Categories Companies Research

Loading the SOTA2 catalog…

Explore

Categories Companies Research

Contribute

Join SOTA2 Contact

SOTA2

© 2026 SOTA2

Build in Bellevue, USA

WizWand is now SOTA2 Research

Papers Benchmarks Datasets Tasks

Text-to-Image Generation research benchmarks – Page 22 · SOTA2 Research

Text-to-Image Generation

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

89.1Fidelity Score (Single Color)

7

Jul 7, 2026

4.69HPSv2.1

7

Jul 7, 2026

SD Attribute-neutral prompts Gender-Profession 1.5

2.7CEO

7

Jul 2, 2026

Visual Genome (train test)

38.61Inception Score

7

Jul 2, 2026

COCO SD 1.5 2017 (val)

20.1FID

7

Jun 30, 2026

Image Quality & Realism Evaluation Set

0.905Reward (Baseline)

7

Jun 29, 2026

LongText-Bench Chinese

94.6Score

7

Jun 26, 2026

89Aesthetic Quality

7

Jun 24, 2026

83Aesthetic Quality Score

7

Jun 24, 2026

MS-COCO subset of 50 prompts 2015

Semantic Browsing

3.34Vendi

7

Jun 23, 2026

Unsafe-1K adversarial prompts

97Total Unsafe Images (T)

7

Jun 8, 2026

I2P sexual prompts

40Total Unsafe Count

7

Jun 8, 2026

87.57VQA Score

7

Jun 8, 2026

Flow-CPS + AdaGRPO

0.327HPS v2

7

Jun 8, 2026

General Prompts SD v3.5

0.793Similarity

7

Jun 2, 2026

DeAttack (Textual Inversion)

0.258CLIP-T

7

May 29, 2026

15.21FID

7

May 29, 2026

Multi-reward CLIP, HPSv2.1, and GenEval (test)

state-aligned latent actor-critic framework

28.96CLIP Score

7

May 28, 2026

OCR-based text-rendering

0.02OCR Score

7

May 28, 2026

DrawWaldoWorlds (Tier C)

56VQA Accuracy

7

May 25, 2026

DrawWaldoWorlds (Tier B)

72VQA Accuracy

7

May 25, 2026

DrawWaldoWorlds (Tier A)

84VQA Accuracy

7

May 25, 2026

0.221PickScore

7

May 21, 2026

PartiPrompt (val)

0.221PickScore

7

May 21, 2026

0.348HPSv2

7

May 20, 2026

Page 22 of 39

Popular datasets

GenEval MS-COCO DrawBench COCO T2I-CompBench

Follow for updates

Join SOTA2 Contact