Machine Learning and AI Seminar Series

Archive: Speaker Abstracts

• Andrej Risteski Carnegie Mellon University

  Title: Architectural Choices in Scientific ML: A View Through the Lens of Theory

   

  Abstract: In deep learning, small architectural changes—such as residual connections or normalization layers—have often had outsized impact. This talk examines how similar effects arise in recent applications of deep learning to the sciences. The central theme is that the architectural changes we identify are not suggested by current benchmarks, which remain much less mature than they are in image or language domains. Instead, they become visible through the right theoretical lenses. We will showcase several vignettes spanning graph neural networks (GNNs), time-dependent partial-differential equations (PDEs), and steady-state PDEs.

   

  The first setting concerns graphs with bottlenecks or hubs: augmenting GNNs with edge-level state yields (provable) gains under constraints on depth and memory. We establish this using techniques from time–space tradeoffs in theoretical computer science, and show that neither “symmetry-only” theoretical accounts nor standard GNN benchmarks would detect this separation. The next setting concerns time-dependent PDEs, where adding an explicit memory layer via state-space models (e.g. S4) has negligible effect under full observability, but substantial impact under partial observation. This kind of phenomenon is predicted by Mori–Zwanzig theory—which also inspired the architectural change. Finally, in steady-state PDEs and operator learning, we show that Deep Equilibrium Model (DEQ)-based architectural changes have efficiency and robustness benefits. Here, the design is motivated by representation-theoretic constructions that simulate “unrolled” gradient descent in function space.

   

  Based on several works including:

   

  Rohatgi, D., Marwah, T., Lipton, Z. C., Lu, J., Moitra, A., & Risteski, A. (2024). Towards characterizing the value of edge embeddings in Graph Neural Networks. arXiv preprint arXiv:2410.09867

   

  Ruiz, R. B., Marwah, T., Gu, A., & Risteski, A. (2024). On the benefits of memory for modeling time-dependent pdes. arXiv preprint arXiv:2409.02313

   

  Marwah, T., Pokle, A., Kolter, J. Z., Lipton, Z., Lu, J., & Risteski, A. (2023). Deep equilibrium based neural operators for steady-state pdes. Advances in Neural Information Processing Systems, 36, 15716-15737

   

  Marwah, T., Lipton, Z. C., Lu, J., & Risteski, A. (2023, July). Neural network approximations of pdes beyond linearity: A representational perspective. In International Conference on Machine Learning (pp. 24139-24172). PMLR

   

  Marwah, T., Lipton, Z., & Risteski, A. (2021). Parametric complexity bounds for approximating PDEs with neural networks. Advances in Neural Information Processing Systems, 34, 15044-15055

   

  Talk Date: Friday, November 21, 2025

• Florentin Guth NYU and Flatiron Institute

  Title: Learning normalized probability models with dual score matching

   

  Abstract: Learning probability models from data is at the heart of many learning tasks. We introduce a new framework for learning normalized energy (log probability) models inspired from diffusion generative models. The energy model is fitted to data by two “score matching” objectives: the first constrains the gradient of the energy (the “score”, as in diffusion models), while the second constrains its *time derivative* along the diffusion. We validate the approach on both synthetic and natural image data: in particular, we show that the estimated log probabilities do not depend on the specific images used during training. Finally, we demonstrate that both image probability and local dimensionality vary significantly with image content, challenging simple interpretations of the manifold hypothesis.

   

  Talk Date: Friday, November 7, 2025

• Furong Huang University of Maryland

  Title: Rethinking Test-Time Thinking: From Token-Level Rewards to Robust Generative Agents

   

  Abstract: We present a unified perspective on test-time thinking as a lens for improving generative AI agents through finer-grained reward modeling, data-centric reasoning, and robust alignment. Beginning with GenARM, we introduce an inductive bias for denser, token-level reward modeling that guides generation during decoding, enabling token-level alignment without retraining. While GenARM targets reward design, ThinkLite-VL focuses on the data side of reasoning. It proposes a self-improvement framework that selects the most informative samples via MCTS-guided search, yielding stronger visual reasoning with fewer labels. Taking this a step further, MORSE-500 moves beyond selection to creation: it programmatically generates targeted, controllable multimodal data to systematically probe and stress-test models’ reasoning abilities. We then interrogate a central assumption in inference-time alignment: Does Thinking More Always Help? Our findings reveal that increased reasoning steps can degrade performance–not due to better or worse reasoning per se, but due to rising variance in outputs, challenging the naive scaling paradigm. Finally, AegisLLM applies test-time thinking in the service of security, using an agentic, multi-perspective framework to defend against jailbreaks, prompt injections, and unlearning attacks–all at inference time. Together, these works chart a path toward generative agents that are not only more capable, but more data-efficient, introspective, and robust in real-world deployment.

   

  Talk Date: Friday, October 24, 2025

• Volodymyr Kuleshov Cornell Tech and Cornell University

  Title: Discrete Diffusion Language Models

   

  Abstract: While diffusion generative models excel at high-quality image generation, prior work reports a significant performance gap between diffusion and autoregressive (AR) methods on discrete data such as text or biological sequences. Our work takes steps towards closing this gap via a simple and effective framework for discrete diffusion. This framework is simple to understand—it optimizes a mixture of denoising (e.g., masking) losses—and can be seen as endowing BERT-like models with principled samplers and variational estimators of log-likelihood. Crucially, our algorithms are not constrained to generate data sequentially, and therefore have the potential to improve long-term planning, controllable generation, and sampling speed.

   

  In the context of language modeling, our framework enables deriving masked diffusion language models (MDLMs), which achieve a new state-of-the-art among diffusion models, and approach AR quality. Combined with novel extensions of classifier-free and classifier-based guidance mechanisms, these algorithms are also significantly more controllable than AR models. Discrete diffusion extends beyond language to science, where it forms the basis of a new generation of DNA foundation models. Our largest models focus on plants and set a new state of the art in genome annotation, while also enabling effective generation. Discrete diffusion models hold the promise to advance progress in generative modeling and its applications in language understanding and scientific discovery.

   

  Talk Date: Friday, October 20, 2025

• Jingfeng Wu University of California Berkeley

  Title: Gradient Descent Dominates Ridge: A Statistical View on Implicit Regularization

   

  Abstract: A key puzzle in deep learning is how simple gradient methods find generalizable solutions without explicit regularization. This talk discusses the implicit regularization of gradient descent (GD) through the lens of statistical dominance. Using least squares as a clean proxy, we present two surprising findings.

   

  First, GD dominates ridge regression: with comparable regularization, the excess risk of GD is always within a constant factor of ridge, but ridge can be polynomially worse even when tuned optimally. Second, GD is incomparable with SGD. While it is known that for certain problems GD can be polynomially better than SGD, the reverse is also true: we construct problems, inspired by benign overfitting theory, where optimally stopped GD is polynomially worse. Finally, GD dominates SGD for a significant subclass of problems — those with fast and continuously decaying covariance spectra — which includes all problems satisfying the standard capacity condition.

  This is joint work with Peter Bartlett, Sham Kakade, Jason Lee, and Bin Yu.

   

  Talk Date: Friday, October 6, 2025