通过分析模型生成前文，实时检测大模型图像幻觉。
PAS : Prelim Attention Score for Detecting Object Hallucinations in Large Vision--Language Models
- 利用生成前文与图像的互信息，发现幻觉时模型忽略图像。
- 提出PAS分数，在推理时无需额外计算即可检测幻觉。
- 轻量无训练，适合部署在多模型、实时系统中。
大型视觉语言模型（LVLM）虽强大，但常出现物体幻觉。本文发现，许多幻觉预测中，模型实际忽略了图像，转而依赖先前生成的输出（prelim）token推断新物体。我们通过计算图像与预测物体在prelim条件下的互信息，量化了这种弱图像依赖性，并证明其与幻觉强相关。基于此，提出轻量级、无需训练的预生成注意力得分（PAS），仅需从prelim token的注意力权重中计算，不需额外前向传播，可在推理过程中实时生成。利用这一被忽视的信号，PAS在多个模型和数据集上实现当前最佳幻觉检测效果，支持实时过滤与干预。
原文摘要 · Abstract (English)
Large vision-language models (LVLMs) are powerful, yet they remain unreliable due to object hallucinations. In this work, we show that in many hallucinatory predictions the LVLM effectively ignores the image and instead relies on previously generated output (prelim) tokens to infer new objects. We quantify this behavior via the mutual information between the image and the predicted object conditioned on the prelim, demonstrating that weak image dependence strongly correlates with hallucination. Building on this finding, we introduce the Prelim Attention Score (PAS), a lightweight, training-free signal computed from attention weights over prelim tokens. PAS requires no additional forward passes and can be computed on the fly during inference. Exploiting this previously overlooked signal, PAS achieves state-of-the-art object-hallucination detection across multiple models and datasets, enabling real-time filtering and intervention.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。