Ph.D. Dissertation Defense Talk: Effective Model Deployment and Data Curation for Foundation Model Development
Speaker: Cheng-Yu Hsieh, Advisors: Ranjay Krishna, Alexander J. Ratner, Supervisory Committee: Ranjay Krishna (Chair), Alexander J. Ratner (Chair), Aylin Caliskan (GSR, Information School), Hanna Hajishirzi, Noah A Smith, Abstract: While scaling, both in terms of model and data size, has led to many recent breakthroughs in AI, this increasingly larger-scale development trajectory faces emergent challenges not seen in traditional small-scale supervised learning settings. On the model side, the exponential growth in parameters has rendered highly capable but massive models inaccessible for many applications. On the data side, simply scaling the volume of data does not always lead to more emergent model capabilities. In this talk, I'll introduce two complementary lines of my research that address these challenges. First, I will present strategies for efficient model deployment and adaptation. Second, I will discuss how better data curation can enable more reliable measurement and effective induction of desired…
Tuesday, May 20, 2025, 11:30 AM – 12:30 PM.
AccessADVANCE Capacity Building Institute
Join AccessADVANCE and colleagues from across the country to share challenges and solutions for creating systemic changes that supports the recruitment and participation of faculty with disabilities in STEM careers. We welcome the participation and perspectives of people with disabilities, administrators, faculty, and disability services professionals. Online via Zoom , In-person on the UW campus in Seattle , Learn more and apply , Apply to participate by April 11, Virtual and in-person participants are invited to propose and lead a flash talk (~15 minutes). You can also suggest panel discussions topics & volunteer to facilitate or join a panel. Virtual participants will receive Zoom links prior to the event. We will explore topics like: Accommodating faculty with disabilities , Funding accommodations , Finding community , Informal accommodations and interdependence , Intersectional issues particularly impacting women faculty with disabilities.
Event interval: Ongoing event. Accessibility Contact: doit@uw.edu. Event Types: Conferences. Diversity Equity Inclusion. Target Audience: People with disabilities, administrators, faculty, and disability services professionals.
Wednesday, May 21, 2025 – Thursday, May 22, 2025.
For more info visit docs.google.com. Ph.D. Dissertation Defense Talk: Algorithimic data efficient learning in the era of large model.
Speaker: Yifang Chen, Advisors: Simon S Du, Kevin Jamieson, Supervisory Committee: SIMON S DU (Chair), KEVIN G JAMIESON (Chair), Yingfei Wang (GSR, Business - Foster School of), Pang Wei Koh, Abstract: In the race towards Artificial General Intelligence, data is the fuel that powers our most advanced models. Vision-Language Models like LLaVA and CLIP are trained on billions of image-text pairs, while Large Language Models (LLMs) like GPT and Claude may process trillions of text samples. Despite the abundance of data, ensuring its quality and effective curation remains more of an art than a science. This process must manage real-world data that is multimodal, noisy, and lacks a guaranteed relationship to target tasks. Furthermore, the process is compounded by the complex training dynamics of neural networks, where the value of each data point depends heavily on the evolving state of model training.
Without principled guidance, these challenges often create systematic blind spots, and their impact remains…
Thursday, May 22, 2025, 10:30 AM – 11:30 AM.
Ph.D. Dissertation Defense Talk: Technology and Power: Examining Imbalances Through Usable Security & Privacy Research
Speaker: Kentrell Owens, Advisors: Tadayoshi Kohno, Franziska Roesner, Supervisory Committee: Tadayoshi Kohno (Chair), Franziska Roesner (Chair), Martin Nisser (GSR, Aeronautics and Astronautics), Joseph Calandrino (Department of Justice), Abstract: Technology is often framed as a neutral tool, but in practice, it amplifies existing power imbalances, particularly when imposed by institutions rather than chosen by those most affected. This talk explores how surveillance, coercion, and informational asymmetry shape the experiences of individuals subject to technology they did not choose. Using methods from usable security and privacy research, I analyze case studies across the U.S. criminal legal system, the U.S. immigration system, and the modified Android app ecosystem to understand how design choices and data practices disproportionately harm those with the least power to resist them. I reflect on the broader implications of these imbalances and discuss potential strategies for mitigating harm in future…
Friday, May 23, 2025, 10:00 AM – 11:00 AM.
Memorial Day
Holidays
No classes. Most University offices and buildings are closed. Check with specific offices to confirm.
