Broader noise schedules and practical discretizations

Extend the geometric–distribution separation guarantees to broader noise schedules and practically used discrete samplers by establishing results that account for finite-step implementations, step-size selection, truncated time horizons, and non‑uniform time weighting while preserving the separation between geometry learning and distribution learning.

Background

The theoretical guarantees are derived in a continuous-time framework with an idealized large-noise training condition and a hybrid SDE/ODE sampler. Practical deployments often use discrete-time samplers and varied noise schedules.

The authors highlight the need to carry over the geometry–distribution separation to these realistic settings and training variations.

References

Several directions remain open: Noise schedules, discretizations, and training idealizations. It would be valuable to extend the theory to broader noise schedules and practically used discretizations, including the effects of finite-step samplers, step-size selection, and common training variations (e.g., truncated time horizons or non-uniform time weighting), while preserving a comparable separation between geometry learning and distribution learning.

— Manifold Generalization Provably Proceeds Memorization in Diffusion Models  (2603.23792 - Shen et al., 24 Mar 2026) in Conclusion, Open directions (2)

It is possible that higher-order discretizations and other more sophisticated procedures could admit different local geometries, and hence potentially different optimal clocks for stepsize allocation.

— The data geometry of masking diffusion: Certified-optimal schedules via unmasking growth complexity  (2608.13520 - Wainwright, 13 Aug 2026) in Section Discussion

Tighter constant estimates---needed only for quantitatively tight versions of the bound---would require spectral norm tracking and are left to future work.

— HybridSB-MoE: Dual-Domain Schrödinger Bridges with Scene-Adaptive Expert Routing for Speech Enhancement  (2608.12715 - Lu et al., 13 Aug 2026) in Appendix, Section \ref{app:assumption_justification}, subsection “Justification of Assumption~\ref{assu:reg}”

Given that the schedule intimately interplays with Gaussian transition kernel approximation, we leave studying the effects it has on the weight computation for future work.

— Steering Diffusion Models to Rare Events with Sequential Monte Carlo  (2610.08652 - Subedi et al., 6 Oct 2026) in Appendix, Section “Sources of Bias,” subsection “The discretised law: grid and integrator,” paragraph “Grid placement”