Discrete Diffusion with Sample-Efficient Estimators for Conditionals
Karthik Elamvazhuthi, Abhijith Jayakumar, Andrey Y. Lokhov · 2026
Read Discrete Diffusion with Sample-Efficient Estimators for Conditionals by Karthik Elamvazhuthi, Abhijith Jayakumar, Andrey Y. Lokhov (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-20293.txt.
Description
Karthik Elamvazhuthi and two co-authors propose a discrete diffusion method for generative modeling over discrete state spaces. Instead of approximating a discrete analogue of a score function, their 2026 approach treats single-site conditional probabilities as the core objects of the reverse diffusion process. Those conditionals are estimated with a sample-efficient method called the Neural Interaction Screening Estimator, paired with round-robin noising and denoising dynamics.
Controlled tests cover synthetic Ising models, MNIST, data from a D-Wave quantum annealer, a synthetic Potts model, and one-dimensional quantum systems. On binary data, the method beats popular ratio-based approaches on total variation, cross-correlations, and kernel density estimates. The work shows that estimating conditionals directly can be a strong basis for discrete diffusion models. It offers a practical route for discrete-data domains.
Catalog record
| CREATOR | Karthik Elamvazhuthi, Abhijith Jayakumar, Andrey Y. Lokhov |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 55 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-20293