Event interval: Single day event. Year: 2025. Quarter: Spring. Event Types: Academics.
Monday, May 26, 2025.
For more info visit www.washington.edu. Ph.D. Dissertation Defense Talk: Title TBA
Speaker: Yizhong Wang, Advisors: Hannaneh Hajishirzi, Noah Smith, Supervisory Committee: Hanna Hajishirzi (Chair), Noah A Smith (Chair), Shane Noah Steinert-Threlkeld (GSR, Linguistics), Pang Wei Koh, Abstract TBA.
Tuesday, May 27, 2025, 11:30 AM – 12:30 PM.
Ph.D. Dissertation Defense Talk: Code in Context: Keeping Developer Context and Interfaces on Evolving Code
Speaker: Edward Misback, Advisors: Steve Tanimoto, Zachary Tatlock, Supervisory Committee: Steven L Tanimoto (Chair), Zachary Tatlock (Chair), Amy J. Ko (GSR, Information School), Rene Just, Abstract: Despite progress in programming tools and environments, developers still struggle to maintain mental models of the systems they build and the context surrounding their code. This dissertation advances a framework for enhancing program understanding and manipulation for all agents working with a codebase—including developers and automated agents such as large language models (LLMs)—through notes on shared context and high-level representations that are semantically attached and synchronized, i.e. external notes and UIs that stay linked to sections of a program where they are meaningful, with their content updated to remain relevant.
I first address the specific case of floating-point error analysis with Odyssey. Odyssey is a workbench for floating-point analysis that transforms an existing low-level,…
Tuesday, May 27, 2025, 12:00 PM – 1:00 PM.
Ph.D. Dissertation Defense Talk: Mind the Metric: Metric-Informed Methods in Machine Learning
Speaker: Sandesh Adhikary, Advisor: Byron Boots, Supervisory Committee: Byron Emereth Boots (Chair), Yen-Chi Chen (GSR, Statistics), Abhishek Gupta, Sewoong Oh, Abstract: Various machine learning algorithms can benefit from exploiting the metric structure inherent in data and model parameters. Such geometric structure can be a useful prior to reduce sample complexity, improve generalization, and ensure feasibility of learned models. We present a range of metric-informed machine learning methods that span reinforcement learning, sampling from probability distributions, and probabilistic graphical models.
Within the domain of reinforcement learning (RL), we use metric structure to define geometry-aware representations of states and policies. Specifically, we introduce Behavioral Eigenmaps (BeigeMaps) -- state representations that preserve the local metric structure induced by bisimulation metrics through neural eigendecompositions of similarity kernels. When added as a drop-in modification, BeigeMaps…
Wednesday, May 28, 2025, 9:00 AM – 10:00 AM.
Ph.D. Dissertation Defense Talk: Off-road Navigation Under Sensing Uncertainty
Speaker: Matthew Schmittle, Advisor: Siddhartha Srinivasa, Supervisory Committee: Siddhartha Srinivasa (Chair), Sawyer Buckminster Fuller (GSR, Mechanical Engineering), Byron Emereth Boots, Sanjiban Choudhury (Cornell), Abstract: This talk dives into novel motion‑planning approaches that enable high‑speed off‑road robots to navigate rugged terrain. Off‑road autonomy requires a robot to navigate unstructured environments relying heavily (or entirely) on onboard sensors and compute. With no traffic rules or roads to guide the robot, it must make split‑second judgments about where it can and cannot go. This is challenging when diverse testing environments and incomplete sensing inject lots of uncertainty into the robot's perception. I will discuss three way of tackling uncertainty to improve planning under these harsh conditions. First, how posterior sampling can tractable reason about sensing uncertainty to produce plans that balance pessimism and optimism. Second, a novel method for leveraging sparse…
Friday, May 30, 2025, 10:00 AM – 11:00 AM.
