Skip to main content
GDPO-Listener: Expressive Interactive Head Generation via Auto-Regressive Flow Matching and Group reward-Decoupled Policy Optimization | Signal Canvas | ScienceToStartup