ResearchBenchmarksRecommendation on ML-1M (rating >= 4.5, test)Follow14.65NDCG@10Exp-RSFT0.36044.07027.7811.4898Mar 10, 2026Evaluation ResultsMethodMethodLinksNDCG@10NDCG@50HR@10HR@50MRRAvg RewardExp-RSFT2026.0314.6520.2927.1252.6112.353.6933RSFT2026.0314.3720.0326.0851.6312.33.6558BC2026.0313.0418.6624.0549.6711.163.8672PPO2026.033.036.176.7321.242.854.4775DPO2026.030.912.311.638.241.223.3737