May 2025 - Gradstudents - mailman.cs.vt.edu

Akshay Bansal's PhD defense, Tuesday, May 27, 2025, 9am
by Sikora, Jamie 23 May '25

23 May '25

Happy Friday! I welcome everyone to attend Akshay Bansal’s PhD defense next week. Details below. Have a great long weekend! -Jamie > Title: Adversarial two-party quantum interactions in cryptography and machine learning > Date/Time: 05/27/2025, 9:00 AM > Location: 3001 Gilbert Place. Zoom option available upon request.

1 0

IMPORTANT* Flight Related Travel Changes
by Teresa Hall 22 May '25

22 May '25

PLEASE READ *BOTH* UPDATES BELOW! *Travelers with a base point of Blacksburg, VA or Roanoke, VA* may elect to fly from any airport within 300 miles without a documented cost comparison. Airports within 300 miles include but are not limited to: * Charlotte International Airport * Dulles International Airport * Piedmont Triad International Airport (Greensboro) * Raleigh-Durham International Airport * Richmond International Airport * Ronald Regan Washington National Airport *Premium economy and business-class fares * * Economy-class tickets must be reserved for all flights with an in-air travel time of less than eight (8) consecutive hours. * Premium economy-class tickets may be reserved for flights with an in-air travel time that exceeds eight (8) consecutive hours. * Business-class tickets may be reserved for long-haul flights with an in-air travel time that exceeds ten (10) consecutive hours when there is also a business meeting conducted within three hours of landing. **Note: The term “consecutive hours” applies to one flight from take-off to landing. The traveler may not combine in-air travel times on a multi-leg itinerary to meet the requirements for the purchase of premium economy or business-class tickets. *Teresa Hall** **Fiscal Technician* Virginia Tech| Department of Computer Science Torgersen Hall, 1180E (0106) 620 Drillfield Drive Blacksburg, VA 24061 540-231-8454 | teresa(a)vt.edu <mailto:teresa@vt.edu> Office Hours: Monday, Wednesday, &Thursday Telework:Tuesday &Friday __

1 0

Fw: Xrai Lab Call for Participants – VR Teams in Military Battle Drill
by Coulson, Sara 19 May '25

19 May '25

Sharing as requested... Sara Coulson Director of Graduate Programs | Computer Science Torgersen 3210C 620 Drillfield Drive Blacksburg, VA 24061 ________________________________ From: McMahan, Ryan <rpm(a)vt.edu> Sent: Monday, May 19, 2025 4:08 PM To: Coulson, Sara <sara83(a)vt.edu> Subject: Xrai Lab Call for Participants – VR Teams in Military Battle Drill Hi Sara, Can you share the IRB-approved Call for Participants below and attached flyer with the graduate CS students? Best regards, Ryan Ryan P. McMahan, Ph.D. Deputy Director, Center for Human-Computer Interaction (CHCI) Professor, Department of Computer Science (CS) Virginia Tech CALL FOR VR TEAMS! Experience a Military Battle Drill! The Extended Reality & Artificial Intelligence (Xrai) Lab at Virginia Tech is inviting you to participate in a research study focused on figuring out how groups of people and virtual humans perform as a military team in virtual reality (VR). The data collected will be used to develop multimodal and machine learning models of team performance based on a VR battle drill simulation. Eligible participants: * Are 18 to 64 years old, * Have 20/20 vision without or with contacts (glasses are not permitted during the study), * Can walk and use both hands, * Are willing to attend a 2-hour in-person experimental session in Gilbert Place, * Are willing to participate as a team with up to three other participants, * Are proficient in reading, writing, listening, and speaking English, * Are U.S. Citizens, * Are NOT pregnant, * Do NOT have pre-existing binocular vision abnormalities or psychiatric disorders, * Have NOT recently undergone any medical procedure (including cosmetic procedures), and * Do NOT suffer from a heart condition or other serious medical condition. Participants will receive a $45 Amazon gift card in compensation for participating in and completing the study, if they are not military personnel. If you are interested in more information, please review the consent form document for the study and verify your eligibility here: https://forms.gle/mFMMrzLx2NuztWZa8. After completing the online survey, the research team will reach out to you via email to confirm a time for the research session. This project is supervised by Dr. Ryan P. McMahan (rpm(a)vt.edu<mailto:rpm@vt.edu>, 540-231-9722) from the Department of Computer Science. This experiment has been approved, as required, by the Virginia Tech Institutional Review Board.

