Abstract: Pre-trained diffusion and flow matching models have made visual generation remarkably powerful, enabling high-fidelity synthesis of images and videos from natural language prompts. However, their behavior is still largely dictated by the pre-training data distribution and likelihood objective, which do not directly encode downstream desiderata such as fine-grained semantic alignment, controllability, or realism. This gap motivates post-training: starting from a base generator and further optimizing it with additional supervision signals derived from human or reward model preferences.This work presents post-training for visual generative models through two complementary case studies. First, Hummingbird addresses the problem of fine-grained contextual alignment in image-text-to-image generation. We introduce a multimodal context evaluator that scores the consistency between rich contextual descriptions and generated images, capturing fine-grained alignment beyond global CLIP similarity. By directly backpropagating these differentiable rewards through the diffusion sampler, Hummingbird substantially improves semantic faithfulness while preserving high visual quality.
Second, PISCES tackles post-training for text-to-video generation, where alignment is inherently semantic-spatio-temporal. We show that naive VLM-based rewards suffer from distributional mismatch and token-level misalignment, leading to reward hacking and suboptimal optimization. PISCES introduces a bi-objective, Optimal Transport (OT)-aligned reward module: distributional OT using Neural Optimal Transport to align text and video embedding distributions, and discrete, partial OT over a spatio-temporal cost matrix to capture semantic alignment at the token level. These rewards are integrated into both direct backpropagation and GRPO-style optimization to post-train state-of-the-art text-to-video generators. Together, Hummingbird and PISCES provide a unified view of how carefully designed visual reward models, coupled with OT-based representation alignment, can reliably improve the downstream behavior of pre-trained image and video generators.

Speaker: Minh Quan Le

Location: NCS 220

Zoom: https://stonybrook.zoom.us/j/94798224254?pwd=CFraer25qnpORbJ14aAVHRwaSJOjJM.1




https://stonybrook.zoom.us/j/91775729097pwd=Qlc5Nks0NmlyKzJwMjR0S0hrdVZ3QT09

Meeting ID: 917 7572 9097
Passcode: 555459


Abstract: As the saying goes, there are many ways to skin a cat.
While we don't want to go around skinning cats, the world of
optimization is rich with different problems, problem formulations,
and methods and approaches, each with different guarantees and
computational benefits. In this talk we will take a tour down the
problem of structured sparsity in sensing to see how one simple
problem can inspire a wide range of analysis and tools. First, I will
present the optimality conditions for a generalized structured sparse
problem, which can be geometrically visualized as alignment of vectors
and matrices. Then I will introduce three approximation methods for
the problem of phase retrieval, which are a twist on stochastic
gradient and coordinate descent methods. These methods leverage
fundamental numerical linear algebra concepts to give fast approximate
solutions to large-scale problems, which then after postprocessing can
produce more reliable sensing results.

Bio: Yifan Sun received her PhD in Electrical Engineering from the
University of California Los Angeles in 2015, with research focusing
on convex optimization and semidefinite programming. She was then
Technicolor Research and Innovation, focusing on machine learning and
data science applications. More recently, she completed two postdocs,
at the University of British Columbia in Vancouver, Canada and
L'Institut National de Recherche en Informatique et Automatique
(INRIA) in Paris, France.
Abstract: DeepSeek-R1-Zero has shown that reinforcement learning (RL) at scale can directly enhance the reasoning capabilities of LLMs without supervised fine-tuning. In this work, we critically examine R1-Zero-like training by analyzing its two core components: base models and RL. We investigate a wide range of base models, including DeepSeek-V3-Base, to understand how pretraining characteristics influence RL performance. Our analysis reveals that DeepSeek-V3-Base already exhibit ''Aha moment'', while Qwen2.5 base models demonstrate strong reasoning capabilities even without prompt templates, suggesting potential pretraining biases. Additionally, we identify an optimization bias in Group Relative Policy Optimization (GRPO), which artificially increases response length (especially for incorrect outputs) during training. To address this, we introduce Dr. GRPO, an unbiased optimization method that improves token efficiency while maintaining reasoning performance. Leveraging these insights, we present a minimalist R1-Zero recipe that achieves 43.3% accuracy on AIME 2024 with a 7B base model, establishing a new state-of-the-art.

