This workshop synthesizes the latest research on the impact of AI usage in education so that you could make informed decisions on whether and how to use AI to facilitate your learning. You might have seen conflicting reports on whether the use of AI is good for learning. In this workshop, we are going to tease out, drawing on the latest research, which types of AI usage are beneficial or harmful for different kinds of learning. At the end of the workshop, you should walk away with more clarity on when and how to use AI for your own learning. Join PRODIG+ fellow on critical AI, Zheng Fu, in this informative workshop.
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. 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.
We meet every other Tuesday at noon in CDSD's Training Room (building 725, room 2-124) 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.
Abstract: Identifying model Hamiltonians is a vital step toward creating predictive models of materials. We combined Bayesian optimization with the EDRIXS numerical package to infer Hamiltonian parameters from resonant inelastic X-ray scattering (RIXS) spectra within the single atom approximation. To evaluate the efficacy of our method, we tested it on experimental RIXS spectra for several materials and demonstrated that it can reproduce results obtained from hand-fitted parameters to a precision similar to expert human analysis while providing a more systematic mapping of parameter space. Our work provides a key first step toward solving the inverse scattering problem to extract effective multi- orbital models from information-dense RIXS measurements, which can be applied to a host of quantum materials.
Biography: Marton Lajer is a postdoctoral researcher at the Condensed Matter Physics and Materials Science Department, Brookhaven National Laboratory. Marton obtained his PhD in theoretical physics at the Eotvos Lorand University, Hungary, in 2021. He was a junior research fellow at the Wigner Research Centre for Physics in Budapest before joining BNL in September 2022. His background spans various analytical and performance-critical numerical methods, mostly in the context of low- dimensional quantum field theories and quantum many-body systems. His research currently focuses on incorporating AI-enhanced methods to various problems in inelastic spectroscopy.
In addition to our speaker, we will have a number of CDS staff in attendance with expertise in AI methods and applications including image analysis, foundation models development, and inverse problem solving.
Location: CDS, Bldg. 725, Training Room
Join ZoomGov Meeting: https://bnl.zoomgov.com/j/1604383624?pwd=ffQ5cUPNxTI7nzClKQO6cnsNbhF9Vf.1
Meeting ID: 160 438 3624
Passcode: 558449
Please Note: Due to a funding shortfall, we are for the time being no longer able to provide pizza and sodas for these events. We will have coffee though, and all are of course welcome to bring their lunch.
Abstract: Self-supervised representation learning (SRL) has emerged as a pivotal advancement in machine learning, offering high-quality data representations without the need for labeled datasets. While SRL has demonstrated enhanced adversarial robustness compared to supervised learning, its resilience against other attack types, particularly backdoor attacks, remains an open question. Recent studies have revealed potential vulnerabilities in SRL, underscoring the necessity for a comprehensive security analysis. However, existing research often extrapolates attacks from supervised learning paradigms, neglecting the unique challenges and opportunities inherent to self-supervised mechanisms.
This thesis proposal aims to address three critical objectives in the realm of self-supervised learning: (1) exploring novel attack vectors, (2) implementing and evaluating practical attacks, and (3) developing robust countermeasures. We focus on two key SRL paradigms: Contrastive Learning and Diffusion Models. For Contrastive Learning, we synthesize existing security vulnerabilities and introduce innovative attack vectors, such as CTRL, to uncover distinctive risks. We conduct a comparative analysis of contrastive and supervised learning approaches in their defense against these threats, exploring potential safeguards and highlighting the limitations of current protective measures in self-supervised contexts. Regarding Diffusion Models, we demonstrate inherent vulnerabilities in their application to adversarial purification.
Our research aims to illuminate the unique challenges posed by emerging attack vectors in self-supervised learning, fostering technical advancements to address underlying security risks in real-world applications. By contributing to the development of more resilient and secure self-supervised representation learning systems, we seek to enhance their reliability and trustworthiness in practical scenarios. This comprehensive examination of SRL's security landscape will provide valuable insights for the broader machine-learning community and pave the way for more robust AI systems.
Join here.
