Seminar Series

Event Details

• Time: 10:00 a.m. - 11:30 a.m. Eastern Time
• Format: In person and Zoom
• Location: Academic Building One, Room 5150

Abstract

Reinforcement learning (RL) has become a central approach to post-training large language models, driving advances in chain-of-thought reasoning and agentic problem solving. A striking feature of this approach is that rewards based only on final outcomes can improve reasoning across many intermediate steps without explicitly supervising those steps. This raises fundamental questions about how sparse feedback guides learning, produces complex behaviors, and supports generalization.

This talk presents recent work toward a theory of reinforcement learning for reasoning in transformers, revealing how training dynamics give rise to multi-step reasoning and search capabilities. It identifies curriculum structure as a key factor in progressing from simple skills to complex behaviors, including an implicit curriculum that emerges when training on mixed-difficulty problems. The work characterizes how the difficulty spectrum governs steady progress versus grokking-like transitions, with prolonged plateaus followed by abrupt improvements. It also explains how RL shapes specialized attention mechanisms that equip transformer-based policies with exploration, memory, and backtracking, yielding depth-first search strategies that generalize from shallow training trees to deeper ones. These findings connect curriculum structure, optimization dynamics, and attention mechanisms, providing new insight into how RL develops reasoning and agentic capabilities from sparse feedback.

Speaker Bio

Yingbin Liang is a Professor in the Department of Electrical and Computer Engineering at The Ohio State University and a core faculty member of the Translational Data Analytics Institute. She serves as Deputy Director of the NSF AI-EDGE Institute and Co-Lead for the Foundational AI Pillar of the OSU AI (X) Hub. She received her Ph.D. in electrical engineering from the University of Illinois Urbana-Champaign in 2005 and held faculty positions at the University of Hawaii and Syracuse University before joining Ohio State.

Her research lies at the intersection of machine learning, large-scale optimization, statistical signal processing, information theory, and wireless networks, with growing applications in other scientific domains. She received the National Science Foundation CAREER Award and the State of Hawaii Governor Innovation Award in 2009, as well as the EURASIP Best Paper Award in 2014. She is an IEEE Fellow and an IEEE Information Theory Society Distinguished Lecturer for 2026-2027.

Event Details

• Time: 4:00 p.m. Eastern Time
• Format: In person and Zoom
• Location: Hitt Hall, Room 335, 1385 Perry Street, Blacksburg, VA 24060

Abstract

This talk presents a fascinating account of how the fields of electrical engineering and computer science were invented. Designed for audiences ranging from first-year students to graduate researchers and professors, it shares stories of the invention, inspiration, and ingenuity that led to today's knowledge. The presentation draws from Rappaport's new textbook, which offers a distinctive approach to teaching first-year electrical engineering and computer science students by connecting the personal and professional lives of pioneering mathematicians and engineers with the discoveries that established the fundamental laws governing these fields. The book has been used at New York University for three years, has been adopted by Purdue University, and is scheduled for worldwide publication by Cambridge University Press.

Speaker Bio

Theodore (Ted) S. Rappaport is the David Lee/Ernst Weber Professor of Electrical Engineering at the NYU Tandon School of Engineering, a professor of computer science at New York University's Courant Institute of Mathematical Sciences, and a professor of radiology at the NYU School of Medicine. He is the founding director of NYU WIRELESS and previously founded the Mobile and Portable Radio Research Group (MPRG), now part of Wireless@Virginia Tech, and the Wireless Networking and Communications Group at the University of Texas at Austin. He has been elected to the American Academy of Arts and Sciences.

Rappaport has pioneered radio-wave propagation measurement and modeling, site-specific wireless network design, millimeter-wave and sub-THz communications, antennas, circuits, and beamforming. His research has influenced international wireless standards and helped advance technologies used in cellular networks, emergency location systems, and 5G products. He has served on the FCC Technological Advisory Council, conducted research for government agencies and telecommunications companies, authored or edited 24 books, and holds more than 100 U.S. or international patents issued or pending.

Rappaport founded TSR Technologies and Wireless Valley Communications, both of which commercialized wireless technologies. He received his B.S., M.S., and Ph.D. degrees in electrical engineering from Purdue University, where he is a Distinguished Engineering Alumnus and the Neal Armstrong Distinguished Visiting Professor. He is also a Hagler Fellow at Texas A&M University.

