{"693576":{"#nid":"693576","#data":{"type":"event","title":"PhD Defense by ShengYun Anthony Peng","body":[{"value":"\u003Cp\u003E\u003Cstrong\u003ETitle:\u003C\/strong\u003E Beyond capability: safe \u0026amp; reliable foundation models at scale\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EDate:\u003C\/strong\u003E Tuesday, November 17, 2026\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003ETime:\u003C\/strong\u003E 12:00 PM to 2:00 PM Eastern Time (US)\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EJoin in Person:\u003C\/strong\u003E\u003Ca href=\u0022https:\/\/nam12.safelinks.protection.outlook.com\/?url=https%3A%2F%2Fmaps.app.goo.gl%2FUgDFKRffYwwDi6tj7\u0026amp;data=05%7C02%7Cannouncements%40grad.gatech.edu%7C4dcba29355b3416d7a2908df344fb509%7C482198bbae7b4b258b7a6d7f32faa083%7C1%7C0%7C639287189083199878%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C\u0026amp;sdata=0sfDlWIKPFR4y3IEbEiyW70CS2L1jhvQlCX29pDC2Tc%3D\u0026amp;reserved=0\u0022\u003E\u0026nbsp;\u003C\/a\u003E\u003Ca href=\u0022https:\/\/nam12.safelinks.protection.outlook.com\/?url=https%3A%2F%2Fmaps.app.goo.gl%2FUgDFKRffYwwDi6tj7\u0026amp;data=05%7C02%7Cannouncements%40grad.gatech.edu%7C4dcba29355b3416d7a2908df344fb509%7C482198bbae7b4b258b7a6d7f32faa083%7C1%7C0%7C639287189083229719%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C\u0026amp;sdata=cBlPczRRB517rLuZZvT8G2Hq6ntPj0F5wEYUBOA8Aj8%3D\u0026amp;reserved=0\u0022\u003ECODA\u003C\/a\u003E 114 (first floor conference room; just walk in, no special access needed)\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EJoin Virtually:\u003C\/strong\u003E\u003Ca href=\u0022https:\/\/nam12.safelinks.protection.outlook.com\/?url=https%3A%2F%2Fgatech.zoom.us%2Fj%2F97908941560%3Fpwd%3DNozeHblEgWLIZ4ZoukMnCBhQk9BuiV.1\u0026amp;data=05%7C02%7Cannouncements%40grad.gatech.edu%7C4dcba29355b3416d7a2908df344fb509%7C482198bbae7b4b258b7a6d7f32faa083%7C1%7C0%7C639287189083253920%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C\u0026amp;sdata=nsudNtQYnrIFK34Abhi0DwdfVu8oB1SuKqMSSz7c6%2Fs%3D\u0026amp;reserved=0\u0022\u003E\u0026nbsp;\u003C\/a\u003E\u003Ca href=\u0022https:\/\/nam12.safelinks.protection.outlook.com\/?url=https%3A%2F%2Fgatech.zoom.us%2Fj%2F97908941560%3Fpwd%3DNozeHblEgWLIZ4ZoukMnCBhQk9BuiV.1\u0026amp;data=05%7C02%7Cannouncements%40grad.gatech.edu%7C4dcba29355b3416d7a2908df344fb509%7C482198bbae7b4b258b7a6d7f32faa083%7C1%7C0%7C639287189083277543%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C\u0026amp;sdata=xy4MDaf0vMefeMPPlUGU9L0yQK1IOHllG7FIe6RKgns%3D\u0026amp;reserved=0\u0022 title=\u0022https:\/\/gatech.zoom.us\/j\/97908941560?pwd=NozeHblEgWLIZ4ZoukMnCBhQk9BuiV.1\u0022\u003EZoom\u003C\/a\u003E\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EShengYun Anthony Peng\u003C\/strong\u003E\u003C\/p\u003E\u003Cp\u003E\u003Ca href=\u0022https:\/\/nam12.safelinks.protection.outlook.com\/?url=https%3A%2F%2Fshengyun-peng.github.io%2F\u0026amp;data=05%7C02%7Cannouncements%40grad.gatech.edu%7C4dcba29355b3416d7a2908df344fb509%7C482198bbae7b4b258b7a6d7f32faa083%7C1%7C0%7C639287189083301272%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C\u0026amp;sdata=hLkLIrnBRuKX4WZ4btLT%2FPWsmL16cEVT0P%2BIydCBa5M%3D\u0026amp;reserved=0\u0022\u003Eshengyun-peng.github.io\u003C\/a\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003ECS Ph.D. Candidate\u003C\/p\u003E\u003Cp\u003ESchool