Are we really tilting? The mechanics of reward guidance in flow and diffusion models 文章

ArXiv CS.AI2026-06-03NEWSen作者: Sanjit Dandapanthula, Nicholas M. Boffi
查看原文 →
关系图谱

Are we really tilting? The mechanics of reward guidance in flow and diffusion models · 相关技术

相关技术

plug-in estimatorquadratic rewardsGaussian Mixture Modelreward-guided diffusiondiffusion trainingneural reward functionsreward hackingDoob h-functionReward guidance algorithms