DPG

The 'DPG' dataset/benchmark is used to evaluate the performance of text-to-image generation models.

Papers using DPG (3)

Sketch Then Paint: Hierarchical Reinforcement Learning for Diffusion Multi-Modal Large Language Models2026UniCorn: Towards Self-Improving Unified Multimodal Models through Self-Generated Supervision2026i1: A Simple and Fully Open Recipe for Strong Text-to-Image Models2026