Research map / Image generation / Autoregressive and latent generation
On Diffusion Process of SE(3)-Invariant Manifold Over Pairwise Distance Matrices
Zihan Zhou, Ruiying Liu, Jiachen Zheng, Xiaoxue Wang, Tianshu Yu
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 Langevin dynamics with the proposed ODE sampler at 1000 steps | diffusion sampler · Human3.6M eating poses | mean AD: 3.7369 -> 2.1620 | Improved |
| replace Langevin dynamics with the proposed SDE sampler at 100 steps | diffusion sampler · Human3.6M eating poses | mean AD: 5.1739 -> 2.3737 | Improved |
| replace Langevin dynamics with the proposed ODE sampler at 100 steps | diffusion sampler · Human3.6M eating poses | mean AD: 5.1739 -> 2.8978 | Improved |
| reduce sampling steps from 1000 to 100 for LD | Langevin dynamics · Human3.6M eating poses | mean AD: 3.7369 -> 5.1739 | Hurt |
| reduce sampling steps from 1000 to 100 for SDE | proposed SDE · Human3.6M eating poses | mean AD: 2.3315 -> 2.3737 | No clear effect |
| reduce sampling steps from 1000 to 100 for ODE | proposed ODE · Human3.6M eating poses | mean AD: 2.1620 -> 2.8978 | Mixed |
| replace LD 5000 steps with ODE 1000 steps | diffusion sampler · GEOM-QM9 | COV-R mean / MAT-R mean: 90.07 -> 90.24; 0.2090 -> 0.2052 | No clear effect |
| replace LD 5000 steps with SDE 1000 steps | diffusion sampler · GEOM-QM9 | COV-R mean / MAT-R mean: 90.07 -> 89.25; 0.2090 -> 0.2215 | No clear effect |
| replace LD 1000 steps with ODE 1000 steps | diffusion sampler · GEOM-Drugs | COV-R mean / MAT-R mean: 82.92 -> 90.64; 0.9525 -> 0.8493 | Improved |
| replace LD 5000 steps with ODE 1000 steps | diffusion sampler · GEOM-Drugs | COV-P mean / MAT-P mean: 61.47 -> 59.67; 1.1712 -> 1.1915 | Mixed |
| reduce ODE steps from 1000 to 100 | proposed ODE · GEOM-Drugs | MAT-P mean: 1.1915 -> 1.8528 | Hurt |
| add the corrector to the reverse SDE/ODE | proposed SDE/ODE · molecular and pose generation | sample quality | No clear effect |
Closest papers by meaning
- Manifold Diffusion Fields (ICLR 2024 · similarity 0.84)
- Scalable Non-Equivariant 3D Molecule Generation via Rotational Alignment (ICML 2025 · similarity 0.84)
- Kinetic Langevin Diffusion for Crystalline Materials Generation (ICML 2025 · similarity 0.83)
- Geometric Trajectory Diffusion Models (NeurIPS 2024 · similarity 0.83)
- Manifolds, Random Matrices and Spectral Gaps: The geometric phases of generative diffusion (ICLR 2025 · similarity 0.83)
- Space Group Equivariant Crystal Diffusion (NeurIPS 2025 · similarity 0.83)
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.
