#52 - Unadversarial Examples (Hadi Salman, MIT)

Research #AI Safety and Robustness 📝 Blog|Analyzed: Jan 3, 2026 07:16•

Published: May 1, 2021 01:02

•

1 min read

Analysis

This article discusses Hadi Salman's research on unadversarial examples, which are designed to improve the robustness of neural networks. It highlights the potential of leveraging the brittleness of neural networks to create models that are more resistant to adversarial attacks and generalize better. The article also touches upon related concepts like transferability and the role of robust features.

Key Takeaways

•Unadversarial examples are designed to improve the robustness of neural networks.
•The research leverages the brittleness of neural networks.
•Robust models are expected to transfer better.

Reference / Citation

View Original

"Hadi actually utilized the brittleness of neural networks to design unadversarial examples or robust objects which_ are objects designed specifically to be robustly recognized by neural networks."

ML Street Talk PodMay 1, 2021 01:02

* Cited for critical analysis under Article 32.

Older

Self-Supervised Vision Models at FAIR

Newer

#51 Francois Chollet - Intelligence and Generalisation