AI Innovation Institute

CSE 600 Seminar Series | Fall 2025


Abstract: Large reasoning models have demonstrated capabilities to solve competition-level math problems, answer deep research questions, and address complex coding needs. Much of this progress has been enabled by scaling of data: pre-training data to learn vast knowledge, fine-tuning data to learn natural language reasoning, and RL environments to refine that reasoning. In this talk, I will describe the current LLM reasoning paradigm, its boundaries, and the future of LLM reasoning beyond scaling. First, I will describe the state of reasoning models and where I think scaling can lead to some additional (though perhaps limited) successes. I will then shift to discussing more fundamental issues with models that scale will not resolve in the next few years. I will touch on four current limitations: outdated knowledge, generator-validator gaps, limited creativity, and poor compositional generalization. In all cases, fundamental limitations of LLMs or of supervised learning in general make these problems challenging, inviting future study and novel solutions beyond scaling.

Bio: Greg Durrett is an associate professor in the Department of Computer Science and the Center for Data Science at New York University. His research is broadly in the areas of natural language processing and machine learning. Currently, his group's focus is on reasoning about knowledge in text, verifying correctness of generation methods, and studying how to make progress on problems that defy LLM scaling. He is a 2023 Sloan Research Fellow and a recipient of a 2022 NSF CAREER award. He has served in numerous roles for ACL conferences, recently as a member of the NAACL Board since 2024 and as Senior Area Chair for ACL 2025 and EMNLP 2025. He received his BS in Computer Science and Mathematics from MIT and his PhD in Computer Science from UC Berkeley, where he was advised by Dan Klein.

The overall purpose of this seminar is to bring together people with interests in Computer Vision theory and techniques and to examine current research issues. This course will be appropriate for people who already took a Computer Vision graduate course or already had research experience in Computer Vision. To enroll in this course, you must either: (1) be in the PhD program or (2) receive permission from the instructors.

Each seminar will consist of multiple short talks (around 10 minutes) by multiple people. Students can register for 1 credit for CSE 656. Registered students must attend and present a minimum of 2 or 3 talks. Everyone else is welcome to attend. Fill in https://forms.gle/pCVXovgfMfQwGqG38 to subscribe to our mailing list for further announcement.

CSE 600 Seminar Series | Fall 2025


Abstract: Vision-language models that see and describe the world are now part of our daily lives, from internet search and accessibility tools to content generation and automatic moderation. However, as these models grow and become more widely used, their limitations have also become increasingly visible. In particular, it has been shown that these models are unable to reliably perform complex tasks that require abstraction and compositional reasoning. For example, they struggle to decompose an image or text into entities, attributes, and relations, and then reason over new combinations of these elements. As a result, we see generated content full of hallucinations, privacy leaks in images, and different types of biases in the model outputs.In this talk, I will outline a research agenda that aims to build trustworthy vision-language models in the age of generative AI. I will begin with compositional reasoning: how natural language inference can be used to decompose complex instructions and captions into atomic, verifiable statements, improving both evaluation and model behavior on tasks that require multi-step reasoning. I will then discuss how synthetic data and simulated environments can be used to train more reliable models, and how they can also stress-test models beyond standard benchmarks, revealing when models drop attributes, break object relations, or fail under distribution shifts. I will also share recent work on using hallucination correction as a signal to improve video-language alignment, and on privacy-preserving image understanding for blind and low-vision users. I will conclude with possible ways we can systematically probe, debug, and repair these models, turning synthetic perception into something we can trust in real-world deployments.



Speaker: Paola Cascante-Bonilla is a tenure-track Assistant Professor in the Department of Computer Science at Stony Brook University (SUNY). Before that, she was a Postdoctoral Associate at the University of Maryland Institute for Advanced Computer Studies (UMIACS), developing methods and metrics related to trustworthy machine learning. She received her Ph.D. in Computer Science at Rice University in 2024, working on Computer Vision, Natural Language Processing, and Machine Learning.Her research focuses on developing systems that enable compositional reasoning and common-sense inference through vision and language, while tackling issues such as cultural biases, data distribution, explainability, and trustworthy AI. Additionally, Cascante-Bonilla creates simulated environments for embodied agents to learn in a safe, controlled setting, aiming to facilitate effective collaboration and problem-solving for complex tasks by leveraging the implicit knowledge of large-scale pre-trained deep learning models.
Cascante-Bonilla is the recipient of the Ken Kennedy Institute SLB Graduate Fellowship (2022/23), she was selected as a Future Faculty Fellow by Rice's George R. Brown School of Engineering (2023) and as a Rising Star in EECS (2023).
Location: NCS 120

