Figureure 1 · Overall pipeline of the methodology, which consist of two parts: (a) SFigure 1. Overall pipeline of the methodology, which consist of two parts: (a) Synthetic Data Generation using Representation Conditioned Diffusion Model (RCDM) and (b) Classifier Based Evaluation for Comparative analysis.这张图概括 Representation-Conditioned Diffusion Models for Guided 的整体方法流程。阅读时先看模块之间传递的训练信号，再看作者如何把目标拆成可优化的子问题。Figureure 2 · Scaling synthetic datasets for classifier trainingFigure 2. Scaling synthetic datasets for classifier training. Top-1 (left) and Top-5 (right) accuracy of a ResNet-50 trained on synthetic ImageNet-100 datasets generated using representation-conditioned diffusion models with DINOv2, DINOv3, and CLIP representations, compared to a class-conditioned LDM baseline and ResNet-50 trained on real ImageNet100 baseline.这张图/表用于判断 Representation-Conditioned Diffusion Models for Guided 的实验收益来自哪里。重点看替换、消融或跨模型设置下趋势是否一致，而不是只看单个最高分。
