ResearchBenchmarksHuman Evaluation on Human Evaluation Set (test)Follow0.65Win RateLongDPO0.615680.624590.63350.64241Feb 4, 2025Evaluation ResultsMethodMethodLinksWin RateTie RateLose RateLongDPOCriteria=DiversityCriteria=Diversity2025.020.650.0830.267LongDPOCriteria=ConsistencyCriteria=Consistency2025.020.6170.1670.216LongDPOCriteria=InformativeCriteria=Informative2025.020.6170.0670.316