自然语言处理解释中的忠实性与对抗敏感性概念
Faithfulness and the Notion of Adversarial Sensitivity in NLP Explanations
AI总结：
针对自然语言处理解释忠实性评估方法存在差异和偏差的问题，提出对抗敏感性新方法，通过捕捉解释器对模型对抗攻击下输入变化的响应评估忠实性，解决现有技术局限并从新范式量化该指标。
摘要：
忠实性可以说是评估可解释人工智能可靠性最关键的指标。在自然语言处理领域，当前的忠实性评估方法存在大量差异和偏差，往往无法捕捉模型的真实推理过程。我们提出对抗敏感性作为一种新的忠实性评估方法，聚焦于模型遭受对抗攻击时解释器的响应。我们的方法通过捕捉解释器对对抗性输入变化的敏感性来衡量其忠实性。这项工作解决了现有评估技术的重大局限性，并且从一个关键但未被充分探索的范式量化了忠实性。
英文摘要：
Faithfulness is arguably the most critical metric to assess the reliability of explainable AI. In NLP, current methods for faithfulness evaluation are fraught with discrepancies and biases, often failing to capture the true reasoning of models. We introduce Adversarial Sensitivity as a novel approach to faithfulness evaluation, focusing on the explainer's response when the model is under adversarial attack. Our method accounts for the faithfulness of explainers by capturing sensitivity to adversarial input changes. This work addresses significant limitations in existing evaluation techniques, and furthermore, quantifies faithfulness from a crucial yet underexplored paradigm.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
