{"682735":{"#nid":"682735","#data":{"type":"event","title":"PhD Defense by  Alex Havrilla","body":[{"value":"\u003Cp\u003EAlex Havrilla\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003ETitle: Towards a Theory and Practice of Open-ended Reasoning with Generative Models\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003EDate: 6\/13\/2025\u003C\/p\u003E\u003Cp\u003ETime: 1 PM\u003C\/p\u003E\u003Cp\u003ELocation:\u003C\/p\u003E\u003Cp\u003E-\u0026nbsp;In-person: Skiles 202\u003C\/p\u003E\u003Cp\u003E-\u0026nbsp;Remote: \u003Ca href=\u0022https:\/\/gatech.zoom.us\/j\/5472104648\u0022\u003Ehttps:\/\/gatech.zoom.us\/j\/5472104648\u003C\/a\u003E\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003EAlexander Havrilla\u003C\/p\u003E\u003Cp\u003EMachine Learning PhD Student\u003C\/p\u003E\u003Cp\u003ESchool of Mathematics\u003C\/p\u003E\u003Cp\u003EGeorgia Institute of Technology\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003ECommittee\u003C\/p\u003E\u003Cp\u003E1 Dr. Wenjing Liao, School of Mathematics (Advisor), Georgia Tech\u003C\/p\u003E\u003Cp\u003E2 Dr. Mark Riedl, School of Interactive Computing, Georgia Tech\u003C\/p\u003E\u003Cp\u003E3 Dr. Tuo Zhao, School of Industrial and Systems Engineering, Georgia Tech\u003C\/p\u003E\u003Cp\u003E4 Dr. Jacob Abernethy, School of Interactive Computing, Georgia Tech\u003C\/p\u003E\u003Cp\u003E5 Dr. David Alvarez-Melis, School of Engineering and Applied Sciences, Harvard\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003EAbstract\u003C\/p\u003E\u003Cp\u003EDriven by advancements in large language modeling (LLMs), the last several years have seen an explosion in AI reasoning capability. In this dissertation, we characterize two distinct types of reasoning: closed-ended reasoning versus open-ended reasoning. We define closed-ended reasoning as the systematic application of a defined set of rules to reach a desired outcome. In contrast, we describe open-ended reasoning as a less structured process, often requiring the creation or adaptation of new rule sets themselves, and characterized by a greater need for exploration and discovery. While LLMs increasingly excel at closed-ended reasoning, they struggle more with problems requiring the open-ended counterpart. We study both types of reasoning in three parts. First, by establishing novel approximation and statistical theory for LLMs. This theory elucidates data complexity as a driving factor behind scaling laws, which themselves have a strong downstream effect on reasoning ability. Then, to improve reasoning ability in practice, we develop a novel RL framework for LLMs, trlX, which is used to fine-tune LLMs on reasoning problems. Our analysis reveals the exploration ability of LLMs as a key bottleneck to future improvement via RL. This leads us to propose SPARQ: a self-improvement style synthetic data generation algorithm drawing on techniques from the quality-diversity (QD) literature to improve both the correctness and diversity of LLM reasoning. We conclude by discussing open problems and future directions for better open-ended AI reasoning.\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E","summary":"","format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003ETowards a Theory and Practice of Open-ended Reasoning with Generative Models\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"Towards a Theory and Practice of Open-ended Reasoning with Generative Models"}],"uid":"27707","created_gmt":"2025-06-09 15:53:42","changed_gmt":"2025-06-09 15:54:43","author":"Tatianna Richardson","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2025-06-13T13:00:00-04:00","event_time_end":"2025-06-16T15:00:00-04:00","event_time_end_last":"2025-06-16T15:00:00-04:00","gmt_time_start":"2025-06-13 17:00:00","gmt_time_end":"2025-06-16 19:00:00","gmt_time_end_last":"2025-06-16 19:00:00","rrule":null,"timezone":"America\/New_York"},"location":"Skiles 202","extras":[],"groups":[{"id":"221981","name":"Graduate Studies"}],"categories":[],"keywords":[{"id":"100811","name":"Phd Defense"}],"core_research_areas":[],"news_room_topics":[],"event_categories":[{"id":"1788","name":"Other\/Miscellaneous"}],"invited_audience":[{"id":"78771","name":"Public"}],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[],"email":[],"slides":[],"orientation":[],"userdata":""}}}