ResearchBenchmarksGenerative Recommendation on RecSim Extreme Noisy Noise DominatedFollow0.318RewardDRPO-Exp0.116240.168620.2210.27338Feb 11, 2026Evaluation ResultsMethodMethodLinksRewardeCPMDRPO-Exp2026.020.3181.59DRPO2026.020.3081.54Adapt. BCmechanism=Adaptive fil...mechanism=Adaptive filtering2026.020.2971.48IQL2026.020.2891.44BC2026.020.2681.34AWR2026.020.2541.27CRR2026.020.2171.08BPPO2026.020.1240.62