This thesis proposal aims to address three critical objectives in the realm of self-supervised learning: (1) exploring novel attack vectors, (2) implementing and evaluating practical attacks, and (3) developing robust countermeasures. We focus on two key SRL paradigms: Contrastive Learning and Diffusion Models. For Contrastive Learning, we synthesize existing security vulnerabilities and introduce innovative attack vectors, such as CTRL, to uncover distinctive risks. We conduct a comparative analysis of contrastive and supervised learning approaches in their defense against these threats, exploring potential safeguards and highlighting the limitations of current protective measures in self-supervised contexts. Regarding Diffusion Models, we demonstrate inherent vulnerabilities in their application to adversarial purification.
Our research aims to illuminate the unique challenges posed by emerging attack vectors in self-supervised learning, fostering technical advancements to address underlying security risks in real-world applications. By contributing to the development of more resilient and secure self-supervised representation learning systems, we seek to enhance their reliability and trustworthiness in practical scenarios. This comprehensive examination of SRL's security landscape will provide valuable insights for the broader machine-learning community and pave the way for more robust AI systems.
Join here.
As part of a grant project funded by the AI3 Institute, a group of instructors participated in a faculty development program, Fostering Writing-to-Learn Skills with Critical AI Literacy: A Faculty Development and Student Support Program. This program was developed to support instructors across campus with navigating/integrating AI in their courses specifically around writing intensive/involved assignments. We would like to invite anyone interested to the culmination of this program, a mini-symposium, where the participants will share practical changes they made or are making around writing intensive/involved assignments and AI.
Location: Wang 201
A light lunch will be served. Please register by Friday, November 7th.
Face Editing with Machine Learning presented by Zhixin Shu
ABSTRACT: The face is the most informative feature of humans and has been a long-standing research topic in Computer Vision and Graphics. Images of faces are also ubiquitous in photography and social media, and people have devoted significant resources to capturing and editing face images. Face editing can be broadly viewed as the encoding, manipulation and the decoding of some representations for face images. The challenges are that we want to manipulate an image in a controllable way and generate results that are both desirable and as realistic as possible. This thesis explores different Machine Learning-based face-editing approaches. I discuss the role of machine learning for achieving desirable edits by learning both the physical aspects as well as the statistical manifold of human faces. In my work for eye-editing, I discuss the importance of understanding multiple physical elements of a face image, such as shape, illumination, pose, etc. In a deep-learning-based approach, I introduce image formation domain knowledge to the construction and training of a neural network. This network provides transparent access to the disentangled representations of the aforementioned physical properties. With this network, we can achieve various face editing tasks in forms of representation manipulation. After that, I introduce Deforming Autoencoders, a network that learns to disentangle shape and appearance in an unsupervised manner. This disentanglement is beneficial for the learning of some other factors of variations, such as illumination and facial expression. In an extension of Deforming Autoencoders, we incorporate non-rigid structure-from-motion to learn a 3D morphable model for faces that only requires an image set for training. At last, I describe an image-to-image network for 3D face reconstruction, which also utilizes structure-from-motion in deep learning. With real face images in training, this network not only reconstructs 3D faces more accurately than prior art but also has better generalization ability in real-life testing cases.
ABSTRACT: The face is the most informative feature of humans and has been a long-standing research topic in Computer Vision and Graphics. Images of faces are also ubiquitous in photography and social media, and people have devoted significant resources to capturing and editing face images. Face editing can be broadly viewed as the encoding, manipulation and the decoding of some representations for face images. The challenges are that we want to manipulate an image in a controllable way and generate results that are both desirable and as realistic as possible. This thesis explores different Machine Learning-based face-editing approaches. I discuss the role of machine learning for achieving desirable edits by learning both the physical aspects as well as the statistical manifold of human faces. In my work for eye-editing, I discuss the importance of understanding multiple physical elements of a face image, such as shape, illumination, pose, etc. In a deep-learning-based approach, I introduce image formation domain knowledge to the construction and training of a neural network. This network provides transparent access to the disentangled representations of the aforementioned physical properties. With this network, we can achieve various face editing tasks in forms of representation manipulation. After that, I introduce Deforming Autoencoders, a network that learns to disentangle shape and appearance in an unsupervised manner. This disentanglement is beneficial for the learning of some other factors of variations, such as illumination and facial expression. In an extension of Deforming Autoencoders, we incorporate non-rigid structure-from-motion to learn a 3D morphable model for faces that only requires an image set for training. At last, I describe an image-to-image network for 3D face reconstruction, which also utilizes structure-from-motion in deep learning. With real face images in training, this network not only reconstructs 3D faces more accurately than prior art but also has better generalization ability in real-life testing cases.
