Researchers introduced the Unforeseen Attack Robustness (UAR) metric to evaluate how reliably neural network classifiers defend against adversarial attacks unseen during training. The method highlights the necessity of measuring performance across a diverse range of unforeseen attacks, addressing limitations in current single-benchmark evaluations.
Termsslug: adversarial-attack
Adversarial Attack
对抗攻击
Techniques that use small perturbations to mislead ML models into making incorrect predictions.
Curated Intelligence
OpenAI 官方博客T1·Models
45 pts
OpenAI 官方博客T1·Industry
42 pts
A collaborative paper by FHI, CSER, and other institutions forecasts how malicious actors might misuse AI technology and outlines potential prevention and mitigation strategies. The outcome of nearly a year of interdisciplinary work, it focuses on threat modeling and defense in AI safety.
OpenAI 官方博客T1·Models
25 pts
The entry contains only the title 'Adversarial attacks on neural network policies' with no body text. Due to the lack of specific methods, experimental data, or conclusions, its technical details and actual impact cannot be assessed.