Event Details

• Time: 2:00 - 3:00 p.m. Eastern Time
• Format: In person and online via Teams
• Location: Arlington

Abstract

Internet of Things (IoT) devices and applications are becoming increasingly common, creating denser networks and larger attack surfaces that malicious users may exploit to compromise security and privacy. This talk first presents a protocol for IoT devices in 6G mobile networks that supports network densification and encrypted communication while enabling higher throughput. It then analyzes a Denial-of-Service threat to Wi-Fi 7 IoT networks and proposes a lightweight Authenticated Encryption with Associated Data (AEAD) framework to support fair medium access for IoT devices. The talk concludes with low-latency, reliable communication, presenting an open-loop proactive cell-association scheme in which Agentic AI exploits latent patterns in Wi-Fi communications to reduce each device's likelihood of outage through a Multi-Agent Proximal Partial Optimization framework.

Event Details

• Reception: 3:30 p.m., Room AB 2100
• Talk: 4:00 p.m., Room AB 2100
• Format: In person and online via Zoom
• Location: Virginia Tech Academic Building One (AB1), 3625 Potomac Avenue, Alexandria, VA 22305

Abstract

The potential for major new 6G spectrum allocations in the upper mid-band and THz ranges has intensified the need for reliable channel models within wireless standards bodies. Such models depend on large empirical, in-situ measurement campaigns across use cases and geographic regions, but those campaigns are expensive, time-consuming, and vary in quality depending on how the data is collected. This talk presents a standardized point-data format that allows empirical channel measurements to be collected, pooled, and reused by standards contributors. The format is designed to improve channel-model accuracy through larger shared data sets while enabling artificial intelligence and machine learning methods to synthesize richer channel models. The approach is demonstrated through a collaboration between New York University and the University of Southern California, where field measurements from different channel-sounder systems are pooled to produce a more representative Urban Microcell channel model that can be processed effectively by machines.

Speaker Bio

Theodore (Ted) S. Rappaport is the David Lee/Ernst Weber Professor of Electrical Engineering at the NYU Tandon School of Engineering, a professor of computer science at New York University's Courant Institute of Mathematical Sciences, and a professor of radiology at the NYU School of Medicine. He is the founding director of NYU WIRELESS, the first academic research center to combine engineering, computer science, and medicine, and he previously founded the Mobile and Portable Radio Research Group (MPRG), now part of Wireless@Virginia Tech.

Rappaport has pioneered radio wave propagation measurement and modeling for cellular and personal communications, millimeter-wave and sub-THz wireless systems, site-specific RF channel modeling, wireless network design, antennas, circuits, and beamforming. His work has influenced international wireless standards and helped advance the use of millimeter-wave and sub-THz spectrum for 5G and beyond. He has served on the FCC Technological Advisory Council, conducted research for government agencies and telecommunications companies, authored or edited 24 books, and holds more than 100 U.S. or international patents issued or pending. He received his B.S., M.S., and Ph.D. degrees in electrical engineering from Purdue University.

Event Details

• Time: 3:30 - 4:30 p.m.
• Format: In person and online via Zoom
• Location: Holtzman Alumni Center Auditorium

Abstract

The past year has brought major steps toward 6G. Beginning with a May 2024 workshop, 3GPP SA1 has been identifying use cases and requirements for new capabilities. In March 2025, 3GPP RAN held a radio-focused workshop, and technical studies are now underway. This presentation will examine the status of 5G, emerging capabilities, areas where 5G has and has not met expectations, and broader challenges facing cellular technology. Particular attention will be given to satellite integration through Non-Terrestrial Networks (NTN), Integrated Sensing and Communications (ISAC), and AI. The talk will also present a vision for technical enhancements that may be incorporated into 6G.

Speaker Bio

Dr. Edward G. Tiedemann, Jr. is a Qualcomm Fellow and Senior Vice President of Engineering at Qualcomm Technologies, Inc., where he leads Qualcomm's worldwide standardization and industry organization activities. He was instrumental in the design and development of the TIA/EIA/IS-95 CDMA system, also known as cdmaOne, and led Qualcomm's, and much of the industry's, work on the third-generation cdma2000 system.

Dr. Tiedemann also led the offline industry group on M2M convergence that resulted in oneM2M. He serves on the boards of the MulteFire Alliance, the Open Mobile Alliance, and the Open Connectivity Foundation.