CSE 600 Seminar Series | Fall 2025
Abstract: Imagine machines that can see the invisible: drones locating wildfire survivors, cameras predicting building failures, and smartphones detecting skin tumors. These applications lie beyond today's vision systems, which focus only on human-visible information. In this talk, I argue that a wealth of scene information is hidden in light properties invisible to the human eye, such as the travel time of photons and polarization of light waves. I will present how co- designing camera hardware, graphics models, and learning algorithms unlocks these invisible properties to create superhuman vision systems. I will present three superhuman vision capabilities: seeing around blind corners, turning objects into cameras, and extracting internal stress fields. By analyzing faint light reflections on diffuse walls and shiny objects, we create virtual cameras that reveal scenes hidden from the line of sight - enabling autonomous systems to navigate safely. Using the polarization of light, we recover mechanical stress fields hidden inside objects - opening new possibilities for non-destructive material characterization. These capabilities point toward a future where machines can see the invisible: around us, beneath our bodies, and beyond our scientific understanding.
Bio:
Akshat Dave is an Assistant Professor in the Department of Computer Science at Stony Brook
University, USA. His research lies at the intersection of applied optics, computer vision, and
machine learning. His work has been recognized by Rice University's Best Thesis Award, Optica Best Paper Prize, SIGGRAPH Asia Doctoral Consortium, and fellowships by Qualcomm, Texas Instruments, and INK Global Foundation. Prior to Stony Brook, he was a Postdoctoral Associate at MIT Media Lab. He holds a Ph.D. from Rice University and a Masters and a Bachelors from Indian Institute of Technology Madras.
Abstract: Imagine machines that can see the invisible: drones locating wildfire survivors, cameras predicting building failures, and smartphones detecting skin tumors. These applications lie beyond today's vision systems, which focus only on human-visible information. In this talk, I argue that a wealth of scene information is hidden in light properties invisible to the human eye, such as the travel time of photons and polarization of light waves. I will present how co- designing camera hardware, graphics models, and learning algorithms unlocks these invisible properties to create superhuman vision systems. I will present three superhuman vision capabilities: seeing around blind corners, turning objects into cameras, and extracting internal stress fields. By analyzing faint light reflections on diffuse walls and shiny objects, we create virtual cameras that reveal scenes hidden from the line of sight - enabling autonomous systems to navigate safely. Using the polarization of light, we recover mechanical stress fields hidden inside objects - opening new possibilities for non-destructive material characterization. These capabilities point toward a future where machines can see the invisible: around us, beneath our bodies, and beyond our scientific understanding.
Bio:
Akshat Dave is an Assistant Professor in the Department of Computer Science at Stony Brook
University, USA. His research lies at the intersection of applied optics, computer vision, and
machine learning. His work has been recognized by Rice University's Best Thesis Award, Optica Best Paper Prize, SIGGRAPH Asia Doctoral Consortium, and fellowships by Qualcomm, Texas Instruments, and INK Global Foundation. Prior to Stony Brook, he was a Postdoctoral Associate at MIT Media Lab. He holds a Ph.D. from Rice University and a Masters and a Bachelors from Indian Institute of Technology Madras.
Abstract:
Recent advances in deep learning have significantly enhanced the capabilities of Natural Language Processing (NLP) and Vision-Language Models (VLMs). However, these advancements come with increased vulnerabilities, notably through backdoor attacks that pose severe security threats. This thesis addresses two critical dimensions of Trustworthy AI and Efficient Multimodal Representation Learning: (1) security through analyzing, detecting, and designing backdoor attacks in NLP and VLMs, and (2) efficiency through advanced multimodal representation methods tailored for clinical and medical imaging applications.
