ResearchBenchmarksConversational Quality Evaluation on TGC 10% human-annotated sampleFollow94.37Factual CorrectnessTGC86.51888.556590.59592.6335Apr 14, 2026Evaluation ResultsMethodMethodLinksFactual CorrectnessTopic ConsistencyCompletenessCoherenceNaturalnessRelevanceTGCEvaluation Protocol=Hu...Evaluation Protocol=Human evaluation (0-100 scale)2026.0494.3797.8288.2294.6797.11—TGCEvaluation Protocol=Hu...Evaluation Protocol=Human evaluation (0-100 scale)2026.0486.82————87.41