操控多模态AI幻觉的可验证性
Steering the Verifiability of Multimodal AI Hallucinations
AI总结：
本文研究多模态AI幻觉的可验证性差异，提出基于激活空间的干预方法，通过区分明显和隐秘幻觉，实现对模型可验证性的精细控制。
摘要：
由多模态大语言模型（MLLMs）驱动的AI应用易产生幻觉，对用户构成重大风险。关键在于，幻觉并非等同问题：一些幻觉内容可被人类检测（即明显幻觉），而另一些则常被忽视或需更多验证（即隐秘幻觉）。这表明多模态AI幻觉在可验证性上有显著差异。然而，很少有研究探讨如何控制这一属性以满足不同安全性和可用性需求的AI应用。为填补这一空白，我们从4,470人对AI生成幻觉的回应中构建了数据集，并根据人类用户可验证性将这些幻觉分为明显和隐秘类型。进一步，我们提出了一种激活空间干预方法，分别学习明显和隐秘幻觉的探测器。我们发现明显和隐秘幻觉会引发不同的干预探测器，从而实现对模型可验证性的精细控制。实证结果证明了该方法的有效性，并显示针对性干预在调节相应可验证性方面表现更优。此外，简单混合这些干预可实现不同场景所需的灵活可验证性控制。
英文摘要：
AI applications driven by multimodal large language models (MLLMs) are prone to hallucinations and pose considerable risks to human users. Crucially, such hallucinations are not equally problematic: some hallucination contents could be detected by human users(i.e., obvious hallucinations), while others are often missed or require more verification effort(i.e., elusive hallucinations). This indicates that multimodal AI hallucinations vary significantly in their verifiability. Yet, little research has explored how to control this property for AI applications with diverse security and usability demands. To address this gap, we construct a dataset from 4,470 human responses to AI-generated hallucinations and categorize these hallucinations into obvious and elusive types based on their verifiability by human users. Further, we propose an activation-space intervention method that learns separate probes for obvious and elusive hallucinations. We reveal that obvious and elusive hallucinations elicit different intervention probes, allowing for fine-grained control over the model's verifiability. Empirical results demonstrate the efficacy of this approach and show that targeted interventions yield superior performance in regulating corresponding verifiability. Moreover, simply mixing these interventions enables flexible control over the verifiability required for different scenarios.
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。
