Statistical Physics Meets Machine Learning I
ORAL · MAR-L63 · ID: MAR-L63
Presentations
-
Evolution of Language Statistics under Renormalization Group Flow
Oral-In-person
–
Presenters
-
Roberto Avalos
- Emory University
Authors
-
Roberto Avalos
- Emory University
-
Tankut Can
-
-
Compression in Neural Networks via Weight Coupling
Oral-In-person
–
Publication: Quantization and the Bottom of the Loss Landscape (ICML workshop paper); Neural Network Quantization via Weight-Weight Coupling (planned paper)
Presenters
-
Daniel Bernstein
- Princeton University
Authors
-
Daniel Bernstein
- Princeton University
-
Luca Di Carlo
-
David Schwab
- The Graduate Center, City University of New York
-
-
Building effective shallow generative models through side information
Oral-In-person
–
Presenters
-
Kyle Bojanek
- University of Chicago
Authors
-
Kyle Bojanek
- University of Chicago
-
Stephanie Palmer
- University of Chicago
-
-
Implementation of Kolmogorov Flow Matching in a GFlowNet Learning Model
Oral-In-person
–
Presenters
-
Sergio Cuadra
- University of Massachusetts Boston
Authors
-
Sergio Cuadra
- University of Massachusetts Boston
-
Sho Inaba
- University of Massachusetts Boston
-
Jacob Adamczyk
- University of Massachusetts Boston
-
Rahul Kulkarni
- University of Massachusetts Boston
-
-
Combinatorial Reasoning: Using physics-inspired methods to improve reasoning on generative language models
Oral-In-person
–
Publication: [1] M. Esencan et al., "Combinatorial Reasoning: Selecting Reasons in Generative AI Pipelines via Combinatorial Optimization," arXiv:2407.00071 (2024)
Presenters
-
Mert Esencan
- University of Oxford
Authors
-
Mert Esencan
- University of Oxford
-
Davide Venturelli
- USRA and NASA Ames Research Center
-
Can Unlu
- Icosa Computing
-
Tarun Advaith Kumar
-
Alan Ho
-
-
The geometry and dynamics of annealed optimization in the coherent Ising machine with hidden and planted solutions
Oral-In-person
–
Presenters
-
Federico Ghimenti
- Stanfod University
Authors
-
Federico Ghimenti
- Stanfod University
-
Adithya Sriram
-
Atsushi Yamamura
- Stanford University
-
Hideo Mabuchi
-
Surya Ganguli
- Stanford University
-
-
Entropic Confinement and Mode Connectivity in Overparameterized Neural Networks
Oral-In-person
–
Publication: Submitted to ICLR 2026
Presenters
-
Chase Goddard
- Princeton University
Authors
-
Chase Goddard
- Princeton University
-
Luca Di Carlo
- Princeton University
-
David Schwab
- The Graduate Center, City University of New York
-
-
Tokenization gauge symmetry in language modeling
Oral-In-person
–
Presenters
-
Kanishk Jain
- Emory University
Authors
-
Kanishk Jain
- Emory University
-
Matthew Day
-
Tankut Can
-
-
EVE: EigenVector-Based Exploration
Oral-In-person
–
Publication: EVE: EigenVector-Based Exploration (paper)
Presenters
-
Adam Kamoski
- University of Massachusetts Boston
Authors
-
Adam Kamoski
- University of Massachusetts Boston
-
Rahul Kulkarni
- University of Massachusetts Boston
-
Sho Inaba
- University of Massachusetts Boston
-
Jacob Adamczyk
- University of Massachusetts Boston
-
-
Scattering-based Optical Computing
Oral-In-person
–
Presenters
-
Wolfgang Losert
- University of Maryland College Park
Authors
-
Eunji Ko
- University of Maryland College Park
-
Wolfgang Losert
- University of Maryland College Park
-
-
Network architecture can shape in-context learning
Oral-In-person · Withdrawn
–
Presenters
-
Pranav Kantroo
- Yale University
Authors
-
Pranav Kantroo
- Yale University
-
Benjamin Machta
- Yale University
-
-
Local Diffusion Models and Phases of Data Distributions
Oral-In-person
–
Publication: Preprint: arXiv:2508.06614
Presenters
-
Guangkuo Liu
- JILA
Authors
-
Guangkuo Liu
- JILA
-
Fangjun Hu
- QuEra Computing Inc.
-
Yifan (Frank) Zhang
- Princeton University
-
Xun Gao
-
-
Large Language Models Think in Curved Space
Oral-In-person
–
Presenters
-
Toni Liu
- Cornell University
Authors
-
Toni Liu
- Cornell University
-
Raphael Sarfati
- University of Colorado, Boulder
-
Nicolas Boulle
-
Christopher Earls
-
-
Data coarse graining can improve model performance
Oral-In-person
–
Publication: A. Nguyen, D. J. Schwab, and V. Ngampruetikorn, arXiv:2509.14498 [cond-mat.stat-mech]
Presenters
-
Vudtiwat Ngampruetikorn
- The University of Sydney
Authors
-
Alex Nguyen
-
David Schwab
- The Graduate Center, City University of New York
-
Vudtiwat Ngampruetikorn
- The University of Sydney
-
-
Variational Inference with Heavy-Tailed Distributions using the Coupled Free Energy
Oral-In-person
–
Publication: Nelson, K., Oliveira, I., Al-Najafi, A., Zhang, F., & Ng, H. K. T. (2025, July 25). Variational inference optimized using the curved geometry of coupled free energy (arXiv:2506.09091). arXiv. https://doi.org/10.48550/arXiv.2506.09091
Nelson, K. P. (2025, August 9). Coupled entropy: A Goldilocks generalization for complex systems (arXiv:2506.17229). arXiv. https://arxiv.org/abs/2506.17229Presenters
-
Igor Oliveira
- Universidade Federal de Pernambuco - UFPE
Authors
-
Igor Oliveira
- Universidade Federal de Pernambuco - UFPE
-
Kenric Nelson
-
Amenah Najaf
-