ResearchBenchmarksBehavioral accuracy evaluation on Synthetic TTS FRFollow71.6Text-Affinity Accuracy (Abase → TCF)Qwen55.37659.58863.868.012Jun 17, 2026Evaluation ResultsMethodMethodLinksText-Affinity Accuracy (Abase → TCF)Dominance Margin (ΔT-A) (Abase → TCF)Text-Affinity Accuracy (ACF → Tbase)Dominance Margin (ΔT-A) (ACF → Tbase)Text-Affinity Accuracy (Tbase → ACF)Dominance Margin (ΔT-A) (Tbase → ACF)Text-Affinity Accuracy (TCF → Abase)Dominance Margin (ΔT-A) (TCF → Abase)Text-Affinity Accuracy (Conflict Avg.)Dominance Margin (ΔT-A) (Conflict Avg.)Accuracy (Aligned Avg.)QwenLabel setting=limited...Label setting=limited labels2026.0671.60.43268.40.36841.2-0.17641.6-0.16855.70.11492.2UltravoxLabel setting=limited...Label setting=limited labels2026.06560.1256.80.13660.40.20859.60.19258.20.16494.6