让自回归图像视频生成更稳定，通过对比真实与损坏前缀来优化下一步预测。
VPG: Visual Prefix Guidance for Autoregressive Image and Video Generation
- 推理时对比生成前缀与损坏前缀，引导模型选择更支持当前生成结果的下一步。
- 在VAR、Infinity和InfinityStar上均提升质量，平均降低FID 0.36。
- 无需重训练，适用于图像与视频生成，适合追求生成稳定性的研究者。
自回归图像和视频生成器在训练时使用教师强制策略，但在推理时需基于自身生成的前缀进行采样，易受暴露偏差和前缀漂移影响。现有方法或修改训练过程，或在采样阶段引入外部语义条件（如类别标签或文本提示）的引导，但未检验下一步预测是否能为已生成前缀提供强后验支持。本文提出视觉前缀引导（VPG），一种无需训练的推理阶段引导方法。VPG通过对比模型在生成前缀与损坏前缀下的输出，将逻辑值外推至能增强生成前缀后验支持的候选方案，从而改善下一步预测。在类条件图像生成（VAR）、文本到图像生成（Infinity）及文本到视频生成（InfinityStar）任务中，VPG不需重训练即可提升生成质量，平均降低FID 0.36，显著提升图像与视频生成基准性能。
原文摘要 · Abstract (English)
Autoregressive image and video generators are trained with teacher-forced histories but must sample from their own generated prefixes at inference time, making them vulnerable to exposure bias and prefix drift. Existing remedies either modify training or apply sampling-time guidance aimed primarily at external semantic conditions, such as class labels or text prompts, rather than testing whether a next-step prediction provides strong posterior support for the generated prefix itself. We propose Visual Prefix Guidance (VPG), a training-free inference-time guidance method for autoregressive image and video generation. VPG improves next-step prediction by contrasting the model's output under the generated prefix with its output under a corrupted prefix, then extrapolating logits toward candidates that strengthen the posterior support of the generated prefix. Across class-conditional image generation with VAR, text-to-image generation with Infinity, and text-to-video generation with InfinityStar, VPG improves generation quality without retraining the base model, reducing FID on VAR by 0.36 on average and improving benchmark performance on both image and video generation.
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。