Research map / Image generation / Fast diffusion sampling
Optimal Stopping in Latent Diffusion Models
Yu-Han Wu, Quentin Berthet, Gérard Biau, Claire Boyer, Romuald Elie, Pierre Marion
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| stop the backward diffusion early instead of at t = T | latent diffusion model · ImageNet-256 (32x32x4 latents) | FID: minimum at t=0.956 rather than at the end | Improved |
| use a larger latent dimension | latent diffusion model · ImageNet-256 | optimal stopping time | Mixed |
| use a noisy autoencoder as proxy for the full LDM | LDM model selection · ImageNet-256 | FID curve crossing and optimal stopping time: LDM min-time 0.956, AE min-time 0.956 | Improved |
| use the exact score and very fine discretization on GMM data (ablation) | latent diffusion · GMM data | FID trajectory | No clear effect |
| project onto the data subspace dimension for low-rank data | latent diffusion · Gaussian with low-rank covariance | Frechet distance | Improved |
| use a larger latent dimension for linear AEs on MNIST | LDM with linear AE · MNIST (latent 64, 256, 400) | FID | Mixed |
| cap score weight norm C on exponentially decaying spectrum | clipped linear score · 10-dim Gaussian | optimal projection dimension: d_min in {5, 6} predicted and observed | Mixed |
| apply noisy AE proxy to pre-trained SD-VAE and FLUX-AE | noisy AE proxy · ImageNet-256 | FID | Improved |
Closest papers by meaning
- Probing the Latent Hierarchical Structure of Data via Diffusion Models (ICLR 2025 · similarity 0.90)
- Dimension-free Score Matching and Time Bootstrapping for Diffusion Models (NeurIPS 2025 · similarity 0.90)
- On Statistical Rates and Provably Efficient Criteria of Latent Diffusion Transformers (DiTs) (NeurIPS 2024 · similarity 0.89)
- On the Edge of Memorization in Diffusion Models (NeurIPS 2025 · similarity 0.89)
- O(d/T) Convergence Theory for Diffusion Probabilistic Models under Minimal Assumptions (ICLR 2025 · similarity 0.89)
- Generalization Dynamics of Linear Diffusion Models (NeurIPS 2026 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