1 0

Fw: Postdoc Opportunity - ARL Distinguished Postdoctoral Fellowships - Please Share!
by Cho, Jin-Hee 13 May '25

13 May '25

Dear PhD students and faculty colleagues, I wanted to share information about an excellent postdoctoral fellowship program—please see the details below. Faculty, feel free to forward this to anyone who may be interested. Best regards, Jin-Hee ________________________________ From: NASEM Fellowships Office Outreach Sent: Tuesday, May 13, 2025 9:05 AM To: Cho, Jin-Hee Subject: Postdoc Opportunity - ARL Distinguished Postdoctoral Fellowships - Please Share! You don't often get email from fellowshipsofficeoutreach(a)nas.edu. Learn why this is important<https://aka.ms/LearnAboutSenderIdentification> [https://outlook.office.com/mail/inbox/id/AAkALgAAAAAAHYQDEapmEc2byACqAC%2FE…] Dear Sir or Madam: We ask your assistance in forwarding this message to graduate students in your department to inform them of an outstanding fellowship opportunity with the Army Research Laboratory (ARL). The Army Research Laboratory (ARL) invites exceptional early career researchers to apply for an ARL Distinguished Postdoctoral Fellowship<http://pull.xmr3.com/p/6373-D47D/202680950/1dd47577-2f43-49b6-a582-852b9019…> that provides recipients the opportunity to pursue independent research of their own choosing that supports the mission of ARL<http://pull.xmr3.com/p/6373-177C/202680954/5cf78680-1db1-4747-b228-4cc6e360…>. Candidates must display extraordinary ability in scientific research, show clear promise of becoming outstanding leaders, and are expected to have already successfully tackled a major scientific or engineering problem during their thesis work or to have provided a new approach or insight, evidenced by a recognized impact in their field. Eligibility and Selection Criteria * Must have completed all the requirements for a Ph.D. or Sc.D. degree in the physical sciences, life sciences, computational sciences, behavioral sciences, or engineering by October 1, 2025. Applicants who intend to defend their dissertation after October 1, 2025, are not eligible. * May not be more than five years beyond the award date of their Ph.D. or Sc.D. by the fellowship application deadline (July 10, 2025 at 5:00 PM Eastern Time).\ * Must demonstrate exceptional qualifications with respect to academic and scholarly achievement, as evidenced by research and publication. * Are expected to have conducted research on a major scientific or engineering problem during their thesis work or have provided a new approach or insight, evidenced by a recognized impact in their field. There are no citizenship requirements; however, selected fellows must pass a Department of Defense (DoD) clearance process, requiring a background security investigation. Foreign nationals must get further DoD approval, requiring a background investigation. Fellowship benefits * Full-time, one-year in-residence appointment at ARL, renewable for up to three years based on performance * Annual stipend of $115,000 * Health insurance (including dental and vision) * Paid relocation and professional travel allowance Interested candidates can find additional information and apply at https://arldpf.nationalacademies.org<http://pull.xmr3.com/p/6373-DA17/201262660/1dc62db9-c1a7-4621-adc6-4fd2dd3d…>. Sincerely yours, Elizabeth M. Prescott, D.Phil. Director, Fellowships Office [https://outlook.office.com/mail/inbox/id/AAkALgAAAAAAHYQDEapmEc2byACqAC%2FE…] ________________________________ If you prefer not to receive additional messages from this sender, please Click here<http://www.xmr3.com/rm/1458401-63182901-2-6373-US3-C74B/JICHO@VT.EDU/HCS87FE> and confirm your request.

