从中间层提取视觉事实以缓解多模态大语言模型中的幻觉
Extracting Visual Facts from Intermediate Layers for Mitigating Hallucinations in Multimodal Large Language Models
AI总结：
本文提出无需训练的 EVA 方法，通过动态选择中间层并对比原始输入与纯文本输入的输出分布来提取视觉事实知识，从而校正多模态大语言模型的输出 logits，显著降低对象幻觉率。
摘要：
多模态大语言模型（MLLMs）通过结合视觉识别和语言理解，在生成连贯且上下文准确的内容方面取得了显著进展。然而，MLLMs 仍然面临对象幻觉问题，即模型会产生看似合理但事实错误的输出，包括图像中并不存在的对象。最近的研究表明，MLLMs 中的先验知识会显著抑制深层中的视觉信息，从而导致幻觉输出。然而，这些先验知识在 MLLMs 的中间层阶段如何抑制视觉信息仍不清楚。我们观察到，视觉事实知识与中间层先验/原始概率分布之间的差异在中间层中表现出相似的演化趋势。受此启发，我们提出了通过提取视觉事实进行解码（EVA），这是一种简单且无需训练的方法，能够动态选择具有最显著视觉事实信息的中间层。通过对比所选层在原始输入和纯文本输入下得到的输出分布，EVA 提取视觉事实知识，并将其按比例融入最终层以校正输出 logits。重要的是，EVA 具有模型无关性，能够无缝集成各种经典解码策略，并适用于不同的 MLLMs。我们在广泛使用的基准上验证了 EVA，结果表明与基线方法相比，它显著降低了幻觉率，凸显了其在缓解幻觉方面的有效性。
英文摘要：
Multimodal Large Language Models (MLLMs) have made significant strides by combining visual recognition and language understanding to generate content that is both coherent and contextually accurate. However, MLLMs continue to struggle with object hallucinations, where models produce seemingly plausible but factually incorrect outputs, including objects that do not exist in the image. Recent work has revealed that the prior knowledge in MLLMs significantly suppresses visual information in deep layers, causing hallucinatory outputs. However, how these priors suppress visual information at the intermediate layer stage in MLLMs remains unclear. We observe that visual factual knowledge and the differences between intermediate-layer prior/original probability distributions show similar evolutionary trends in intermediate layers. Motivated by this, we introduce Decoding by Extracting Visual Facts (EVA), a simple, training-free method that dynamically selects intermediate layers with the most significant visual factual information. By contrasting the output distributions of the selected layer derived from the original input and pure-text input, EVA extracts visual factual knowledge and proportionally incorporates it into the final layer to correct the output logits. Importantly, EVA is model-agnostic, seamlessly integrates with various classic decoding strategies, and is applicable across different MLLMs. We validate EVA on widely-used benchmarks, and the results show that it significantly reduces hallucination rates compared to baseline methods, underscoring its effectiveness in mitigating hallucinations.
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。