Ph.D. Dissertation Defense Talk: Ubiquitous Computing Platforms as Scalable Surrogates in Public Health and Well-Being
Speaker: Joe Breda, Advisor: Shwetak Patel, Supervisory Committee: Shwetak N. Patel (Chair), Steve Mooney (GSR, Epidemiology - Public Health), Kurtis L. Heimerl, Vikram Iyer, Abstract: Computing platforms with sensing components have become ubiquitous across modern society – present in vehicles, across environments, in our pockets, and on our bodies. While these sensors may have been originally designed and included in these devices with a specific purpose in mind, they can be re-contextualized to offer new insights for secondary applications – a process we call side-channel sensing. We can then use these already ubiquitous sensors as surrogates for more direct measures to realize the benefits of scale at the cost of some directness. In this thesis, I explore 3 approaches of re-contextualizing or extending ubiquitous devices for new sensing applications in public health and well-being. First, I demonstrate how the internal temperature sensor of smartphones can be re-contextualized as a thermometry to make…
Friday, May 30, 2025, 10:30 AM – 11:30 AM.
Ph.D. Dissertation Defense Talk: Beyond Scaling: Frontiers of Retrieval-Augmented Language Models
Speaker: Akari Asai, Advisor: Hannaneh Hajishirzi, Supervisory Committee: Hanna Hajishirzi (Chair), Lucy Lu Wang (GSR, Information School), Pang Wei Koh, Yulia Tsvetkov, Luke Zettlemoyer, Abstract: Large Language Models (LMs) have achieved remarkable progress by scaling training data and model sizes. However, they continue to face critical limitations, including hallucinations and outdated knowledge, which hinder their reliability—especially in expert domains such as scientific research and software development. In this talk, I will argue that addressing these challenges requires moving beyond monolithic LMs and toward Augmented LMs—a new AI paradigm that designs, trains, and deploys LMs alongside complementary modules to enhance reliability and efficiency. Focusing on my research on Retrieval-Augmented LMs, one of the most impactful and widely adopted forms of Augmented LMs today, I will begin by presenting systematic analyses of current LM shortcomings and demonstrating how retrieval augmentation offers a…
Tuesday, June 3, 2025, 12:00 PM – 1:00 PM.
Allen School Colloquium: Shaping AI for the Public Good via the Berkeley Lab Model
ABSTRACT:
This three-part talk summarizes the underlying philosophy and best practices of the Berkeley lab model, offers a perspective on improving AI’s impact, and describes a new research funding opportunity starting this Fall.
Senior computer scientists and rising stars in AI developed a balanced perspective on AI's societal impact informed by interviews with 24 experts including John Jumper, Barack Obama, Susan Rice, and Eric Schmidt (see www.ShapingAI.com). We believe we are in the early days of practical AI, and focused efforts by practitioners, policymakers, and stakeholders can still increase AI's upsides and reduce its downsides.
We suggest guidelines for enhancing AI development: i) Human-AI collaboration exceeds capabilities of either AI or people; ii) Focus on productivity improvements that generate more jobs; iii) Initially prioritize eliminating tedious tasks; and iv) Consider that AI impact varies by geography.
Rather than simply predict long-term AI impact, we've identified research…
Event interval: Single day event. Campus room: Amazon Auditorium | Room G20 (CSE2). Accessibility Contact: dso@u.washington.edu. Event Types: Lectures/Seminars.
Tuesday, June 3, 2025, 3:30 PM – 4:30 PM.
Gates Center (CSE2), G20 | Amazon Auditorium.
For more info visit www.cs.washington.edu. Instruction Ends - Spring Quarter
Dates of Instruction
Instruction ends.
Event interval: Single day event. Year: 2025. Quarter: Spring. Event Types: Academics.
Friday, June 6, 2025.
For more info visit www.washington.edu. Ph.D. Dissertation Defense Talk: Modeling, analyzing, and exploring protein-protein interactions using high-throughput datasets and deep learning
Speaker: Alyssa La Fleur, Advisor: Georg Seelig, Supervisory Committee: Georg Seelig (Chair, Electrical and Computer Engineering), Armita Nourmohammad (GSR, Physics), Pang Wei Koh, Jeffrey Nivala, Abstract TBA.
Friday, June 6, 2025, 10:00 AM – 11:00 AM.
Final Examinations - Spring Quarter
Dates of Instruction
Week of final examinations for spring quarter.
Event interval: Ongoing event. Year: 2025. Quarter: Spring. Event Types: Academics.
Saturday, June 7, 2025 – Friday, June 13, 2025.
For more info visit www.washington.edu. Ph.D. Dissertation Defense Talk: Scaling Human Supervision for Robot Manipulation
Speaker: Michael Murray, Advisor: Maya Cakmak, Supervisory Committee: Maya Cakmak (Chair), Martin Nisser (GSR, Aeronautics and Astronautics), Abhishek Gupta, Joshua R. Smith, Abstract TBA.
