CNN-based generative modelling has evolved to produce synthetic images indistinguishable from real images in the RGB pixel space. Recent works have observed that CNN-generated images share a systematic shortcoming in replicating high frequency Fourier spectrum decay attributes. Furthermore, these works have successfully exploited this systematic shortcoming to detect CNN-generated images reporting up to 99% accuracy across multiple state-of-the-art GAN models. In this work, we investigate the validity of assertions claiming that CNN-generated images are unable to achieve high frequency spectral decay consistency. We meticulously construct a counterexample space of high frequency spectral decay consistent CNN-generated images emerging from our handcrafted experiments using DCGAN, LSGAN, WGAN-GP and StarGAN, where we empirically show that this frequency discrepancy can be avoided by a minor architecture change in the last upsampling operation. We subsequently use images from this counterexample space to successfully bypass the recently proposed forensics detector which leverages on high frequency Fourier spectrum decay attributes for CNN-generated image detection. Through this study, we show that high frequency Fourier spectrum decay discrepancies are not inherent characteristics for existing CNN-based generative models--contrary to the belief of some existing work--, and such features are not robust to perform synthetic image detection. Our results prompt re-thinking of using high frequency Fourier spectrum decay attributes for CNN-generated image detection. Code and models are available at https://keshik6.github.io/Fourier-Discrepancies-CNN-Detection/

Keshigeyan Chandrasegaran Ngoc-Trung Tran Ngai-Man Cheung

We develop a theory of quasiparticle localization in superconductors in situations without spin rotation invariance. We discuss the existence, and properties of superconducting phases with localized/delocalized quasiparticle excitations in such systems in various dimensionalities. Implications for a variety of experimental systems, and to the properties of random Ising models in two dimensions, are briefly discussed.

T. Senthil Matthew P. A. Fisher

Recent results on the nature of the quantum critical point between Neel and valence bond solid(VBS) ordered phases of two dimensional quantum magnets are examined by an attack from the VBS side. This approach leads to an appealingly simple physical description, and further insight into the properties of the transition.

Michael Levin T. Senthil

A simple physical realization of an integer quantum Hall state of interacting two dimensional bosons is provided. This is an example of a "symmetry-protected topological" (SPT) phase which is a generalization of the concept of topological insulators to systems of interacting bosons or fermions. Universal physical properties of the boson integer quantum Hall state are described and shown to correspond to those expected from general classifications of SPT phases.

T. Senthil Michael Levin

Using the Fermi LAT data on the gamma ray emission from dwarf spheroidal galaxies, we get the upper bound on the probability of gamma rays from dark matter decay for the validity of explanation of the anomalous Kolar events as dark matter decay.

R. Thiru Senthil G. Rajasekaran

Designing model architectures requires decisions such as selecting operators (e.g., attention, convolution) and configurations (e.g., depth, width). However, evaluating the impact of these decisions on model quality requires costly pretraining, limiting architectural investigation. Inspired by how new software is built on existing code, we ask: can new architecture designs be studied using pretrained models? To this end, we present grafting, a simple approach for editing pretrained diffusion transformers (DiTs) to materialize new architectures under small compute budgets. Informed by our analysis of activation behavior and attention locality, we construct a testbed based on the DiT-XL/2 design to study the impact of grafting on model quality. Using this testbed, we develop a family of hybrid designs via grafting: replacing softmax attention with gated convolution, local attention, and linear attention, and replacing MLPs with variable expansion ratio and convolutional variants. Notably, many hybrid designs achieve good quality (FID: 2.38-2.64 vs. 2.27 for DiT-XL/2) using <2% pretraining compute. We then graft a text-to-image model (PixArt-Sigma), achieving a 1.43x speedup with less than a 2% drop in GenEval score. Finally, we present a case study that restructures DiT-XL/2 by converting every pair of sequential transformer blocks into parallel blocks via grafting. This reduces model depth by 2x and yields better quality (FID: 2.77) than other models of comparable depth. Together, we show that new diffusion model designs can be explored by grafting pretrained DiTs, with edits ranging from operator replacement to architecture restructuring. Code and grafted models: https://grafting.stanford.edu

Keshigeyan Chandrasegaran Michael Poli Daniel Y. Fu Dongjun Kim Lea M. Hadzic Manling Li Agrim Gupta Stefano Massaroli Azalia Mirhoseini Juan Carlos Niebles Stefano Ermon Li Fei-Fei

Model Inversion (MI) attacks aim to reconstruct private training data by abusing access to machine learning models. Contemporary MI attacks have achieved impressive attack performance, posing serious threats to privacy. Meanwhile, all existing MI defense methods rely on regularization that is in direct conflict with the training objective, resulting in noticeable degradation in model utility. In this work, we take a different perspective, and propose a novel and simple Transfer Learning-based Defense against Model Inversion (TL-DMI) to render MI-robust models. Particularly, by leveraging TL, we limit the number of layers encoding sensitive information from private training dataset, thereby degrading the performance of MI attack. We conduct an analysis using Fisher Information to justify our method. Our defense is remarkably simple to implement. Without bells and whistles, we show in extensive experiments that TL-DMI achieves state-of-the-art (SOTA) MI robustness. Our code, pre-trained models, demo and inverted data are available at: https://hosytuyen.github.io/projects/TL-DMI

Sy-Tuyen Ho Koh Jun Hao Keshigeyan Chandrasegaran Ngoc-Bao Nguyen Ngai-Man Cheung

We present a theory of the effects of impurity scattering in d_{x^2-y^2} superconductors and their quantum disordered counterparts, based on a non-linear sigma model formulation. We show the existence, in a quasi-two-dimensional system, of a novel spin-metal phase with a non-zero spin diffusion constant at zero temperature. With decreasing inter-layer coupling, the system undergoes a quantum phase transition (in a new universality class) to a localized spin-insulator. Experimental implications for spin and thermal transport in the high-temperature superconductors are discussed.

T. Senthil Matthew P. A. Fisher Leon Balents Chetan Nayak

The precise theoretical characterization of a fractionalized phase in spatial dimensions higher than one is through the concept of ``topological order''. We describe a physical effect that is a robust and direct consequence of this hidden order that should enable a precise experimental characterization of fractionalized phases. In particular, we propose specific ``smoking-gun'' experiments to unambiguously settle the issue of electron fractionalization in the underdoped cuprates.

T. Senthil Matthew P. A. Fisher

Recently, we have elucidated the physics of electron fractionalization in strongly interacting electron systems using a $Z_2$ gauge theory formulation. Here we discuss the connection with the earlier U(1) gauge theory approaches based on the slave boson mean field theory. In particular, we identify the relationship between the holons and Spinons of the slave-boson theory and the true physical excitations of the fractionalized phases that are readily described in the $Z_2$ approach.

T. Senthil Matthew P. A. Fisher