ResearchBenchmarksBandit Optimization on Bandit TaskFollow-0.621Initial NMREvoTune-0.65205-0.636525-0.621-0.605475Feb 2, 2026Evaluation ResultsMethodMethodLinksInitial NMRFinal NMRImprovementEvoTuneType=BaselineType=Baseline2026.02-0.621-0.50918EvoDPOType=AblationType=Ablation2026.02-0.621-0.50718.3ATLASType=OursType=Ours2026.02-0.621-0.49320.6