The infrastructure decisions being made around AI today will shape competitive advantage, cost, control, and institutional trust for years to come.

Organizations will invest in AI either way. The strategic question is whether they remain dependent on systems they cannot fully inspect or influence, or help shape shared infrastructure that they can adapt, govern, and build upon.

We would like to invite you to a live executive briefing from the AI Alliance. The discussion will focus on what leaders should understand now:

  • The state of the movement: how 205 organizations across 29 countries are collaborating on more than 30 open projects.
  • Project Tapestry: our global initiative to build frontier-capable, open-sourced AI through shared development, while allowing participating institutions to retain control over their data, resources, and strategic priorities.
  • The business and policy landscape: how cost, sovereignty, trust, and access are reshaping AI strategy.
  • Paths to participation: how organizations can contribute through the AI Alliance Open R&D Community and the Open Source AI Innovation Association.

Register here.
TITLE: Towards a Theory of Encode/Decoder Architectures by Andrej Risteski of CMU

ABSTRACT: A common choice of architecture in representation learning (i.e., learning a good embedding of the data) is an encoder/decoder architecture, which tries to map a part of the input into a good latent representation (via an encoder), and predict the remaining part of the input (via a decoder). Two common examples are universal machine translation: where one tries to learn to translate between any pair of a set of languages via a common latent language, given paired up corpora for only a part of the pairs; and contextual encoders -- where one tries to predict a part of the image, given the rest of the image.
 
We will give a framework for analyzing the sample complexity of such architectures -- i.e., how many pairs of languages do we need to have paired up corpora for? How many image prediction tasks do we have to solve to get a good representation?
Defending Software Systems from Cyber Attack Campaigns Presented by R. Sekar The DNC hack of 2016, the Equifax breach of 2017, and the spate of ransomware campaigns in 2019 demonstrate the formidable challenges we face in securing our network and software systems against highly stealthy and sophisticated adversaries. In this talk, I will describe two avenues of research we have been pursuing to help tilt the table against such powerful adversaries. The first is software hardening techniques that make software vulnerabilities harder to exploit. To maximize their applicability and ease of use, our techniques are implemented into compilers, or they directly transform binary code. I will outline some of the exciting new developments we have had in this area over the years, including randomization, memory safety, information-flow tracking, control-flow integrity, and code-pointer integrity. We complement this first line of defense with techniques for analyzing and understanding attack campaigns that manage to slip past all deployed defenses. Our techniques can sift through logs consisting of hundreds of millions of events to zoom in on attack activity that may span just a few hundred events. I will describe our experience in mapping out several DARPA-sponsored red team attack campaigns.
Mind Brain Lecture: Constructing the World of Taste in Your Head You fork the morsel into your mouth and say yum...chocolate cake. The appreciation of your dessert's taste seems to follow directly, quickly and simply from the placement of the food on your tongue. The truth, however, is far more interesting and complex: your brain actually begins determining whether you will enjoy a bite of food even before the fork approaches your mouth and continues to work the problem well after. Information about your food's color, smell, texture and taste activates multiple parts of your brain, where that information collides with your pre-mouthful beliefs about how it should taste. The coming-together and shuffling of that information around the brain takes time, as networks of neurons work together to help you decide whether the morsel in your mouth is worth swallowing. Referring to work from psychology, biology and computational neuroscience, Professor Katz will de-mystify and reveal the beauty of these complexities of the neuroscience of taste. Donald Katz, Professor of Psychology, Departments of Neuroscience, Psychology, and the Volen National Center for Complex Systems, Brandeis University Free presentation intended for a general audience. Reception to follow. https://www.stonybrook.edu/commcms/mind/

Understand Prompting the crucial part to interface with models

Discover how to prompt effectively by exploring the details behind your AI interactions. This isn't just about basic prompting; it's about understanding how to articulate your ideas clearly. We'll showcase a few prompts and how they work. Discover how giving AI the right details can truly boost your productivity and help you reclaim valuable time in your day.

In this session, you will

  1. Utilize AI models effectively
  2. Understanding different prompts
  3. Find out tips that we use with AI

Register: https://stonybrookuniversity.co1.qualtrics.com/jfe/form/SV_dht1o3rNzlZhHka?source=event+manager&session=0805251000ai

Abstract:

Many real world complex problems are multi-step reasoning tasks. These range from analytic tasks such as answering questions to automation tasks where agents complete tasks on behalf of users.. Evaluation, datasets, and models for such tasks can be unreliable for multiple reasons. (i) Datasets often have annotation artifacts and biases, allowing models to take reasoning shortcuts. Such shortcuts can allow models to make effective guesses -- or, in a sense, cheat -- to achieve high performance without any multi-step reasoning. This issue is further exacerbated for complex tasks because as the number of the required reasoning steps increases, so do the avenues for bypassing those steps. (ii) Models trained on such dataset/s learn to solve the task by taking reasoning shortcuts instead of proper multi-step reasoning. As a result, these models are not robust (reliable) when evaluated in an out-of-distribution evaluation setting. (iii) Lastly, recent works have shown that language models can solve complex multi-step tasks by producing a step-by-step explanation without any training. However, these methods often hallucinate factually incorrect (i.e., unreliable) explanations when posed with knowledge-intensive tasks.