1 0

Andrew Zhang's MS defense, 12:00 PM, May 13th
by Thomas, Chris 13 May '25

13 May '25

Hello, You are kindly invited to attend the MS thesis defense of Andrew Zhang, May 13th, 12:00 PM. Details follow. Title: Discrete Diffusion for Text Infilling Date: 05/13/25 Time & Location: 12:00 pm, Virtual Zoom: https://virginiatech.zoom.us/j/8956885181 Committee Members: Chris Thomas, Pinar Yanardag, Xuan Wang Abstract: Discrete diffusion models are a new class of text generators which offer advantages such as bidirectional context use, parallelizable generation, and flexible prompting compared to autoregressive models. However, existing text diffusion models are unable to handle flexible-length text infilling tasks without ground-truth positional information. To address this limitation, we propose Discrete Diffusion with Optimal Transport Position Coupling (DDOT), the first discrete diffusion model capable of flexible-length text infilling. DDOT introduces a diffusion process that simultaneously denoises discrete token values and continuous token positions with a sample-level optimal transport (OT) coupling to ensure smooth token flows throughout the diffusion process. Our method is orthogonal to existing discrete text diffusion methods and is compatible with various pretrained text denoisers. Extensive experiments on constrained text generation benchmarks demonstrate that DDOT achieves competitive performance among non-autoregressive models, approaches autoregressive methods, and leads to significant improvements in training efficiency and flexibility. Thank you! Chris Thomas

1 0

Anushka Sivakumar's MS defense, 10:30 AM, May 13th
by Thomas, Chris 13 May '25

13 May '25

Hello, You are kindly invited to attend the MS thesis defense of Anushka Sivakumar, May 13th, 10:30 AM. Details follow. Title: Model Control through Lightweight Activation Steering for Vision Language Models Date: 13th May 2025 Time & Location: 10:30am-12:00pm, Data and Decision Sciences (Zoom) Zoom: https://virginiatech.zoom.us/j/8956885181 Committee Members: Dr. Chris Thomas, Dr. Pinar Yanardag, Dr. Xuan Wang Abstract: This work proposes a lightweight steerer, a parameterized model that guides large vision language models (VLMs), toward outputs that better adhere to desired instructions. Our aim is to achieve effective model control, steering behavior towards a target and away from its converse, using paired prompts. We introduce a robust, lightweight steering module that operates in the activation space, adapting steering to VLMs by connecting the language modality with the image context. This approach enables inference-time control over complex output semantics, such as entailment, hallucination, tone, and perspective, and does not directly modify model weights thereby preserving performance on off-target tasks. A key feature of our method is a layer-agnostic steering approach that does not require parameter tuning to identify optimal steering layers or pre-extracted steering vectors to capture complex relationships between target and converse prompts. In addition to our proposed steering module, we introduce a novel multimodal steering dataset, VNIA (Visual Narrative Intent Alignment), designed to facilitate the development and evaluation of vision-language model steering. Thank you! Chris Thomas

1 0

Swapnil Singh's MS defense, 1:00 pm Wednesday 5/14
by Zhang, Liqing 12 May '25

12 May '25

Dear Colleagues and Students, I am delighted to invite you to Swapnil Singh's MS defense on Wednesday, May 14. Title: Multi-modal Multi-Level Neuroimaging Fusion with Modality-Aware Mask-Guided Attention and Deep Canonical Correlation Analysis to Improve Dementia Risk Prediction Date: Wednesday May 14, 2025 Time: 1:00 pm EST Location: Torgersen Hall, Room 1210 Zoom link: https://virginiatech.zoom.us/j/81949848609 Committee Members: Liqing Zhang (co-Chair), Da Ma (co-Chair), T M Murali, Chris Thomas Abstract: Alzheimer’s Disease (AD) is a progressive neurodegenerative disorder characterized by structural and molecular changes in the brain. Early diagnosis and accurate subtyping are essential for timely intervention and therapeutic planning. This thesis presents a novel multimodal deep learning framework that integrates T1-weighted MRI and Amyloid PET imaging to improve the diagnosis and stratification of AD. The proposed architecture leverages a two-stage pipeline involving modality-specific feature extraction using ResNet50 backbones, followed by middle fusion enhanced with a Modality-Aware Mask-Guided Attention (MAMGA) mechanism. To address missing modalities and inter-modal misalignment, the model incorporates Random Modality Masking and Deep Canonical Correlation Analysis (DCCA) for cross-modal feature alignment. Experiments on the ADNI dataset demonstrate that the proposed MRI+PET (MAMGA+DCCA) model achieves a balanced accuracy of 0.998 and an AUC-ROC of 0.999 in distinguishing stable normal cognition (sNC) from stable Alzheimer's Disease (sDAT). For the more challenging task of separating stable and progressive MCI (sMCI vs. pMCI), the best-performing fusion model achieved a balanced accuracy of 0.732 and an AUC of 0.789. Extensive ablation studies confirm the contributions of MAMGA, DCCA, and dual-optimizer strategies in enhancing diagnostic robustness. This work highlights the clinical potential of multimodal deep learning frameworks in improving early Alzheimer's detection and stratification. -- Liqing Zhang, Ph.D. Professor, Department of Computer Science, Virginia Tech Fralin Biomedical Research Institute, Virginia Tech FBRI Cancer Research Center, Washington, DC Professor of Health Sciences, Virginia Tech

