AI RESEARCH
Curriculum Sampling: A Two-Phase Curriculum for Efficient Training of Flow Matching
arXiv CS.LG
•
ArXi:2603.12517v1 Announce Type: new Timestep sampling $p(t)$ is a central design choice in Flow Matching models, yet common practice increasingly favors static middle-biased distributions (e.g., Logit-Normal). We show that this choice induces a speed--quality trade-off: middle-biased sampling accelerates early convergence but yields worse asymptotic fidelity than Uniform sampling. By analyzing per-timestep