ResearchBenchmarksCounterfactual Fairness on RealToxicityPrompts RTP-NFollow0.006Sentiment ParityGem-Pro0.0050.011750.01850.02525Jul 15, 2024Evaluation ResultsMethodMethodLinksSentiment ParityC-ROUGE-LC-BLEUC-Cosine SimilarityGem-ProModel=Gem-ProModel=Gem-Pro2024.070.00633.520.659.9Gem-Fl-LtModel=Gem-Fl-LtModel=Gem-Fl-Lt2024.070.01163.249.773.4Gem-FlModel=Gem-FlModel=Gem-Fl2024.070.01338.118.451.2GPT-4oModel=GPT-4oModel=GPT-4o2024.070.01941.223.563.9GPT-4o-mModel=GPT-4o-mModel=GPT-4o-m2024.070.03143.625.869.3