Get our free extension to see links to code for papers anywhere online!

Chrome logo  Add to Chrome

Firefox logo Add to Firefox

Diffusion Art or Digital Forgery? Investigating Data Replication in Diffusion Models


Dec 12, 2022
Gowthami Somepalli, Vasu Singla, Micah Goldblum, Jonas Geiping, Tom Goldstein

Add code

* Updated draft with the following changes (1) Clarified the LAION Aesthetics versions everywhere (2) Correction on which LAION Aesthetics version SD - 1.4 is finetuned on and updated figure 12 based on this (3) A section on possible causes of replication 

   Access Paper or Ask Questions

How Much Data Are Augmentations Worth? An Investigation into Scaling Laws, Invariance, and Implicit Regularization


Oct 12, 2022
Jonas Geiping, Micah Goldblum, Gowthami Somepalli, Ravid Shwartz-Ziv, Tom Goldstein, Andrew Gordon Wilson

Add code

* 27 pages, 20 figures. Code at https://github.com/JonasGeiping/dataaugs 

   Access Paper or Ask Questions

Can Neural Nets Learn the Same Model Twice? Investigating Reproducibility and Double Descent from the Decision Boundary Perspective


Mar 15, 2022
Gowthami Somepalli, Liam Fowl, Arpit Bansal, Ping Yeh-Chiang, Yehuda Dar, Richard Baraniuk, Micah Goldblum, Tom Goldstein

Add code

* To appear in CVPR 2022 

   Access Paper or Ask Questions

PatchGame: Learning to Signal Mid-level Patches in Referential Games


Nov 02, 2021
Kamal Gupta, Gowthami Somepalli, Anubhav Gupta, Vinoj Jayasundara, Matthias Zwicker, Abhinav Shrivastava

Add code

* To appear at NeurIPS 2021 

   Access Paper or Ask Questions

SAINT: Improved Neural Networks for Tabular Data via Row Attention and Contrastive Pre-Training


Jun 02, 2021
Gowthami Somepalli, Micah Goldblum, Avi Schwarzschild, C. Bayan Bruss, Tom Goldstein

Add code


   Access Paper or Ask Questions

What Doesn't Kill You Makes You Robust(er): Adversarial Training against Poisons and Backdoors


Feb 26, 2021
Jonas Geiping, Liam Fowl, Gowthami Somepalli, Micah Goldblum, Michael Moeller, Tom Goldstein

Add code

* 17 pages, 14 figures 

   Access Paper or Ask Questions