- Gradstudents - mailman.cs.vt.edu

2025-26 CSGC Election Results!
by Woo, Wesley 01 May '25

01 May '25

Hi folks, Please join me in congratulating the CS Graduate Council officers for the upcoming 2025-26 school year: President: Bernie Moussad Vice President: Vee Pettit Treasurer: Syed Ibrahim Mustafa Bukhari Secretary: Sindhuja Madabushi MEng Rep: William Furgerson In the end, it was a tight election between a lot of highly qualified candidates. Thank you to everyone who voted and a special thank you to everyone who ran — CSGC isn't possible without people who are willing to volunteer their time and energy towards serving the community. It's been a pleasure serving as president this past year and I'd like to conclude my term with a reminder to everyone that anyone can join the CSGC meetings and contribute to the discussion. You don't have to have an official title to help improve our graduate student community! -- Wesley

3 2

Sindhura Kommu's MS defense, 10 am 5/1, both in DDS 445 and on Zoom
by Wang, Xuan 01 May '25

01 May '25

Hi everyone, You are cordially invited to Sindhura Kommu's MS defense on Thursday. See details below. Title: Towards Network-Guided Large-Scale Foundation Models on Single-Cell Transcriptomics Date: Thursday, May 1, 2025 Time & Location: 10 am, DDS 445 Zoom: https://virginiatech.zoom.us/j/81006101573?from=addon Committee Members: Xuan Wang (Chair), Yue Wang, Dawei Zhou Abstract: Large pretrained models, known as foundation models, have driven substantial progress in natural language processing and related fields. Recently, foundation models tailored for single-cell RNA sequencing (scRNA-seq) have shown significant potential in interpreting the transcriptomic 'languages' of cells through self-supervised learning on extensive, large-scale gene expression datasets. These models could significantly enhance our understanding of cellular functions and disease mechanisms. However, unlike text data, scRNA-seq data are inherently noisy and sparse, posing unique challenges. We hypothesize that a major limitation of current single-cell foundation models (scFMs) lies in their inability to effectively leverage prior biological knowledge encoded within gene regulatory networks (GRNs), which could provide valuable complementary insights. scRNA-seq technology enables the fine-grained inference of GRNs, illuminating cellular phenotypes at the molecular scale. This thesis initially addresses this limitation by incorporating experimentally validated transcription factor-DNA binding data, employing supervised machine learning methods that leverage known regulatory interactions to enhance GRN inference framed as a gene regulatory link prediction task. To address gene regulatory link prediction, we developed scRegNet, an innovative framework combining scFMs with graph-based learning. scRegNet achieved state-of-the-art results when compared to nine baseline methods across seven scRNA-seq benchmark datasets and demonstrated greater robustness in the presence of noisy training data. In the second part of the thesis, we systematically explored incorporating prior GRNs into the pretraining of single-cell foundation models. This exploration provided valuable insights into the benefits and limitations of network guidance, revealing varied effects on predictive accuracy across different downstream tasks related to chromatin and network dynamics. We present detailed findings from these experiments offering insights to inform future research directions. Best, Xuan -------- Assistant Professor Department of Computer Science Sanghani Center for AI and Data Analytics Virginia Tech

3 2

Priya Pitre's MS defense, 10 am 5/8, both in Innovation Campus Room 5736 and on Zoom
by Wang, Xuan 01 May '25

01 May '25

Hi everyone, You are cordially invited to Priya Pitre's MS defense on Thursday next week. See details below. Title: Toward Deliberative AI: Multi-Agent LLMs for Real-World Reasoning Date: Thursday, May 8, 2025 Time & Location: 10 am, Innovation Campus Room 5736 Zoom: https://virginiatech.zoom.us/my/priyap Committee Members: Xuan Wang (Chair), Naren Ramakrishnan (Co-Chair), Pinar Yanardag Abstract: In this work, we study how to design and evaluate multi-agent systems built on large language models (LLMs) to simulate human-like deliberation, consensus, and reasoning. We present two core frameworks. First, we introduce CONSENSAGENT, a system designed to improve both the effectiveness and efficiency of multi-agent debates on objective reasoning tasks. Through a novel trigger mechanism and prompt optimization step, CONSENSAGENT reduces sycophancy between agents and addresses prompt ambiguity—two major failure modes in agentic collaboration. We evaluate CONSENSAGENT on six benchmark reasoning datasets and demonstrate substantial gains in both accuracy and speed of convergence compared to existing single- and multi-agent baselines. Second, we introduce TRUCE, a reinforcement learning based framework for real-world simulations, where agents trained on real policy data engage in structured debate and compromise. We construct domain-specific datasets from interviews with city stakeholders, public statements by U.S. senators, and legal opinions by Supreme Court judges. These agents engage in deliberative simulations under varied experimental conditions including moderation, social pressure, credibility manipulation, and emotional argumentation. We evaluate not only whether consensus is reached, but also how much each agent compromises using ranking metrics such as nDCG and MAP. Together, these frameworks demonstrate that multi-agent LLMs can be used to study both logical coordination and complex real-world negotiation dynamics. Best, Xuan -------- Assistant Professor Department of Computer Science Sanghani Center for AI and Data Analytics Virginia Tech

