Research map / Image generation / Fast diffusion sampling
Discrete Diffusion Samplers and Bridges: Off-Policy Algorithms and Applications in Latent Spaces
Arran Carter, Sanghyeok Choi, Kirill Tamogashev, Víctor Elvira, Esmeralda S. Whitammer
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 importance-weighted buffer and MCMC exploration to TB training | TB discrete diffusion sampler · Ising 16x16 (beta=1.2) | Sinkhorn distance: 127.82 -> 0.02 | Improved |
| add importance-weighted buffer to TB training | TB discrete diffusion sampler · Ising 16x16 (beta=1.2) | Sinkhorn distance: 127.82 -> 50.94 | Improved |
| add MCMC exploration on top of buffer for TB | TB + Buffer · Potts q=3 (beta=1.2) | Sinkhorn distance: 156.12 -> 12.37 | Improved |
| add buffer and MCMC to TB | TB discrete diffusion sampler · 40GMM (d=32) | Sinkhorn distance: 2142.65 -> 4.25 | Improved |
| add buffer and MCMC to LV | LV discrete diffusion sampler · Ising 16x16 (beta=1.2) | Sinkhorn distance: 126.89 -> 0.04 | Improved |
| add MCMC exploration on top of buffer for TB | TB + Buffer · Ising 16x16 (beta=0.4407) | Sinkhorn distance: 46.44 -> 46.22 | No clear effect |
| add buffer and MCMC to TB | TB discrete diffusion sampler · ManyWell (d=80) | EUBO: 57.48 -> 52.84 | Improved |
| use TB with buffer and MCMC instead of MDNS (WDCE) | discrete diffusion sampler · 40GMM (d=32) | Sinkhorn distance: 349.31 -> 4.25 | Improved |
| add buffer and MCMC to TB (vs on-policy) | TB discrete diffusion sampler · 40GMM (d=16) | Sinkhorn distance: 156.58 -> 0.35 | Improved |
Closest papers by meaning
- Scalable Discrete Diffusion Samplers: Combinatorial Optimization and Statistical Physics (ICLR 2025 · similarity 0.92)
- Improved off-policy training of diffusion samplers (NeurIPS 2024 · similarity 0.92)
- Outsourced Diffusion Sampling: Efficient Posterior Inference in Latent Spaces of Generative Models (ICML 2025 · similarity 0.90)
- Underdamped Diffusion Bridges with Applications to Sampling (ICLR 2025 · similarity 0.89)
- On scalable and efficient training of diffusion samplers (NeurIPS 2025 · similarity 0.89)
- Particle Denoising Diffusion Sampler (ICML 2024 · 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.
