ResearchBenchmarksText-to-Image Generation on NegGenBench 10 Selected Prompts Human Labelled (test)Follow1Positive ScoreNAG0.8960.9230.950.977Aug 11, 2025Evaluation ResultsMethodMethodLinksPositive ScoreNegative ScoreQuality ScoreNAGvariant=defaultvariant=default2025.0810.10.895NAGvariant=Strong, hyperp...variant=Strong, hyperparameters=re-tuned2025.080.950.250.675NASA2025.080.950.150.685VSFvariant=Qualityvariant=Quality2025.080.90.550.823