描述-再生成瓶颈：VLM描述如何改变图像生成结果
The Describe-Then-Generate Bottleneck: How VLM Descriptions Alter Image Generation Outcomes
AI总结：
本文通过150对图像实验量化了视觉-语言-视觉管线中描述-再生成瓶颈导致的信息损失，发现99.3%样本存在显著感知退化，证明该瓶颈是多模态系统的可测量一致性局限。
摘要：
随着多模态AI系统在创意工作流中的日益融合，理解视觉-语言-视觉管线中的信息损失对于评估系统局限性已变得重要。然而，视觉内容经过文本中介时发生的退化仍缺乏定量刻画。在本工作中，我们对描述-再生成瓶颈进行了实证分析，其中自然语言充当视觉信息的中间表示。我们通过描述-再生成管线生成了150对图像，并应用现有指标（LPIPS、SSIM和颜色距离）来测量感知、结构和色彩维度上的信息保持情况。我们的评估显示，99.3%的样本表现出显著的感知退化，91.5%的样本表现出显著的结构信息损失，从而提供了实证证据，表明描述-再生成瓶颈代表了当代多模态系统中一种可测量且一致性的局限性。
英文摘要：
With the increasing integration of multimodal AI systems in creative workflows, understanding information loss in vision-language-vision pipelines has become important for evaluating system limitations. However, the degradation that occurs when visual content passes through textual intermediation remains poorly quantified. In this work, we provide empirical analysis of the describe-then-generate bottleneck, where natural language serves as an intermediate representation for visual information. We generated 150 image pairs through the describe-then-generate pipeline and applied existing metrics (LPIPS, SSIM, and color distance) to measure information preservation across perceptual, structural, and chromatic dimensions. Our evaluation reveals that 99.3% of samples exhibit substantial perceptual degradation and 91.5% demonstrate significant structural information loss, providing empirical evidence that the describe-then-generate bottleneck represents a measurable and consistent limitation in contemporary multimodal systems.
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。
