Autoregressive Models Rival Diffusion Models at ANY-ORDER Generation

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Du, Tianqi, Fang, Lizhe, Yang, Weijie, Zhang, Chenheng, Wei, Zeming, Wang, Yifei, Wang, Yisen
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866918296514723840
author Du, Tianqi
Fang, Lizhe
Yang, Weijie
Zhang, Chenheng
Wei, Zeming
Wang, Yifei
Wang, Yisen
author_facet Du, Tianqi
Fang, Lizhe
Yang, Weijie
Zhang, Chenheng
Wei, Zeming
Wang, Yifei
Wang, Yisen
contents Diffusion language models enable any-order generation and bidirectional conditioning, offering appealing flexibility for tasks such as infilling, rewriting, and self-correction. However, their formulation-predicting one part of a sequence from another within a single-step dependency-limits modeling depth and often yields lower sample quality and stability than autoregressive (AR) models. To address this, we revisit autoregressive modeling as a foundation and reformulate diffusion-style training into a structured multi-group prediction process. We propose Any-order Any-subset Autoregressive modeling (A3), a generalized framework that extends the standard AR factorization to arbitrary token groups and generation orders. A3 preserves the probabilistic rigor and multi-layer dependency modeling of AR while inheriting diffusion models' flexibility for parallel and bidirectional generation. We implement A3 through a two-stream attention architecture and a progressive adaptation strategy that transitions pretrained AR models toward any-order prediction. Experiments on question answering, commonsense reasoning, and story infilling demonstrate that A3 outperforms diffusion-based models while maintaining flexible decoding. This work offers a unified approach for a flexible, efficient, and novel language modeling paradigm.
format Preprint
id arxiv_https___arxiv_org_abs_2601_13228
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Autoregressive Models Rival Diffusion Models at ANY-ORDER Generation
Du, Tianqi
Fang, Lizhe
Yang, Weijie
Zhang, Chenheng
Wei, Zeming
Wang, Yifei
Wang, Yisen
Computation and Language
Artificial Intelligence
Diffusion language models enable any-order generation and bidirectional conditioning, offering appealing flexibility for tasks such as infilling, rewriting, and self-correction. However, their formulation-predicting one part of a sequence from another within a single-step dependency-limits modeling depth and often yields lower sample quality and stability than autoregressive (AR) models. To address this, we revisit autoregressive modeling as a foundation and reformulate diffusion-style training into a structured multi-group prediction process. We propose Any-order Any-subset Autoregressive modeling (A3), a generalized framework that extends the standard AR factorization to arbitrary token groups and generation orders. A3 preserves the probabilistic rigor and multi-layer dependency modeling of AR while inheriting diffusion models' flexibility for parallel and bidirectional generation. We implement A3 through a two-stream attention architecture and a progressive adaptation strategy that transitions pretrained AR models toward any-order prediction. Experiments on question answering, commonsense reasoning, and story infilling demonstrate that A3 outperforms diffusion-based models while maintaining flexible decoding. This work offers a unified approach for a flexible, efficient, and novel language modeling paradigm.
title Autoregressive Models Rival Diffusion Models at ANY-ORDER Generation
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2601.13228