Monday, June 9, 2025, 9:30 AM – 10:30 AM.
Ph.D. Dissertation Defense Talk: Identifying, Evaluating, and Mitigating Security and Privacy Threats in Augmented Reality Systems
Speaker: Kaiming Cheng, Advisors: Tadayoshi Kohno, Franziska Roesner, Supervisory Committee: Tadayoshi Kohno (Chair), Franziska Roesner (Chair), Tivon Rice (GSR, Digital Arts and Experimental Media), JON FROEHLICH, Abstract TBA.
Wednesday, June 11, 2025, 1:00 PM – 2:00 PM.
Ph.D. Dissertation Defense Talk: Reimagining Wearable Sensing: Smart Jewelry for Unobtrusive Health Monitoring
Speaker: Shirley Xue, Advisors: Vikram Iyer, Shwetak Patel, Supervisory Committee: Vikram Iyer (Chair), Shwetak N. Patel (Chair), Martin Nisser (GSR, Aeronautics and Astronautics), James Fogarty, Abstract TBA.
Wednesday, June 11, 2025, 2:00 PM – 3:00 PM.
Ph.D. Dissertation Defense Talk: Towards Sociotechnical Security & Privacy and Against Online Abuse
Speaker: Miranda Wei, Advisors: Tadayoshi Kohno, Franziska Roesner, Supervisory Committee: Tadayoshi Kohno (Chair), Franziska Roesner (Chair), Mary Fan (GSR, Law - School of), Richard Anderson, Elissa Redmiles (Georgetown University), Abstract TBA.
Thursday, June 12, 2025, 12:00 PM – 1:00 PM.
Quarter Break - Summer
Dates of Instruction
Break between spring and summer quarters.
Event interval: Ongoing event. Year: 2025. Quarter: Summer. Event Types: Academics.
Saturday, June 14, 2025 – Sunday, June 22, 2025.
For more info visit www.washington.edu. Juneteenth
Holidays
No classes. Most University offices and buildings are closed. Check with specific offices to confirm.
Event interval: Single day event. Year: 2025. Quarter: Summer. Event Types: Academics.
Thursday, June 19, 2025.
For more info visit www.washington.edu. Instruction Begins - Summer Quarter - Full and A-term
Dates of Instruction
Instruction begins.
Event interval: Single day event. Year: 2025. Quarter: Summer. Event Types: Academics.
Monday, June 23, 2025.
For more info visit www.washington.edu. Independence Day
Holidays
No classes. Most University offices and buildings are closed. Check with specific offices to confirm.
Event interval: Single day event. Year: 2025. Quarter: Summer. Event Types: Academics.
Friday, July 4, 2025.
For more info visit www.washington.edu. Instruction Ends - Summer Quarter - A-term
Dates of Instruction
Instruction ends.
Event interval: Single day event. Year: 2025. Quarter: Summer. Event Types: Academics.
Wednesday, July 23, 2025.
For more info visit www.washington.edu. Instruction Begins - Summer Quarter - B-term
Dates of Instruction
Instruction begins.
Event interval: Single day event. Year: 2025. Quarter: Summer. Event Types: Academics.
Thursday, July 24, 2025.
For more info visit www.washington.edu. Ph.D. Dissertation Defense Talk: Primal-dual Methods for Policy Learning in Bandits and Reinforcement Learning
Speaker: Zhihan Xiong, Advisor: Maryam Fazel, Supervisory Committee: Maryam Fazel (Chair, Electrical and Computer Engineering), Yingfei Wang (GSR, Business - Foster School of), KEVIN G JAMIESON, Pang Wei Koh, Lin Xiao (Electrical and Computer Engineering), Abstract TBA.
Monday, July 28, 2025, 2:00 PM – 3:00 PM.
Ph.D. Dissertation Defense Talk: Goal-Centered Personal Informatics Tools
Speaker: Yasaman Sefidgar, Advisor: James Fogarty, Supervisory Committee: James Fogarty (Chair), ANDREA HARTZLER (GSR, Biomedical Informatics & Medical Educ), Jeffrey Heer, Sean Munson (Human Centered Design and Engineering), Jina Suh (Microsoft), Abstract TBA.
Tuesday, July 29, 2025, 8:30 AM – 9:30 AM.