Please join us on Friday for a CSE 600 talk by CS Faculty, Stanley Bak. During this semester, please periodically check the CSE 600 schedule for the latest talk updates.

Title:  Formal Verification Methods for Cyber-Physical Systems and Neural Networks

Time: Friday 4/1, 2:40 PM

Location:  NCS 120

Abstract: Formal verification methods in Computer Science strive to prove properties about all possible executions of a system, and are an alternative development approach to testing when correctness is paramount. Traditionally these have been applied to hardware circuits, state-machine protocols, or software source code. Prof. Stanley Bak will discuss his research on extending formal verification approaches to more complex areas including cyber-physical systems and neural networks.

Speaker Bio: Stanley Bak is an assistant professor in the Department of Computer Science at Stony Brook University investigating the verification of autonomy, cyber-physical systems, and neural networks. He received a PhD from the University of Illinois at Urbana-Champaign (UIUC) in 2013, and worked for four years in the Verification and Validation (V&V) group in the Aerospace Systems Directorate at the Air Force Research Laboratory (AFRL). He received the AFOSR Young Investigator Research Program (YIP) award in 2020.

Abstract Driving intelligence test is critical to the development and deployment of autonomous vehicles. The prevailing approach tests autonomous vehicles in life- like simulations of the naturalistic driving environment. However, due to the high dimensionality of the environment and the rareness of safety-critical events, hundreds of millions of miles would be required to demonstrate the safety performance of autonomous vehicles, which is severely inefficient. We discover that sparse but adversarial adjustments to the naturalistic driving environment, resulting in the naturalistic and adversarial driving environment, can significantly reduce the required test miles without loss of evaluation unbiasedness. By training the background vehicles to learn when to execute what adversarial maneuver, the proposed environment becomes an intelligent environment for driving intelligence testing. We demonstrate the effectiveness of the proposed environment in a highway-driving simulation. Comparing with the naturalistic driving environment, the proposed environment can accelerate the evaluation process by multiple orders of magnitude. ZOOM LINK: Meeting ID: 950 6760 3617; Passcode: 426506 https://stonybrook.zoom.us/j/95067603617?pwd=dXQybEprSkNlTFY3WHlWYjViUG95UT09 Bio Professor Henry Liu is a professor in the Department of Civil and Environmental Engineering at the University of Michigan, Ann Arbor. He is also a Research Professor at the University of Michigan Transportation Research Institute and the Director for the Center for Connected and Automated Transportation (USDOT Region 5 University Transportation Center). Prof. Liu conducts interdisciplinary research at the interface between civil and mechanical engineering. Specifically, his scholarly interests concern traffic flow monitoring, modeling, and control, as well as testing and evaluation of connected and automated vehicles. He has published more than 100 refereed journal papers and is listed as one of the top 50 leading authors in the past 50 years (1969-2019) in the prestigious Transportation Research journal. Professor Liu and his work have been widely recognized in public media for promoting smart transportation innovations. He has appeared on media outlets including CNBC, Forbes, Technode, etc. In 2019, Professor Liu was invited to testify on national transportation research agenda in front of the US House Subcommittee on Research and Technology. Professor Liu has nurtured a new generation of scholars, and some of his PhD students and postdocs have joined first class universities such as Columbia University, Purdue University, RPI, etc. Prof. Liu is the managing editor of Journal of Intelligent Transportation Systems.