Speaker: Md. Saqib Hasan

Location: CS2311
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
Join the Center of Excellence in Wireless and Information Technology (CEWIT) and their co-host IEEE-USA for a livestream panel discussion on Generative Artificial Intelligence (Gen AI). In this engaging livestream, we will dive into the technologies that continue to transform what is possible and explore the dynamic intersection of innovation, creativity, ethics, and Gen AI.

CEWIT is joined by Stony Brook University experts who will provide their insights and perspectives on this rapidly changing technology.

Meet the Panel

Laura Lindenfeld, PhD

Executive Director
Alan Alda Center for Communicating Science®
Dean
School of Communication & Journalism
BIO

Margaret Schedel, PhD
Associate Professor
Composition and Computer Music
Co-Founder
Lyrai
BIO

Steven Skiena, PhD

Interim Director
AI Innovation Institute
Distinguished Professor
Computer Science
BIO

Vivian Zhang
CTO/School Director
NYC Data Science Academy
Chief Data Officer
GoDental.ai
BIO


Register here.
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 656 Seminars in Computer Vision - Wednesdays 11:30am-12:50pm, Room NCS 120

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 CSE656. 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.

The first meeting will be Wed Jan 29 at 11.30am, room 120 New CS. The meeting will deal with organizational matters and we will start right away with some presentations. Send David Paredes Merino <dparedesmeri@cs.stonybrook.edu> an email if you are interested but cannot attend the first meeting. Please forward to people outside the CS department that you think might be interested.

You are cordially invited to attend the biweekly Brookhaven AI Mixer (BAM). BAM includes one short talk on AI research happening at BNL, followed by an open mixer over coffee and snacks for everyone to network and discuss all things AI. The first half hour will consist of presentations that will be available via ZOOM, and the second half hour will be for in person only networking.

Join us every other Tuesday at noon in CDSD's Training Room (building 725, 2nd floor) to learn about interesting AI methods and applications, engage with potential collaborators, prepare for pending FASST funding calls, and build a community of AI for Science at BNL.

Experiencing Machine Learning in Collider-Accelerator Control System

Abstract: The Relativistic Heavy Ion Collider (RHIC) at Collider-Accelerator Department (C-AD) of BNL provides the world's only high-energy polarized proton beam. It is in the unique position to study where nuclei obtain their spin. During 25 years of operation at RHIC, the C-AD controls group has developed its own control system to tune the accelerator performance, which contains millions of control points. The successful operation of this system will highly affect the machine performance. RHIC's successor, the Electron-Ion Collider (EIC), will be one of the most complex scientific instruments ever built, with the capability of colliding polarized proton and electron beams. The increasing complexity of instruments will require new, sophisticated control methods/tools to tune and optimize the accelerator performance. In this talk, I will summarize some projects developed in recent years that utilize machine learning in the C-AD controls group.

Biography: Dr. Yuan Gao is an assistant scientist at the Collider-Accelerator Department (C-AD) at Brookhaven, primarily working on developing new machine learning schemes in the control group to enhance system performance. His research interests include game theory, algorithm design, anomaly detection, and simulation modeling.

Location: CDS, Bldg. 725, Training Room

Join ZoomGov Meeting: https://bnl.zoomgov.com/j/1604302440?pwd=0x2I95PIvbkkzIi6rA0MNnon5k2sux.1

Meeting ID: 160 430 2440
Passcode: 478223

Stony Brook, NY, Nov 12, 2026 - In a recent study published in IEEE Transactions on Visualization and Computer Graphics, Stony Brook researchers analyzed how artificial intelligence connects concepts with colors in semantically meaningful ways. As humans associate “ocean” and “peace” with blue, and “cherry” and “anger” with red, can AI also represent nuanced, semantically resonant colors for diverse tangible and abstract concepts?