FlowDPO
PulseAugur coverage of FlowDPO — every cluster mentioning FlowDPO across labs, papers, and developer communities, ranked by signal.
-
FlowCPO introduces unified divergence view for preference alignment in flow models
Researchers have introduced FlowCPO, a novel method for aligning flow and diffusion models using an offline forward-KL objective. This approach unifies existing online reinforcement learning and offline preference optim…
-
New ThermoDPO method tackles manifold drift in generative model alignment
Researchers have identified "manifold drift" as a key issue in extending preference optimization methods to continuous-time dynamics for generative models. This phenomenon occurs when reward-driven updates in flow match…
-
New unsupervised framework APO predicts 3D atomic structures
Researchers have developed APO (Atomic Policy Optimization), a novel unsupervised framework for predicting the 3D structures of atomic systems. This method eliminates the need for expensive experimental labels by adapti…