ResearchBenchmarksOpen-ended Generation on NQ-SwapFollow44.91EMBaseline31.951635.315838.6842.0442Feb 10, 2026Evaluation ResultsMethodMethodLinksEMF1TruthfulnessBaselineModel=Qwen-2.5-14BModel=Qwen-2.5-14B2026.0244.910.541558.6DiverModel=Qwen-2.5-14BModel=Qwen-2.5-14B2026.02420.512857.1CoCoASIGModel=Qwen-2.5-14B, Mo...Model=Qwen-2.5-14B, Mode=fMLDS, Alpha=2.52026.0241.10.521559.7DoLaModel=Qwen-2.5-14BModel=Qwen-2.5-14B2026.0232.450.433747.35