ResearchBenchmarksText-to-Image Generation on DDPO BERTScoreText-to-Image Generation on DDPO BERTScoreFollow191EpochsDDPO144.2156.35168.5180.65Sep 30, 2025Evaluation resultsMethodMethodLinksEpochsSpeedupDDPOTarget Level=0.52, Rew...Target Level=0.52, Reward=BERTScore2025.09191—PCPOTarget Level=0.52, Bas...Target Level=0.52, Baseline Algorithm=DDPO, Reward=BERTScore2025.0914630.8