1 0

5/2 CS Grad Council End of Semester Picnic!
by Moussad, Bernard 30 Apr '25

30 Apr '25

Hey everyone, Just a reminder that this Friday is the end of semester picnic! Time: Friday, 5/2 from 5 to 8pm Location: Shelter 9 at Blacksburg Municipal Park (https://maps.app.goo.gl/iQiXoKowe8sGD2C49) (~920 Turner St NE, Blacksburg, VA 24060) We will need a few volunteers to help with setup, walking groups, and tear down so if you feel up to it, please put your name down! https://docs.google.com/spreadsheets/d/1p8XqUTXgSvwIVwRFDtqRhKur-a6fbVyCar_… Also attached a picture of the park with Shelter 9 circled! Hope that helps! Can't wait to see everyone there! Sincerely, Bernard Moussad [cid:892c4f3c-910b-4ac9-a974-78c24370ce00]

1 0

PARTICIPANTS NEEDED - Optimization of the Plan of Study academic planning process
by Guo, Peiqing 29 Apr '25

29 Apr '25

Hi everyone, The Code World lab is doing a research survey on optimizing the plan of study process by creating a interactive website that brings the process of searching for classes, checking for requirements, and setting up a personalized plan of study into a single website. the website is published at plan-of-study.cs.vt.edu<http://plan-of-study.cs.vt.edu/>. There is a walkthrough of how to use the website on the FAQ page. Please read through the instructions first when using the website. Once you are done messing around the website, please take this quick survey to give us your thoughts on the website! https://docs.google.com/forms/d/e/1FAIpQLSdxiULMT7Y-tG3SUGPOUPwK-6x9T1muHT2…. This process should take less than 30 minutes to complete, and it can help you to visualize your own plan of study. Best, PeiQing Guo MS Thesis student in CS @ VT

1 1

CS Grad council meeting tomorrow - last one of Spring 2025
by Ibn Arif, Kazi Hasan 29 Apr '25

29 Apr '25

Hi everyone, This is a quick reminder about our next CS Grad Council meeting tomorrow, Wednesday, April 30, at 12 PM in Gilbert 3001 (third floor). You can join over zoom too! Here is the link: https://virginiatech.zoom.us/j/87009667028 We look forward to seeing you there! Best, Hasan -- Kazi Hasan Ibn Arif PhD Student at Virginia Tech Dept. of Computer Science LinkedIn<https://www.linkedin.com/in/kazi-hasan-ibn-arif-8b78a61a3?utm_source=share&…> | Google Scholar<https://scholar.google.com/citations?user=3OhNWLcAAAAJ&hl=en>

1 0

CCI Workshop on Testing and Securing User Interfaces in CI/CD Pipelines
by Brown, Chris 29 Apr '25

29 Apr '25

Dear All, We are excited to invite you to participate in our upcoming workshop on Testing and Securing User Interfaces in CI/CD Pipelines. This event, sponsored by the Commonwealth Cyber Initiative, will be organized as a "working group" for researchers and practitioners to discuss challenges and brainstorm opportunities to enhance user interface (UI) testing and security in CI/CD pipelines (all experience levels are welcome). We are also thrilled to feature a virtual keynote from Andrew "Pandy" Knight---the Automation Panda. The workshop will take place on Saturday, May 10th, 2025, from 10:00 AM - 2:30 PM ET in-person in Torgersen 1100 and virtually via Zoom (link shared with registrants). More details about the workshop are available on the Workshop Website<https://code-world-no-blanket.github.io/cci-workshop/> and please register for the event here<https://forms.gle/WqPJrje1H1uE3Fsh9> by May 7th. Please feel free to contact me with any questions for additional information. Thanks, Chris

1 0

Grad Student Updates: GTA App Deadline, Course Withdrawal, Fall Registration & More
by Coulson, Sara 25 Apr '25

