Paper recorded by Signals 4 on 2026-09-18 in cs.LG. Abstract reproduced from arXiv; link to the original below.
Published 2026-09-18 on arXiv · recorded by Signals 4 on 2026-09-21
Category: cs.LG · 机器学习 · first seen 2026-09-21
Reinforcement learning is increasingly used to align image generators with reward signals, and Flow-GRPO recently extended this paradigm to flow-matching models by treating the denoising sampler as a stochastic policy that can be optimized from reward feedback. Training in this setting is unstable in a way specific to multi-step denoising: the policy update changes systematically across denoising