How Can We Harness Pre-Training to Develop Robust Models?We explore a simple principle for harnessing pre-training to develop robust models.
Ask Your Distribution Shift if Pre-Training is Right for YouWe study the robustness benefits of pre-training and characterize failure modes that pre-training can and cannot address.
DsDm: Model-Aware Dataset Selection with DatamodelsSelecting better data by approximating how models learn from data.
How Training Data Guides Diffusion ModelsWe introduce a new framework for data attribution in generative settings, and propose an efficient method to attribute diffusion models.
Rethinking Backdoor AttacksWe introduce a new perspective on backdoor attacks and defenses in deep learning.
TRAK-ing Model Behavior with DataWe introduce TRAK, a new data attribution method that scales to large(r) models!
Dataset Interfaces: Diagnosing Model Failures Using Controllable Counterfactual GenerationWe introduce dataset interfaces, a scalable framework that synthesizes counterfactual examples under user-specified shifts
Tailored Data Augmentation to Mitigate Model FailuresWe demonstrate how we can use Stable Diffusion to target a model's failure modes