AI RESEARCH
Stable GFlowNets with Probabilistic Guarantees
arXiv CS.LG
•
ArXi:2605.01729v1 Announce Type: new Generative Flow Networks (GFlowNets) is often unstable, exhibiting severe loss spikes and mode collapse. To tackle this, we first assess the sensitivity of GFlowNet objectives, nstrating that a small Total Variation (TV) distance between the learned and target distributions does not preclude unbounded