List your company
ResearchFormerly WizWand

Text-to-image synthesis evaluation (Human Correlation - Overall) on PaintSkills

0.6437Kendall's TauLLMScore
0.0447640.2002570.355750.511243May 18, 2023
Updated 8mo ago

Evaluation results

MethodLinks
LLMScoreHuman ranking criteria=Error Counting2023.050.64370.7325
BLIP-ITCHuman ranking criteria=Error Counting2023.050.61780.6511
LLMScore2023.050.560.6853
LLMScoreHuman ranking criteria=Overall2023.050.560.6853
NegCLIP2023.050.49110.6313
NegCLIPHuman ranking criteria=Overall2023.050.49110.6313
BLIP-ITM2023.050.47550.6214
BLIP-ITMHuman ranking criteria=Overall2023.050.47550.6214
BLIP-ITC2023.050.47430.5864
BLIP-ITCHuman ranking criteria=Overall2023.050.47430.5864
CLIP2023.050.47150.5869
CLIPHuman ranking criteria=Overall2023.050.47150.5869
BLIP-ITMHuman ranking criteria=Error Counting2023.050.46220.4997
NegCLIPHuman ranking criteria=Error Counting2023.050.45630.4908
DescCLIPtype=LLMScore variant2023.050.38250.4917
CLIPHuman ranking criteria=Error Counting2023.050.30520.2891
CapCLIPtype=LLMScore variant2023.050.30350.3751
CapMETEORtype=LLMScore variant2023.050.23150.3056
DescMETEORtype=LLMScore variant2023.050.06780.0827