Abstract: In high-dimensional data spaces, vast empty regions often exist where no known data points are present. These empty spaces are not merely gaps but hold untapped potential for discovering novel configurations, optimizing parameters, and improving decision-making processes. However, traditional exploration techniques struggle to identify and leverage these regions due to the curse of dimensionality. To address this, we introduce the Empty Space Search Algorithm (ESA), a scalable, physics-inspired method that systematically identifies and explores these uncharted voids. ESA operates by modeling the data space as a dynamic system, using a repulsion-attraction mechanism to locate optimal empty space configurations (ESCs) without requiring exhaustive search. Building upon ESA, we present GapMiner, a visual analytics system that integrates human-in-the-loop AI to iteratively refine and validate ESCs. GapMiner combines parallel coordinate visualization, interactive optimization, and deep learning-based predictive modeling to enhance the efficiency of empty space exploration. This methodology has broad applications, including accelerating convergence in evolutionary algorithms through a more diverse initial population, optimizing adversarial learning strategies, and discovering novel parameter configurations in reinforcement learning. Our approach demonstrates that empty space is not just an absence of data but a frontier for new possibilities in high-dimensional problem-solving.
Bio: Xinyu Zhang received his B.E. in Computer Science from Shandong University, Taishan College, in 2019. He is currently a final-year Ph.D. candidate in the Department of Computer Science at Stony Brook University, advised by Prof. Klaus Mueller. His research focuses on multivariate data analysis, scientific visualization, and reinforcement learning. He has published multiple papers in top-tier journals and conferences, including IEEE TVCG and NeurIPS.
*this seminar will be held in person (food provided on a first come, first serve basis), and online (zoom link below)!
Topic: IACS Student Seminar Speaker: Xinyu Zhang
Time: Feb 26, 2025 12:00 PM Eastern Time (US and Canada)
Join Zoom Meeting
https://stonybrook.zoom.us/j/91848218975?pwd=lfITFa61GaXZ2Wsa1B1OnbLQMmXvOE.1

Meeting ID: 918 4821 8975
Passcode: 027337

University Libraries Presents: The Library AI Club is a welcoming space for students, faculty, and staff to explore AI in a supportive, low-pressure environment. Meeting every two weeks, the club features discussions, collaborative projects, guest speakers, and hands-on experiments. Join us to learn, share ideas, and engage with AI responsibly and creatively. We'd love to see you at an upcoming meeting! Location: Melville Library, Scholarly Communication Seminar Room

Abstract: Molecular learning has become an emerging field of AI, driving breakthroughs in drug discovery, protein design, and materials design. For high-stakes scientific tasks, however, predictive accuracy alone is not sufficient: models must also be interpretable and trustworthy. Our work aims to study molecular learning under a unified explainability perspective across two major model families: Graph Neural Networks (GNNs) and Large Language Models (LLMs).

GNNs are natural choices for molecular graphs and achieve strong performance on many molecular tasks. To enhance explainability, many GNN explanation methods have been proposed and work well for 2D GNNs. However, 3D GNNs introduce two key challenges: producing chemically meaningful substructures and reducing fidelity loss caused by dense geometric graphs. To address these challenges, I present two methods. 3DGraphX decomposes dense 3D graphs into chemically meaningful 3D motifs, enabling compact explanations that align with chemical intuition. EDMA introduces an energy-based discrete mask approximation approach to reduce the discrepancy between the soft mask optimized during training and the hard mask used for explanation, improving explanation fidelity.

LLMs present different characteristics and challenges compared with GNNs. LLMs can provide a certain level of explanation through step-by-step reasoning, and their natural-language outputs are easy for humans to understand and interpret. However, because LLMs are trained for general-purpose tasks, their performance on scientific tasks often lags behind specialized GNNs. To improve performance, existing methods guide LLMs by providing suggestions through brief feedback, retrieval-augmented generation (RAG), or planner agents. However, these approaches face several limitations, such as vague guidance, introduced bias problems, and high computational cost. To fill the gap, I propose RL-Guider, a lightweight reinforcement-learning agent that converts evaluation feedback into input-specific guidance for molecular optimization. RL-Guider improves over time by accumulating historical experience and transfers efficiently across different LLMs while preserving interpretability.

Together, these efforts aim to provide explanations that are scientifically meaningful and faithful, while also preserving or improving performance on molecular tasks to better meet real scientific needs.

Speaker: Xufeng Liu

Location: New Computer Science-1-Room 115