Vision‑Language‑Vision Auto‑Encoder: Scalable Knowledge Distillation from Diffusion Models | Read Paper on Bytez