SUPR
Postdoc on Diffusion & AR
Dnr:

NAISS 2025/22-539

Type:

NAISS Small Compute

Principal Investigator:

Ziyun Li

Affiliation:

Kungliga Tekniska högskolan

Start Date:

2025-04-03

End Date:

2026-05-01

Primary Classification:

10210: Artificial Intelligence

Webpage:

Allocation

Abstract

Combining Diffusion Models and Autoregression: this project aims to unify the strengths of Autoregressive (AR) and diffusion models, leveraging AR’s sequential factorization for language modeling and diffusion’s iterative refinement for visual generation. The goal is to create a flexible framework that integrates both paradigms for improved generative performance across domains.