Diffusion Language Models are Super Data Learners

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ni, Jinjie, Liu, Qian, Dou, Longxu, Du, Chao, Wang, Zili, Yan, Hang, Pang, Tianyu, Shieh, Michael Qizhe
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909887548620800
author Ni, Jinjie
Liu, Qian
Dou, Longxu
Du, Chao
Wang, Zili
Yan, Hang
Pang, Tianyu
Shieh, Michael Qizhe
author_facet Ni, Jinjie
Liu, Qian
Dou, Longxu
Du, Chao
Wang, Zili
Yan, Hang
Pang, Tianyu
Shieh, Michael Qizhe
contents Under strictly controlled pre-training settings, we observe a Crossover: when unique data is limited, diffusion language models (DLMs) consistently surpass autoregressive (AR) models by training for more epochs. The crossover shifts later with more or higher-quality data, earlier with larger models, and persists across dense and sparse architectures. We attribute the gains to three compounding factors: (1) any-order modeling, (2) super-dense compute from iterative bidirectional denoising, and (3) built-in Monte Carlo augmentation; input or parameter noise improves AR under data constraint but cannot close the gap. At scale, a 1.7B DLM trained with a ~1.5T-token compute budget on 10B unique Python tokens overtakes an AR coder trained with strictly matched settings. In addition, a 1B-parameter DLM achieves > 56% accuracy on HellaSwag and > 33% on MMLU using only 1B tokens, without any special tricks, just by repeating standard pre-training data. We also show that rising validation cross-entropy does not imply degraded downstream performance in this regime.
format Preprint
id arxiv_https___arxiv_org_abs_2511_03276
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Diffusion Language Models are Super Data Learners
Ni, Jinjie
Liu, Qian
Dou, Longxu
Du, Chao
Wang, Zili
Yan, Hang
Pang, Tianyu
Shieh, Michael Qizhe
Machine Learning
Under strictly controlled pre-training settings, we observe a Crossover: when unique data is limited, diffusion language models (DLMs) consistently surpass autoregressive (AR) models by training for more epochs. The crossover shifts later with more or higher-quality data, earlier with larger models, and persists across dense and sparse architectures. We attribute the gains to three compounding factors: (1) any-order modeling, (2) super-dense compute from iterative bidirectional denoising, and (3) built-in Monte Carlo augmentation; input or parameter noise improves AR under data constraint but cannot close the gap. At scale, a 1.7B DLM trained with a ~1.5T-token compute budget on 10B unique Python tokens overtakes an AR coder trained with strictly matched settings. In addition, a 1B-parameter DLM achieves > 56% accuracy on HellaSwag and > 33% on MMLU using only 1B tokens, without any special tricks, just by repeating standard pre-training data. We also show that rising validation cross-entropy does not imply degraded downstream performance in this regime.
title Diffusion Language Models are Super Data Learners
topic Machine Learning
url https://arxiv.org/abs/2511.03276