I address these challenges by carefully characterizing the requirements of robust multi-step reasoning and designing reliable evaluation datasets and training methods that necessitate thorough multi-step reasoning. In DiRe, I first formalize and introduce Disconnected Reasoning, i.e., reasoning that allows models to arrive at the correct answer by bypassing necessary reasoning steps, and use this formalization to measure how much multi-step reasoning a model does on a dataset. In MuSiQue, I built a multi-step reasoning dataset for QA from scratch that avoids cheatability via disconnected reasoning, providing a more reliable evaluation. In TeaBReaC, I developed a synthetically generated multi-step QA pretraining dataset designed to force models to avoid disconnected reasoning and learn reliable multi-step reasoning. In IRCoT, I address the reliability of model-generated multi-step reasoning chains by interleaving models' step-by-step reasoning with a step-by-step retrieval from an external corpus, resulting in more factually correct reasoning. Finally, in AppWorld, I built a multi-step reasoning dataset that requires highly interactive problem-solving in an environment carefully designed to ensure models need thorough reasoning to succeed.
Speaker: Harsh Trivedi

Location: NCS 220 or Zoom

https://stonybrook.zoom.us/j/99096379762?pwd=zYCJZQVxRuZd9BboscO4nlodCwsKBr.1

Join the Collaborative for the Earth, the Department of Political Science, and the AI Innovation Institute as we welcome Lav Varshney, Ph.D., Della Pietra Infinity Professor and Inaugural Director of the AI Innovation Institute.

The Physical Economy of Intelligence: Marketcraft, Data Centers, and Climate

Artificial intelligence is often described as immaterial, yet it is a rapidly growing physical industry built from memory chips, servers, concrete, land, electricity, water, and capital. This talk develops a marketcraft perspective on AI and climate, treating memory, compute, and intelligence as interdependent goods shaped by supply constraints, demand shocks, long investment cycles, environmental externalities, and the design of markets and institutions. This framing helps explain why sustainable AI cannot be achieved through more efficient algorithms alone: it also requires better lifecycle accounting, procurement standards, flexible computing, and contracts that shift computational activity toward environmentally favorable times and places. The talk will connect these system-level ideas to work on generative design and experimental validation of low-carbon concrete for data-center construction, as well as AI-enabled optimization of wastewater operations that have each reduced energy/carbon by 30 percent or more. Together, these examples show how the co-design of algorithms, physical infrastructure, markets, and public policy can reduce AI's environmental footprint while using AI to accelerate climate solutions.

Date & Time: Wednesday, September 23, 2026 | 12:30 PM - 1:50 PM

Location: Laufer Center

Registration is compulsory

For any questions and accessibility requests or accommodations, please contact Jennifer Gilday at c4e@stonybrook.edu or call 631-632-4625.

Abstract: Making large language models usable requires post-training methods that align models to human values, robustly handle underspecified inputs, and generalize to diverse instructions. This talk addresses the challenge of developing responsible AI through post-training from the following angles. First, I address contextual robustness. Preference data, for example, is often underspecified, and I show how underspecification in preference data can lead to diverging preferences. Standard reward models fail to properly handle these disagreements, often making decisive choices even when human annotators are split. I argue that more consequential outputs demand more context, and propose clarification question generation as one solution. Second, I will discuss how we can train models to be better instruction followers. I will show that most models severely overfit on a small set of instruction-following constraints and are not able to generalize well to unseen output constraints. I propose to train models with reinforcement learning from verifiable rewards for verifiable instruction following, and show how this leads to improved generalization on constraint following. Throughout the presentation, I will outline how I have applied these insights into developing open generative models, like Tülu and OLMo, and I will conclude with my research agenda for responsible post-training: critical AI evaluation, broader generalization, and expanding what models can reliably do.


Bio: Valentina Pyatkin is a postdoctoral researcher at the Allen Institute for AI and the University of Washington, advised by Prof. Hanna Hajishirzi and Prof. Yejin Choi. Additionally, she is part- time affiliated with the ETH AI Center, where she mentors students and works on post-training for the Swiss AI Initiative. She obtained her PhD in Computer Science from Bar Ilan University. Her work has been awarded an ACL Outstanding Paper Award and the ACL Best Theme Paper Award, and has been supported by a Schmidt Sciences Postdoctoral Award. During her doctoral studies, she conducted research internships at Google and the Allen Institute for AI, where she received the AI2 Outstanding Intern of the Year Award.

Location: NCS 120

The Association for Computational Linguistics is the international scientific and professional society for people working on problems involving natural language and computation. Membership includes the ACL quarterly journals, Computational Linguistics and Transactions of the ACL, reduced registration at most ACL-sponsored conferences, discounts on ACL-sponsored publications, and participation in ACL Special Interest Groups.

An annual meeting is held each summer in locations where significant computational linguistics research is carried out.

For more information and registration, visit the official website.