Research map / Image generation / Fast diffusion sampling
Information-Geometric Adaptive Sampling for Graph Diffusion
Yuhui Lu, Wenjing Liu, Kun Zhan
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 |
|---|---|---|---|
| replace fixed-step schedule with DVS adaptive steps | GruM Euler-Maruyama · QM9 | FCD: 0.107 -> 0.095 | Improved |
| replace fixed-step schedule with DVS adaptive steps | GruM Heun · ZINC250k | FCD: 2.182 -> 2.086 | Improved |
| replace quadratic schedule with DVS adaptive steps | GruM Euler-Maruyama · ZINC250k | FCD: 2.194 -> 2.092 | Improved |
| replace fixed-step schedule with DVS in Heun | GruM Heun · ZINC250k | Valid (%): 98.39 -> 98.45 | No clear effect |
| replace fixed-step Euler-Maruyama with DVS sampler | GDSS · QM9 | FCD: 2.551 -> 2.482 | Improved |
| replace fixed-step Euler-Maruyama with DVS sampler | GDSS · Ego-small | Clustering MMD: 0.0184 -> 0.0149 | Improved |
| replace fixed-step Euler-Maruyama with DVS sampler | GruM · SBM | Spectral MMD: 0.0051 -> 0.0030 | Improved |
| replace fixed-step Euler-Maruyama with DVS sampler | GruM · QM9 | number of steps (NFE): 1000 -> 787 | Improved |
| replace fixed-step with DVS sampler (runtime) | GruM Euler-Maruyama · QM9 | total sampling time (s): 843 -> 678 | Improved |
| replace fixed-step with DVS sampler (runtime) | GruM Euler-Maruyama · ZINC250k | total sampling time (s): 5695 -> 5803 | Hurt |
| restrict DVS to an active time range instead of full trajectory (ablation) | GruM DVS · Planar | Spectral MMD: 0.0058 -> 0.0052 | Improved |
| restrict DVS to an active time range instead of full trajectory (ablation) | GruM DVS · SBM | Degree MMD: 0.0002 -> 0.0007 | Hurt |
Closest papers by meaning
- DiPhon: Diffusion on Graphons for Scalable Graph Generation (NeurIPS 2026 · similarity 0.88)
- Toward a Unified Geometry Understanding : Riemannian Diffusion Framework for Graph Generation and Prediction (NeurIPS 2025 · similarity 0.88)
- Discrete-state Continuous-time Diffusion for Graph Generation (NeurIPS 2024 · similarity 0.87)
- Graph Generation with Diffusion Mixture (ICML 2024 · similarity 0.87)
- Generative Modelling of Structurally Constrained Graphs (NeurIPS 2024 · similarity 0.86)
- Elucidating Flow Matching ODE Dynamics via Data Geometry and Denoisers (ICML 2025 · similarity 0.86)
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.
