Restrained Diffusion Bridges: Targeted Denoiser Training and Refinement
Organizations: Courant Institute School of Mathematics, Computing, and Data Science, New York University, New York, New York 10012, United States · Department of Chemistry and James Franck Institute, University of Chicago, Chicago, Illinois 60637, United States
Abstract
Diffusion models are often inaccurate in critical, data-sparse regions of the target distribution. Sampling from restrained distributions, biased distributions used in classical enhanced sampling to concentrate data near chosen configurations, can control where training data are collected but biases the learned distribution. We establish that this bias cancels exactly when forward noising from restrained data is conditioned on the restrained configuration at an appropriately chosen intermediate time. This allows a pretrained model fine-tuned on restrained data to achieve high accuracy in any chosen region of the target distribution, regardless of its probability, using standard denoising or flow matching objectives without importance weights. When no target data are available, Gibbs sampling enables inference from a model trained exclusively on restrained data.
Figures & tables
Appendix figures & tables8 assets
Supplementary material from the paper’s appendix.
Appendix
| Basin | minimum (rad) | minimum (rad) | |
|---|---|---|---|
| Baseline model training dataset | RDB model training dataset | |||
|---|---|---|---|---|
| Matched MD budget ( s) | Production interval per trajectory ( s) | Training configurations | Restrained-MD probes | Retained training pairs |
| 5.25 | 100,000 | 1,000 | 30,000 | |
| 10.25 | 200,000 | 2,000 | 60,000 | |
| 15.25 | 300,000 | 3,000 | 90,000 | |
| 20.25 | 400,000 | 4,000 | 120,000 | |
| 25.25 | 500,000 | 5,000 | 150,000 | |
| MD budget ( s) | Source interval ( s) | Frames |
|---|---|---|
| 7 | 500,000 | |
| 12 | 1,000,000 | |
| 17 | 1,500,000 | |
| 22 | 2,000,000 |
| MD budget ( s) | Method | (%) | (%) | 2D NO JS distance | |
|---|---|---|---|---|---|
| Median [Q1, Q3] | Median | [Q1, Q3] | |||
| 7 | Base | 0.2330 [0.2302, 0.2440] | 37.8 | 0.13290 | [0.13179, 0.13386] |
| Synthetic, | 0.3320 [0.3132, 0.3558] | 11.4 | 0.14217 | [0.14167, 0.14316] | |
| Synthetic, | 0.3010 [0.2833, 0.3145] | 19.7 | 0.13552 | [0.13453, 0.13655] | |
| 12 | Base | 0.3130 [0.2993, 0.3158] | 16.5 | 0.11184 | [0.11153, 0.11268] |
| Synthetic, | 0.3930 [0.3862, 0.3970] | 4.9 | 0.12443 | [0.12254, 0.12498] | |
Diffusion Models
Diffusion models learn to reverse gradual data corruption, enabling flexible generation, editing, forecasting, and inference.
See all 1,275 papers