of Computational Science and Engineering\u003C\/p\u003E\u003Cp\u003ECollege of Computing\u003C\/p\u003E\u003Cp\u003EGeorgia Institute of Technology\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003ECommittee:\u003C\/strong\u003E\u003C\/p\u003E\u003Cp\u003EDr. Duen Horng (Polo) Chau \u2014 Advisor and Committee Chair, School of Computational Science and Engineering, Georgia Institute of Technology\u003C\/p\u003E\u003Cp\u003EDr. Wei Xu \u2014 School of Computer Science, Georgia Institute of Technology\u003C\/p\u003E\u003Cp\u003EDr. Alan Ritter \u2014 School of Interactive Computing, Georgia Institute of Technology\u003C\/p\u003E\u003Cp\u003EDr. Jianfeng Chi \u2014 Member of Technical Staff, OpenAI\u003C\/p\u003E\u003Cp\u003EDr. Pin-Yu Chen \u2014 Principal Research Scientist, IBM Research\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EAbstract:\u003C\/strong\u003E\u003C\/p\u003E\u003Cp\u003EFoundation models achieve remarkable benchmark performance, but growing capabilities do not guarantee safety, reliability, or efficiency in real-world deployment. My dissertation empowers foundation models to operate safely and reliably at scale through three complementary thrusts:\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003E(1) Aligning for Safety.\u003C\/strong\u003E We discover a universal\u0026nbsp;\u003Ca href=\u0022https:\/\/nam12.safelinks.protection.outlook.com\/?url=https%3A%2F%2Fshengyun-peng.github.io%2Fpapers%2Fllm-safety-landscape\u0026amp;data=05%7C02%7Cannouncements%40grad.gatech.edu%7C4dcba29355b3416d7a2908df344fb509%7C482198bbae7b4b258b7a6d7f32faa083%7C1%7C0%7C639287189083327025%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C\u0026amp;sdata=%2BqjGg6FOWN56%2F01oC3dCjo38KRUyRMiD91nXK1tW3hU%3D\u0026amp;reserved=0\u0022\u003E\u003Cem\u003Esafety basin\u003C\/em\u003E\u003C\/a\u003E phenomenon across popular open-source LLMs: a local parameter region preserving safety. This discovery, featured in the \u003Cem\u003EInternational AI Safety Report\u003C\/em\u003E, reveals why finetuning erodes alignment and inspires\u0026nbsp;\u003Ca href=\u0022https:\/\/nam12.safelinks.protection.outlook.com\/?url=https%3A%2F%2Fshengyun-peng.github.io%2Fpapers%2Fllm-safety-shaping\u0026amp;data=05%7C02%7Cannouncements%40grad.gatech.edu%7C4dcba29355b3416d7a2908df344fb509%7C482198bbae7b4b258b7a6d7f32faa083%7C1%7C0%7C639287189083350922%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C\u0026amp;sdata=Z6j36%2FApT7Z7OZP%2FfVFqboMYqIW%2FV%2BNnQ%2BDKnY9JXbs%3D\u0026amp;reserved=0\u0022\u003ESTAR-DSS\u003C\/a\u003E, our novel dynamic safety shaping method to mitigate this erosion. For safer reasoning, we introduce\u0026nbsp;\u003Ca href=\u0022https:\/\/nam12.safelinks.protection.outlook.com\/?url=https%3A%2F%2Fshengyun-peng.github.io%2Fpapers%2Flrm-safety\u0026amp;data=05%7C02%7Cannouncements%40grad.gatech.edu%7C4dcba29355b3416d7a2908df344fb509%7C482198bbae7b4b258b7a6d7f32faa083%7C1%7C0%7C639287189083374771%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C\u0026amp;sdata=k%2FlWXYdQA0Mxb5lwLg1lCEyYoV%2F6E4%2FJl2I6QhmjOz4%3D\u0026amp;reserved=0\u0022\u003ERECAP\u003C\/a\u003E, a principled reinforcement learning (RL) method for post-training that teaches models to recover from flawed thinking.\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003E(2) Designing for Reliability.