Tag: Adversarial Attack

All the papers with the tag "Adversarial Attack".

• Adversarial Cooperative Rationalization: The Risk of Spurious Correlations in Even Clean Datasets

  grok-3-latest

  Score: 0.48

  Published:2025年5月4日 at  at 14:00

  #LLM, #Rationalization, #Spurious Correlations, #Adversarial Attack, #Interpretability

  本文提出 A2I 方法，通过对抗性攻击检测并纠正自解释合理化框架中模型引入的虚假相关性，显著提升 Rationale 质量。