Work

One-step Nonautoregressive Natural Language Generation with Shortcut Flow Matching Models

Year: 2026

Type: conference-paper

Abstract: While having a significant potential for parallel processing in theory, diffusion-based nonautoregressive text generation remains inefficient due to the need for multiple denoising steps.Performance degrades sharply if a low number of steps is used, such as in flow matching.To enable accurate one-step generation, we propose a novel shortcut flow-matching model that learns to directly predict multi-step denoising outcomes in a single step.Experiments conducted on three datasets demonstrate consistent improvements over classic flow-matching, with BLEU scores more than doubling on two datasets.We also tested five different ways of extending shortcut models with commonly used techniques.

Source: Proceedings of the conference - Association for Computational Linguistics. Meeting

DOI: 10.18653/v1/2026.acl-short.53

Authors: Jędrzej Warczyński, Ondrej Dusek, Mateusz Lango

Institutions: Charles University, Poznań University of Technology

Language: English

Topic: Topic Modeling

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: hybrid

Funders: European Commission, Infrastruktura PL-Grid, Academic Computer Centre Cyfronet, AGH University of Science and Technology

Award: G4976355562