ResearchBenchmarksMCQ selection similarity on BiasedLevel SyntheticFollow100Average SimilarityOracle85.4489.229396.78Mar 29, 2026Evaluation ResultsMethodMethodLinksAverage SimilarityOraclealpha=0alpha=02026.03100Oraclealpha=0.25alpha=0.252026.0398.3Oraclealpha=0.5alpha=0.52026.0397.5Oraclealpha=1alpha=12026.0397.4Oraclealpha=0.75alpha=0.752026.0396.7DQNalpha=0alpha=02026.0388.8DQNalpha=0.25alpha=0.252026.0388.5DQNalpha=1alpha=12026.0386.9DQNalpha=0.5alpha=0.52026.0386.2DQNalpha=0.75alpha=0.752026.0386