1 0

Mounika Abbineni's MS defense, 1:30 pm 5/6
by Hooshangi, Sara 09 May '25

09 May '25

Hello, I am delighted to invite our CS community to Mounika Abbineni's MS defense tomorrow (Tuesday). Please see details below: Title: Mixed Methods Study of How Computer Science Students Make Elective Course Choices Date: Tuesday May 6, 2025 Time & Location: 1:30 pm, on Zoom Zoom: https://virginiatech.zoom.us/j/88202275099?pwd=3CbKfvTnvJU3J3bIoMKpQQUWuC42… Committee Members: Sara Hooshangi (Chair), Jin-Hee Cho, Brandi Weiss (GWU) Abstract: Elective course selection is an important part of the undergraduate Computer Science learning experience. These courses often shape a student’s future career along with teaching them new skills and forming their niche in the vast world of Computer Science. While computer science students in Virginia Tech have a designated checklist along with required courses that they are supposed to take, the curriculum offers them a set of technical electives where they are able to learn topics of interest in the vast field of Computer Science. By the time of graduation most Computer Science students have taken about three or four elective courses, where they are able to explore their interests and strengths in Computer Science. This mixed-methods study utilized responses from undergraduate, graduate, and recent alumni from Computer Science, Computer Engineering, and Data Analytics majors at Tech. For the qualitative analysis, a thematic analysis was done from 14 student interviews. A quantitative analysis was performed through survey responses. A statistical analysis of survey responses was done in order to analyze the data and make conclusions on the factors that influence student choices of CS electives. It was observed that as students gain more exposure, they approach elective decisions with a focus on the impact of their career and the utility it offers. Students consistently prioritized taking courses that built their skill set and added value to their career goals. The only gender difference found in how students choose electives was in how important course format factors were, like class timing, modality, and grade distribution. Secondly, students in different years of study prioritized the value of professor-related and course format factors differently in their elective choices. This research aims to help young students find their grounding coming into college so they are aware of how to make better course choices. It also aims to introduce better academic advising strategies and curriculum development such that a students' experiences are better in college with easier transition into a successful and enjoyable work life.} ---------------------------------- Sara Hooshangi, Ph.D. Collegiate Associate Professor Director of Master of Engineering Program Department of Computer Science Virginia Tech

1 2

Reminder: Seminar TODAY: May 9th, 11am: Title: "Large Language Models: Opportunities, Challenges and Pitfalls"
by Kantarcioglu, Murat 09 May '25

