大模型训练或无需“纯净数据”！北大团队新研究：随机噪声影响有限，新方法让模型更抗噪
"模型鲁棒性" 相关文章
亚军团队！全球AI攻防挑战赛系列分享（二）
Understanding Adversarial Attacks Using Fast Gradient Sign Method
Enhancing NLP Models for Robustness Against Adversarial Attacks: Techniques and Applications
Circuit Breakers for AI: Interrupting Harmful Outputs Through Representation Engineering
The Hidden Danger in AI Models: A Space Character’s Impact on Safety