In the first dimension, we explore the internal mechanisms exploited by backdoor attacks, identifying the distinctive phenomenon of attention focus drifting in compromised transformer models, where trigger tokens consistently hijack attention. Leveraging these insights, we propose robust detection frameworks, including the attention-based Trojan detector (AttenTD) and a task-agnostic logit-based detection method (TABDet), achieving effective identification of backdoored NLP models across diverse tasks. We further introduce novel backdoor attack methodologies: the Trojan Attention Loss (TAL), enhancing attack efficiency and stealth through direct attention manipulation, and BadCLM, demonstrating critical vulnerabilities in clinical decision-support systems by effectively compromising clinical language models.
Extending our security exploration to multimodal settings, we investigate backdoor attacks on Vision-Language Models (VLMs), particularly in complex image-to-text generation tasks, proposing innovative techniques (TrojVLM, VLOOD) capable of embedding backdoors without direct access to original training data, thus showcasing practical risks in real-world scenarios.
In the second dimension, we address efficiency and interpretability challenges in clinical and pathology applications. We introduce TCP-LLaVA, the first multimodal large language model (MLLM) designed explicitly for Whole Slide Image (WSI) Visual Question Answering (VQA). Utilizing a novel token compression mechanism inspired by transformer-based models, TCP-LLaVA substantially reduces computational resource consumption while maintaining superior VQA performance across multiple tumor subtypes. Additionally, we present a multimodal transformer model integrating structured Electronic Health Records (EHR) with clinical notes, demonstrating enhanced predictive accuracy and interpretability for in-hospital mortality prediction through integrated gradient-based interpretability methods.
Together, these contributions present a comprehensive approach to ensuring AI models are not only secure against malicious manipulation but also efficient and interpretable for critical clinical applications, underscoring the essential need for trustworthy and effective AI systems.
Speaker: Weimin Lyu
Zoom: https://stonybrook.zoom.us/j/
Meeting ID: 239 232 6575
Passcode: 436192
Recent advances in deep learning have significantly enhanced the capabilities of Natural Language Processing (NLP) and Vision-Language Models (VLMs). However, these advancements come with increased vulnerabilities, notably through backdoor attacks that pose severe security threats. This thesis addresses two critical dimensions of Trustworthy AI and Efficient Multimodal Representation Learning: (1) security through analyzing, detecting, and designing backdoor attacks in NLP and VLMs, and (2) efficiency through advanced multimodal representation methods tailored for clinical and medical imaging applications.
In the first dimension, we explore the internal mechanisms exploited by backdoor attacks, identifying the distinctive phenomenon of attention focus drifting in compromised transformer models, where trigger tokens consistently hijack attention. Leveraging these insights, we propose robust detection frameworks, including the attention-based Trojan detector (AttenTD) and a task-agnostic logit-based detection method (TABDet), achieving effective identification of backdoored NLP models across diverse tasks. We further introduce novel backdoor attack methodologies: the Trojan Attention Loss (TAL), enhancing attack efficiency and stealth through direct attention manipulation, and BadCLM, demonstrating critical vulnerabilities in clinical decision-support systems by effectively compromising clinical language models.
Extending our security exploration to multimodal settings, we investigate backdoor attacks on Vision-Language Models (VLMs), particularly in complex image-to-text generation tasks, proposing innovative techniques (TrojVLM, VLOOD) capable of embedding backdoors without direct access to original training data, thus showcasing practical risks in real-world scenarios.
In the second dimension, we address efficiency and interpretability challenges in clinical and pathology applications. We introduce TCP-LLaVA, the first multimodal large language model (MLLM) designed explicitly for Whole Slide Image (WSI) Visual Question Answering (VQA). Utilizing a novel token compression mechanism inspired by transformer-based models, TCP-LLaVA substantially reduces computational resource consumption while maintaining superior VQA performance across multiple tumor subtypes. Additionally, we present a multimodal transformer model integrating structured Electronic Health Records (EHR) with clinical notes, demonstrating enhanced predictive accuracy and interpretability for in-hospital mortality prediction through integrated gradient-based interpretability methods.
Together, these contributions present a comprehensive approach to ensuring AI models are not only secure against malicious manipulation but also efficient and interpretable for critical clinical applications, underscoring the essential need for trustworthy and effective AI systems.
Speaker: Weimin Lyu
Zoom: https://stonybrook.zoom.us/j/
Meeting ID: 239 232 6575
Passcode: 436192
As artificial intelligence continues to transform higher education and the world beyond, how are students engaging with this change? Join us for a student-led discussion that explores how AI is influencing academic integrity, learning practices, and students' perspectives on its role in future workplaces.
