ResearchBenchmarksMCQ Selection Similarity on BiasedTopic Synthetic DatasetFollow99.9Average SimilarityOracle84.50888.50492.596.496Mar 29, 2026Evaluation ResultsMethodMethodLinksAverage SimilarityOraclealpha=1alpha=12026.0399.9Oraclealpha=0alpha=02026.0399.7Oraclealpha=0.25alpha=0.252026.0399.5Oraclealpha=0.5alpha=0.52026.0399.4Oraclealpha=0.75alpha=0.752026.0399.4DQNalpha=0alpha=02026.0390.3DQNalpha=0.25alpha=0.252026.0388DQNalpha=0.5alpha=0.52026.0387DQNalpha=0.75alpha=0.752026.0386DQNalpha=1alpha=12026.0385.1