Efficient-DLM: From Autoregressive to Diffusion Language Models, and Beyond in Speed

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Fu, Yonggan, Whalen, Lexington, Ye, Zhifan, Dong, Xin, Diao, Shizhe, Liu, Jingyu, Wu, Chengyue, Zhang, Hao, Xie, Enze, Han, Song, Khadkevich, Maksim, Kautz, Jan, Lin, Yingyan Celine, Molchanov, Pavlo
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915967949340672
author Fu, Yonggan
Whalen, Lexington
Ye, Zhifan
Dong, Xin
Diao, Shizhe
Liu, Jingyu
Wu, Chengyue
Zhang, Hao
Xie, Enze
Han, Song
Khadkevich, Maksim
Kautz, Jan
Lin, Yingyan Celine
Molchanov, Pavlo
author_facet Fu, Yonggan
Whalen, Lexington
Ye, Zhifan
Dong, Xin
Diao, Shizhe
Liu, Jingyu
Wu, Chengyue
Zhang, Hao
Xie, Enze
Han, Song
Khadkevich, Maksim
Kautz, Jan
Lin, Yingyan Celine
Molchanov, Pavlo
contents Diffusion language models (dLMs) have emerged as a promising paradigm that enables parallel, non-autoregressive generation, but their learning efficiency lags behind that of autoregressive (AR) language models when trained from scratch. To this end, we study AR-to-dLM conversion to transform pretrained AR models into efficient dLMs that excel in speed while preserving AR models' task accuracy. We achieve this by identifying limitations in the attention patterns and objectives of existing AR-to-dLM methods and then proposing principles and methodologies for more effective AR-to-dLM conversion. Specifically, we first systematically compare different attention patterns and find that maintaining pretrained AR weight distributions is critical for effective AR-to-dLM conversion. As such, we introduce a continuous pretraining scheme with a block-wise attention pattern, which remains causal across blocks while enabling bidirectional modeling within each block. We find that this approach can better preserve pretrained AR models' weight distributions than fully bidirectional modeling, in addition to its known benefit of enabling KV caching, and leads to a win-win in accuracy and efficiency. Second, to mitigate the training-test gap in mask token distributions (uniform vs. highly left-to-right), we propose a position-dependent token masking strategy that assigns higher masking probabilities to later tokens during training to better mimic test-time behavior. Leveraging this framework, we conduct extensive studies of dLMs' attention patterns, training dynamics, and other design choices, providing actionable insights into scalable AR-to-dLM conversion. These studies lead to the Efficient-DLM family, which outperforms state-of-the-art AR models and dLMs, e.g., our Efficient-DLM 8B achieves +5.4%/+2.7% higher accuracy with 4.5x/2.7x higher throughput compared to Dream 7B and Qwen3 4B, respectively.
format Preprint
id arxiv_https___arxiv_org_abs_2512_14067
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Efficient-DLM: From Autoregressive to Diffusion Language Models, and Beyond in Speed
Fu, Yonggan
Whalen, Lexington
Ye, Zhifan
Dong, Xin
Diao, Shizhe
Liu, Jingyu
Wu, Chengyue
Zhang, Hao
Xie, Enze
Han, Song
Khadkevich, Maksim
Kautz, Jan
Lin, Yingyan Celine
Molchanov, Pavlo
Computation and Language
Artificial Intelligence
Machine Learning
Diffusion language models (dLMs) have emerged as a promising paradigm that enables parallel, non-autoregressive generation, but their learning efficiency lags behind that of autoregressive (AR) language models when trained from scratch. To this end, we study AR-to-dLM conversion to transform pretrained AR models into efficient dLMs that excel in speed while preserving AR models' task accuracy. We achieve this by identifying limitations in the attention patterns and objectives of existing AR-to-dLM methods and then proposing principles and methodologies for more effective AR-to-dLM conversion. Specifically, we first systematically compare different attention patterns and find that maintaining pretrained AR weight distributions is critical for effective AR-to-dLM conversion. As such, we introduce a continuous pretraining scheme with a block-wise attention pattern, which remains causal across blocks while enabling bidirectional modeling within each block. We find that this approach can better preserve pretrained AR models' weight distributions than fully bidirectional modeling, in addition to its known benefit of enabling KV caching, and leads to a win-win in accuracy and efficiency. Second, to mitigate the training-test gap in mask token distributions (uniform vs. highly left-to-right), we propose a position-dependent token masking strategy that assigns higher masking probabilities to later tokens during training to better mimic test-time behavior. Leveraging this framework, we conduct extensive studies of dLMs' attention patterns, training dynamics, and other design choices, providing actionable insights into scalable AR-to-dLM conversion. These studies lead to the Efficient-DLM family, which outperforms state-of-the-art AR models and dLMs, e.g., our Efficient-DLM 8B achieves +5.4%/+2.7% higher accuracy with 4.5x/2.7x higher throughput compared to Dream 7B and Qwen3 4B, respectively.
title Efficient-DLM: From Autoregressive to Diffusion Language Models, and Beyond in Speed
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2512.14067