Research map / Adversarial robustness and security / Certified and empirical robustness

Exploring The Forgetting in Adversarial Training: A Novel Method for Enhancing Robustness

ICLR 2025 · Adversarial robustness and security › Certified and empirical robustness · approach: Adversarial training refinements

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes a new adversarial training method that addresses catastrophic forgetting in AT to improve clean and robust accuracy.

Problem

Adversarial training still leaves a gap in achieving strong clean and robust accuracy on many practical tasks; the paper views AT through the lens of catastrophic forgetting.

Approach

Connects adversarial training to catastrophic forgetting and uses this perspective to design a new AT method for enhancing robustness (details not available in the abstract provided).

Experimental setup

Metrics: clean accuracy robust accuracy

Builds on: adversarial training

Closest papers by meaning

• The Effectiveness of Random Forgetting for Robust Generalization (ICLR 2024 · similarity 0.92)
• Robustness Under Data Scarcity: Few-Shot Continual Adversarial Training for Evolving Threats (CVPR 2026 · similarity 0.91)
• A comprehensive, application-oriented study of catastrophic forgetting in DNNs (ICLR 2019 · similarity 0.90)
• Towards Understanding Catastrophic Forgetting in Two-layer Convolutional Neural Networks (ICML 2025 · similarity 0.89)
• Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.89)
• Unlocking the Power of Function Vectors for Characterizing and Mitigating Catastrophic Forgetting in Continual Instruction Tuning (ICLR 2025 · similarity 0.88)