Robustness May Be at Odds with Accuracy
1 version
LE
Logan Engstrom
cs.LGstat.MLcs.CVcs.NEcs.CRcs.ROcs.AIcs.CL
On Valency
published · living versionsBlack-box Adversarial Attacks with Limited Queries and Information
Exploring the Landscape of Spatial Robustness
Preprints & journals
29 papers in the corpus · 2017–2025OpenAI GPT-5 System Card
26 citations
Small-to-Large Generalization: Data Influences Models Consistently Across Scale
0 citations
DataMIL: Selecting Data for Robot Imitation Learning with Datamodels
0 citations
MAGIC: Near-Optimal Data Attribution for Deep Learning
0 citations
Optimizing ML Training with Metagradient Descent
0 citations
DsDm: Model-Aware Dataset Selection with Datamodels
2 citations
FFCV: Accelerating Training by Removing Data Bottlenecks
36 citations
Dataset Interfaces: Diagnosing Model Failures Using Controllable Counterfactual Generation
10 citations
When does Bias Transfer in Transfer Learning?
15 citations
Datamodels: Predicting Predictions from Training Data
20 citations
3DB: A Framework for Debugging Computer Vision Models
16 citations
Unadversarial Examples: Designing Objects for Robust Vision
27 citations
Do Adversarially Robust ImageNet Models Transfer Better?
147 citations
Noise or Signal: The Role of Image Backgrounds in Object Recognition
161 citations
Implementation Matters in Deep Policy Gradients: A Case Study on PPO and TRPO
146 citations
From ImageNet to Image Classification: Contextualizing Progress on Benchmarks
75 citations
Identifying Statistical Bias in Dataset Replication
14 citations
Image Synthesis with a Single (Robust) Classifier
92 citations
Adversarial Robustness as a Prior for Learned Representations
107 citations
Adversarial Examples Are Not Bugs, They Are Features
940 citations
A Closer Look at Deep Policy Gradients
38 citations
Evaluating and Understanding the Robustness of Adversarial Logit Pairing
83 citations
Prior Convictions: Black-Box Adversarial Attacks with Bandits and Priors
261 citations
Robustness May Be at Odds with Accuracy
1,069 citations
Black-box Adversarial Attacks with Limited Queries and Information
806 citations
DNase-capture reveals differential transcription factor binding modalities.
1 citation
Query-Efficient Black-box Adversarial Examples (superceded)
41 citations
Exploring the Landscape of Spatial Robustness
300 citations
Synthesizing Robust Adversarial Examples
973 citations
Profile built from the corpus for this byline.
Author records are still filling in while the Hub is in beta. If this is your page, you'll be able to claim it soon. Spot a mistake? Tell us.