09 May '25

Dear All, Please see the details below for an upcoming seminar scheduled for May 9th at 11:00 AM in Room 1100 , Torgersen Hall (Corrected Location). We encourage you to attend if your schedule allows. ------------------------------------------------------------------------------------------------------------------------------------- Title: "Large Language Models: Opportunities, Challenges and Pitfalls" Abstract: Ever since ChatGPT was released, Large language models (LLMs) have captured the imagination of researchers and users worldwide and led to a capacity race. We examine their utility for detecting and explaining deceptive attacks such as fake news, fake product reviews, and phishing. Then, we consider the challenges in the context of the detectability of LLM-generated content. We conclude with our recent outside-the-class experiences with LLMs at the University of Houston and at the Naval Surface Warfare Center (NSWC), Dahlgren. Directions for future research and insights into LLMs revealed by our research will be discussed including the consequences of aligning LLM models for different purposes, such as 1) ability to mimic humans on natural language tasks and 2) safety. (This talk is based on joint work with Dainis Boumber, Nachum Dershowitz, David Marchette, and Bryan Tuck) Bio: Rakesh Verma is a Professor of Computer Science at the University of Houston (UH), where he pioneered a Security Analytics course over a decade ago. He received UH's 2013 Lifetime Faculty Mentoring Award for his extensive mentoring of undergraduate students in research. He is the author of Cybersecurity Analytics (CRC Press, 2019) with David Marchette. He is a cybersecurity section associate editor of the Frontiers in Big Data journal. He has been organizing the ACM Annual International Workshop on Security and Privacy Analytics (IWSPA) since 2015. He is currently the PC Co-chair of ICISS 2025. His research interests include formal methods, algorithm design and analysis, and computer security. He is on sabbatical at NSWC Dahlgren for the 2024-2025 academic year. See Cybersecurity Analytics (CRC Press, 2019) at https://www2.cs.uh.edu/~rmverma/book.html Date and Time: May 9th at 11:00 AM in Location: Room 1100 Torgersen Hall (Corrected Location). Zoom info: https://virginiatech.zoom.us/j/86821347393?pwd=IH4ntEcThTIaojr0xYcxbxbyc828… Meeting ID: 868 2134 7393 Passcode: 318285 ------------------------------------------------------------------------------------------------------------------------------------- Best Regards, Murat ==================================== Murat Kantarcioglu, Ph.D. Professor, and CCI Faculty Fellow Virginia Tech, Department of Computer Science www.kantarcioglu.net ====================================

1 0

Sophia Stil's MS Thesis Defense - 5/13 7pm ET
by Rho, Eugenia 08 May '25

08 May '25

Hi All, You are cordially invited to the Master's Thesis Defense of Sophia Stil. The details are as follows: Date: 05/13/2025 Time: 7:00 PM ET Title: Understanding the Applications of AI for Autistic STEM Student Mentorship Location: 3218 Gilbert Place Zoom: https://virginiatech.zoom.us/j/81187632432 Committee: Eugenia Rho (Chair), Myounghoon Jeon, Sang Won Lee Abstract: Autistic college students in STEM face significant challenges in employment, such as low-pay and hiring discrimination compared to their neurotypical peers, despite existing mentorship programs for career development. This study investigates how artificial intelligence (AI) can be leveraged to enhance STEM mentorship for autistic students. Specifically, it examines students’ acceptance of AI as a complementary support tool alongside human mentors, their previous mentorship experiences, their experiences and perceptions of AI, as well as their deeper visions for future AI integration in mentorship. Chapter 2 presents findings from an exploratory online survey of 203 autistic students, offering insights into their perceptions of AI-augmented mentorship, including perceived benefits, limitations, and areas of concern. Building on these results, Chapter 3 describes a follow-up interview study with 15 participants that explores their current academic and professional experiences typically supported by mentorship and their vision for how AI could support key mentorship activities. Thematic analysis reveals a strong desire for integrated, intelligent tools to facilitate online communication, time management, and emotional well-being, alongside valid concerns regarding the limitations and potential harms of current AI systems. The findings inform design considerations for future AI-supported mentorship and underscore the importance of accessibility and inclusion in the development of such systems. We conclude with recommendations for the design of future mentorship programs and AI to support the unique needs of autistic students in STEM. Eugenia Rho<https://eugeniarho.com/> Assistant Professor<https://eugeniarho.com/> Computer Science | Virginia Tech<https://cs.vt.edu/> [cid:a4758e9e-9d99-4f67-a691-0edc4c4e3b8e]<https://eugeniarho.com/sail/>

1 0