25 Apr '25

CS Grad Students, Some friendly reminders and updates that are relevant to you. Hope you all have a lovely weekend! GTA Applications Priority Deadline- May 2 Planning to be a GTA in Fall 2025? Submit your application<https://hosting.cs.vt.edu/gta/> by May 2, even if you have guaranteed funding. We still need your form! Graduate Course Withdrawal Deadline – May 2 Need to drop a Spring 2025 grad course? Submit the Graduate Course Withdrawal<https://graduateschool.vt.edu/content/dam/graduateschool_vt_edu/forms/Cours…> (signed by you and your instructor) via the CS Grad Form Submission Portal. Welcome, Andrea Sirles! Andrea Sirles has joined the Grad Program Office as our new Graduate Program Coordinator, working with Blacksburg MS and Accelerated MS students. Fall 2025 Add/Drop Open Through Aug. 29 The Fall 2025 timetable<https://onecampus.vt.edu/task/all/timetable> is live! You can now make schedule changes through August 29. Research Credit Enrollment (CS 5994 / CS 7994) Starting Fall 2025, each advisor will have their own section of CS 5994 & 7944. If you don’t see your advisor’s section, please let me know and I’ll get it fixed! No advisor yet? Use these CRNs: * MS – CRN 83780 * PhD – CRN 83791 Need Help Adjusting Research Credits in Add/Drop? Watch this quick Add/Drop Instructional Video for step-by-step guidance. Best, Sara Sara Coulson Director of Graduate Programs | Computer Science Torgersen 3210C 620 Drillfield Drive Blacksburg, VA 24061

1 0

Master's thesis defense: Masaki Hasegawa (3 pm, April 25@VTRC)
by MATSUO, SHINICHIRO 24 Apr '25

24 Apr '25

Hi all, You're invited to the Master's thesis defense of Masaki Hasegawa. Date: 04/25/2025 (canary:event?ts=767289605.00) Time: 3:00 PM (canary:event?ts=767214005.00) Title: LLM-Assisted Detecting and Redacting Confidential Information for Government Information Disclosure Place: 3-024 VT Research Ctr-Arlington (canary:map?Street=3-024%20VT%20Research%20Ctr&City=Arlington) Committee: Matsuo Shinichiro(Chair), Lou Wenjing, Cameron Melissa Abstract: Generative AI, especially large language models (LLMs), has advanced rapidly, with real-world applications growing steadily. However, the use of generative AI in the public sector has lagged behind the private sector. This thesis focuses on the "Governmental Information Disclosure Process," which is vital in democratic countries' administrative systems. Many developed nations require government agencies to disclose information to citizens, excluding confidential data such as personal information. Although agencies must confirm the presence of confidential information and redact or mask it before release, this process is still manual, creating significant room for improvement. Additionally, since the information to be masked is defined in natural language, such as legal text, interpreting documents' contexts to determine what qualifies as confidential is resource-intensive. In this context, LLMs, capable of inferring context and general knowledge, could efficiently identify parts of documents that require masking. This thesis first reviews the existing literature on sensitive or confidential information detection using LLMs, clarifying the use cases and the category of information identified in both the private and public sectors. Then, as a case study, we create sample documents modeled after Japanese administrative texts and compare the detection and masking results performed by testers with administrative experience, following legal requirements, with those generated by an LLM. This study contributes by proposing a zero-shot approach where LLMs directly generate masked text with dynamically determined granularity. This resolves the fundamental trade-off in previous methods by allowing the model to decide appropriate masking units (characters, words, or phrases) based on contextual requirements rather than predetermined structural units. Shin’ichiro Matsuo, Ph.D., Research Professor Virginia Tech | Commonwealth Cyber Initiative (CCI) 900 N. Glebe Road, Arlington, VA 22203 (571) 858-3115 (tel://(571)%20858-3115) | matsuol(a)vt.edu (mailto:matsuol@vt.edu)

1 0

End of Year CS Grad Picnic - 5 PM Fri May 2nd at Blacksburg Municipal Park, Shelter 9
by Woo, Wesley 23 Apr '25

23 Apr '25

Hi folks, On Friday May 2nd, we'll be having an End of the Year Picnic for CS graduate students at Blacksburg Municipal Park! The event will be at Shelter 9 (https://maps.app.goo.gl/iQiXoKowe8sGD2C49) from 5 PM to 8 PM. Please fill out the following form to help us plan for food: https://forms.gle/2U7JWK7R8Mw9yoBh6 Wesley

1 1