Test-Time Anchoring for Discrete Diffusion Posterior Sampling
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917233111859200 |
|---|---|
| author | Rout, Litu Lugmayr, Andreas Jafarian, Yasamin Varadharajan, Srivatsan Caramanis, Constantine Shakkottai, Sanjay Kemelmacher-Shlizerman, Ira |
| author_facet | Rout, Litu Lugmayr, Andreas Jafarian, Yasamin Varadharajan, Srivatsan Caramanis, Constantine Shakkottai, Sanjay Kemelmacher-Shlizerman, Ira |
| contents | While continuous diffusion models have achieved remarkable success, discrete diffusion offers a unified framework for jointly modeling text and images. Beyond unification, discrete diffusion provides faster inference, finer control, and principled training-free guidance, making it well-suited for posterior sampling. Existing approaches to posterior sampling using discrete diffusion face severe challenges: derivative-free guidance yields sparse signals, continuous relaxations limit applicability, and split Gibbs samplers suffer from the curse of dimensionality. To overcome these limitations, we introduce Anchored Posterior Sampling (APS), built on two key innovations: quantized expectation for gradient-like guidance in discrete embedding space, and anchored remasking for adaptive decoding. APS achieves state-of-the-art performance among discrete diffusion samplers on both linear and nonlinear inverse problems across the standard image benchmarks. We demonstrate the generality of APS through training-free stylization and text-guided editing. We further apply APS to a large-scale diffusion language model, showing consistent improvement in question answering. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_02291 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Test-Time Anchoring for Discrete Diffusion Posterior Sampling Rout, Litu Lugmayr, Andreas Jafarian, Yasamin Varadharajan, Srivatsan Caramanis, Constantine Shakkottai, Sanjay Kemelmacher-Shlizerman, Ira Machine Learning Computer Vision and Pattern Recognition While continuous diffusion models have achieved remarkable success, discrete diffusion offers a unified framework for jointly modeling text and images. Beyond unification, discrete diffusion provides faster inference, finer control, and principled training-free guidance, making it well-suited for posterior sampling. Existing approaches to posterior sampling using discrete diffusion face severe challenges: derivative-free guidance yields sparse signals, continuous relaxations limit applicability, and split Gibbs samplers suffer from the curse of dimensionality. To overcome these limitations, we introduce Anchored Posterior Sampling (APS), built on two key innovations: quantized expectation for gradient-like guidance in discrete embedding space, and anchored remasking for adaptive decoding. APS achieves state-of-the-art performance among discrete diffusion samplers on both linear and nonlinear inverse problems across the standard image benchmarks. We demonstrate the generality of APS through training-free stylization and text-guided editing. We further apply APS to a large-scale diffusion language model, showing consistent improvement in question answering. |
| title | Test-Time Anchoring for Discrete Diffusion Posterior Sampling |
| topic | Machine Learning Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2510.02291 |