Research map / Image generation / Fast diffusion sampling
Bridge Matching Sampler: Scalable Sampling via Generalized Fixed-Point Diffusion Matching
Denis Blessing, Lorenz Richter, Julius Berner, Egor Malitskiy, Gerhard Neumann
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 damping (eta=10) (ablation) | BMS · ALA2 | JS divergence: 0.311 -> 0.068 | Improved |
| add damping (eta=10) (ablation) | BMS · ALA2 | TICA W2: 2.707 -> 0.900 | Improved |
| add damping (eta=10) (ablation) | ASBS · ALA2 | JS divergence: 0.232 -> 0.111 | Improved |
| add damping (eta=10) (ablation) | ASBS · ALA2 | TICA W2: 1.263 -> 1.526 | Hurt |
| add damping (eta=50) (ablation) | BMS · ALA4 | TICA W2: 280.3 -> 1.697 | Improved |
| add damping (eta=50) (ablation) | ASBS · ALA4 | JS divergence: 0.489 -> 0.230 | Improved |
| use BMS instead of ASBS with damping | BMS-damped · ALA2 | RMSE free-energy: 0.128 -> 0.079 | Improved |
| use BMS instead of ASBS with damping | BMS-damped · ALA4 | TICA W2: 1.606 -> 1.697 | Hurt |
| use BMS instead of ASBS | BMS · LJ-55 | energy W2: 28.10 -> 2.36 | Improved |
| use BMS instead of ASBS | BMS · LJ-13 | energy W2: 1.99 -> 0.53 | Improved |
| use BMS instead of ASBS | BMS · DW-4 | energy W2: 0.20 -> 0.21 | No clear effect |
| use BMS instead of AS | BMS · LJ-55 | W2: 4.50 -> 3.80 | Improved |
Closest papers by meaning
- On scalable and efficient training of diffusion samplers (NeurIPS 2025 · similarity 0.91)
- Underdamped Diffusion Bridges with Applications to Sampling (ICLR 2025 · similarity 0.91)
- Wasserstein Gradient Flows and Forward-Only Diffusion Are Not Enough for Multimodal Sampling (NeurIPS 2026 · similarity 0.90)
- Feedback Schrödinger Bridge Matching (ICLR 2025 · similarity 0.89)
- Test-time Scaling of Diffusions with Flow Maps (NeurIPS 2026 · similarity 0.89)
- Adjoint Sampling: Highly Scalable Diffusion Samplers via Adjoint Matching (ICML 2025 · 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.
