Research map / Probabilistic and scientific modeling / Variational and generative sampling
Unfolding Generative Flows with Koopman Operators: Trajectory-Preserving Linearization
Erkan Turan, Ari Siozopoulos, Louis Martinez, Julien Gaubil, Emery Pierson, Maks Ovsjanikov
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 |
|---|---|---|---|
| add consistency loss (weight 0.01 target) (ablation) | Koopman-CFM · FFHQ | FID: 7.5 -> 10.1 | Hurt |
| add consistency loss (ablation) | Koopman-CFM · CIFAR-10 | FID: 14.1 -> 16.7 | Hurt |
| add consistency loss (ablation) | Koopman-CFM · MNIST | FID: 6.43 -> 7.1 | Hurt |
| add target loss to recon+phase (ablation) | Koopman-CFM · FFHQ | FID: 41 -> 7.5 | Improved |
| add target loss to recon+phase (ablation) | Koopman-CFM · CIFAR-10 | FID: 64.5 -> 14.1 | Improved |
| add target loss to recon+phase (ablation) | Koopman-CFM · MNIST | FID: 143.5 -> 6.43 | Improved |
| add consistency loss (ablation) | Koopman-CFM · FFHQ trajectory fidelity | mean MSE (x10^-5): 130 -> 0.5 | Improved |
| add consistency loss (ablation) | Koopman-CFM · CIFAR-10 trajectory fidelity | mean MSE (x10^-5): 174 -> 1.0 | Improved |
| add consistency loss (ablation) | Koopman-CFM · FFHQ semantic modes | average max coherence: .45 -> .72 | Improved |
| increase Koopman dimension up to 1026 (ablation) | Koopman-CFM · FFHQ | FID | Improved |
| include matrix exponential in latent evolution | Koopman-CFM · Latency | latency (ms): 36.631 -> 38.838 | No clear effect |
Closest papers by meaning
- Hierarchical Koopman Diffusion: Fast Generation with Interpretable Diffusion Trajectory (NeurIPS 2025 · similarity 0.90)
- One-Step Offline Distillation of Diffusion-based Models via Koopman Modeling (NeurIPS 2025 · similarity 0.90)
- Expanding Flow Maps (NeurIPS 2026 · similarity 0.89)
- How to build a consistency model: Learning flow maps via self-distillation (NeurIPS 2025 · similarity 0.88)
- Flow Density Control: Generative Optimization Beyond Entropy-Regularized Fine-Tuning (NeurIPS 2025 · similarity 0.88)
- Improving Flow Matching by Aligning Flow Divergence (ICML 2025 · similarity 0.88)
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.
