faculty\/wp-json\/wp\/v2\/media\/16631"}],"wp:attachment":[{"href":"https:\/\/www.bu.edu\/cds-faculty\/wp-json\/wp\/v2\/media?parent=16416"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.bu.edu\/cds-faculty\/wp-json\/wp\/v2\/categories?post=16416"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.bu.edu\/cds-faculty\/wp-json\/wp\/v2\/tags?post=16416"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}