GenEval

The 'GenEval' dataset/benchmark is used to evaluate generative models, specifically focusing on their performance in generating high-quality images and text-to-image generation tasks.

Papers using GenEval (5)

Refinement via Regeneration: Enlarging Modification Space Boosts Image Refinement in Unified Multimodal Models2026CoF-T2I: Video Models as Pure Visual Reasoners for Text-to-Image Generation2026RAISE: Requirement-Adaptive Evolutionary Refinement for Training-Free Text-to-Image Alignment2026Rethinking UMM Visual Generation: Masked Modeling for Efficient Image-Only Pre-training2026Asymmetric Flow Models2026