\u003C\/strong\u003E Safe behavior must be paired with reliable predictions. In\u0026nbsp;\u003Ca href=\u0022https:\/\/nam12.safelinks.protection.outlook.com\/?url=https%3A%2F%2Fshengyun-peng.github.io%2Fpapers%2Frobust-principles\u0026amp;data=05%7C02%7Cannouncements%40grad.gatech.edu%7C4dcba29355b3416d7a2908df344fb509%7C482198bbae7b4b258b7a6d7f32faa083%7C1%7C0%7C639287189083398929%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C\u0026amp;sdata=YfVC1p0GWA%2FlxvoC%2F3h6EhzMEckKZXuB5BlQWkPLDfA%3D\u0026amp;reserved=0\u0022\u003E\u003Cem\u003ERobust Principles\u003C\/em\u003E\u003C\/a\u003E, we synthesize a suite of three generalizable robust architectural design principles for CNNs, achieving state-of-the-art (SOTA) adversarial robustness on RobustBench. This work earned the Best Poster (Paper) Award at BMVC and guided DARPA\u2019s research on adversarially robust architectures. These insights also enable accurate, efficient business table recognition with lightweight encoders.\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003E(3) Scaling Model Capabilities.\u003C\/strong\u003E Complementing architectural design, we unlock accurate table recognition with simpler encoders through self-supervised pretraining. Our\u0026nbsp;\u003Ca href=\u0022https:\/\/nam12.safelinks.protection.outlook.com\/?url=https%3A%2F%2Fshengyun-peng.github.io%2Fpapers%2Funitable\u0026amp;data=05%7C02%7Cannouncements%40grad.gatech.edu%7C4dcba29355b3416d7a2908df344fb509%7C482198bbae7b4b258b7a6d7f32faa083%7C1%7C0%7C639287189083422452%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C\u0026amp;sdata=LLF9F3W2GtVB9vmjE%2FqygIKD9ya5YxkMck8O2laYnkQ%3D\u0026amp;reserved=0\u0022\u003EUniTable\u003C\/a\u003E training framework unifies both the training paradigm and training objective of table recognition, achieving SOTA performance. UniTable (530+ GitHub stars) powers table extraction in Open Parse (3.2K GitHub stars) and is used by startups for retrieval-augmented generation (RAG). This research has been selected for oral presentations and presented in invited talks. To scale visual understanding efficiently, we establish a principled framework revealing optimal inference compute allocations for video VLMs.\u003C\/p\u003E\u003Cp\u003EMy dissertation research significantly impacts academia, industry, and society, advancing foundation models through collaborations with NVIDIA, Meta, IBM, Intel, and ADP. It has contributed to the AI foundations of industry research grants totaling over $1.7 million.\u003C\/p\u003E","summary":"","format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003EBeyond capability: safe \u0026amp; reliable foundation models at scale\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"Beyond capability: safe \u0026 reliable foundation models at scale"}],"uid":"27707","created_gmt":"2026-10-28 15:53:23","changed_gmt":"2026-10-28 15:53:59","author":"Tatianna Richardson","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2026-11-17T12:00:00-05:00","event_time_end":"2026-11-17T14:00:00-05:00","event_time_end_last":"2026-11-17T14:00:00-05:00","gmt_time_start":"2026-11-17 17:00:00","gmt_time_end":"2026-11-17 19:00:00","gmt_time_end_last":"2026-11-17 19:00:00","rrule":null,"timezone":"America\/New_York"},"location":"CODA 114 (first floor conference room","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":""}}}