ResearchBenchmarksEfficiency Analysis on Alignment (train)Follow5Training TimeRLOO-16125.75267.5409.25Mar 5, 2025Evaluation ResultsMethodMethodLinksTraining TimeRLOOTraining Mode=Single-I...Training Mode=Single-Iteration2025.035ZOPrOTraining Mode=Single-I...Training Mode=Single-Iteration2025.03120DPOTraining Mode=Single-I...Training Mode=Single-Iteration2025.03149PPOTraining Mode=Single-I...Training Mode=Single-Iteration2025.03530