Our panelists will share their experiences and reflections on questions such as:
1. What counts as appropriate and inappropriate use of AI in coursework?
2. How do faculty approach AI and talk about its implications in class?
3. What does AI mean for students' learning and ethical decision-making?
4. How are students building their understanding of AI tools and their potential uses in professional contexts?
This conversation offers an authentic look at how students are navigating the promises and challenges of AI--both in their studies and as they look ahead to applying these technologies responsibly in their fields.
Register here.
Our panelists will share their experiences and reflections on questions such as:
1. What counts as appropriate and inappropriate use of AI in coursework?
2. How do faculty approach AI and talk about its implications in class?
3. What does AI mean for students' learning and ethical decision-making?
4. How are students building their understanding of AI tools and their potential uses in professional contexts?
This conversation offers an authentic look at how students are navigating the promises and challenges of AI--both in their studies and as they look ahead to applying these technologies responsibly in their fields.
Register here.
AI can help you write, you hear. AI can save you time, leverage your skills, enhance your productivity. . . . But you also hear: AI output is not reliable, not adequate for advanced tasks/learning, not ethical to use -- you could get in deep trouble for using AI tools without adequate mastery and caution. Which way is it?
Come join this hands-on workshop where you will explore AI tools and their affordances. Engage in writing tasks to learn how to use AI tools effectively and responsibly.
Sign up for a seat now: https://docs.google.com/forms/d/e/1FAIpQLSd0iDTKkTYnkxFd4LkgqbtP97zQSS4FI_MiPVm7p6IY5SGwSg/viewform
Come join this hands-on workshop where you will explore AI tools and their affordances. Engage in writing tasks to learn how to use AI tools effectively and responsibly.
Sign up for a seat now: https://docs.google.com/forms/d/e/1FAIpQLSd0iDTKkTYnkxFd4LkgqbtP97zQSS4FI_MiPVm7p6IY5SGwSg/viewform
Do Natural Language Understanding Systems Learn to Understand or to
Find Shortcuts? (Naoya Inoue, http://naoya-i.github.io/)
ABSTRACT: Recent studies have suggested that natural language understanding (NLU) systems learn to exploit superficial, task-unrelated cues (a.k.a. annotation artifacts) in current datasets. This prevents the community from reliably measuring the progress of NLU systems. In this talk, I will discuss two latest studies from our research team: (i) analysis of annotation artifacts in commonsense causal reasoning and (ii) creation of benchmark for evaluating NLU systems' internal reasoning.
---------------------------------------------------------------------------------------------------------------------------------------------
---------------------------------------------------------------------------------------------------------------------------------------------
Learning graph-structured sparse models (Baojian Zhou, https://baojianzhou.github.io/)
ABSTRACT: Learning graph-structured sparse models has recently received significant attention thanks to their broad applicability to many important real-world problems. However, such models, of more effective and stronger interpretability compared with their counterparts, are difficult to learn due to optimization challenges. In this talk, we will discuss how to learn graph-structured sparse models under stochastic and online learning settings. Some interesting related problems will also be discussed.
Find Shortcuts? (Naoya Inoue, http://naoya-i.github.io/)
ABSTRACT: Recent studies have suggested that natural language understanding (NLU) systems learn to exploit superficial, task-unrelated cues (a.k.a. annotation artifacts) in current datasets. This prevents the community from reliably measuring the progress of NLU systems. In this talk, I will discuss two latest studies from our research team: (i) analysis of annotation artifacts in commonsense causal reasoning and (ii) creation of benchmark for evaluating NLU systems' internal reasoning.
---------------------------------------------------------------------------------------------------------------------------------------------
---------------------------------------------------------------------------------------------------------------------------------------------
Learning graph-structured sparse models (Baojian Zhou, https://baojianzhou.github.io/)
ABSTRACT: Learning graph-structured sparse models has recently received significant attention thanks to their broad applicability to many important real-world problems. However, such models, of more effective and stronger interpretability compared with their counterparts, are difficult to learn due to optimization challenges. In this talk, we will discuss how to learn graph-structured sparse models under stochastic and online learning settings. Some interesting related problems will also be discussed.