Paper Detail

Improved Large Language Diffusion Models

Shen Nie, Qiyang Min, Shaoxuan Xu, Zihao Huang, Yuxuan Song, Yong Shan, Yankai Lin, Wayne Xin Zhao, Chongxuan Li, Ji-Rong Wen

huggingface Score 14.4

Published 2026-06-24 · First seen 2026-06-25

General AI

Abstract

Modern large language models are predominantly trained with autoregressive factorization and causal attention. We present iLLaDA, an 8B masked diffusion language model trained from scratch with fully bidirectional attention. iLLaDA keeps the masked diffusion objective throughout pre-training and supervised fine-tuning (SFT), scaling pre-training to 12T tokens and fine-tuning on a 25B-token instruction corpus for 12 epochs. We further use variable-length generation for efficiency and introduce confidence-based scoring for multiple-choice evaluation. Compared with LLaDA, iLLaDA improves broadly across general, mathematical, and code benchmarks; for example, iLLaDA-Base improves by 21.6 points on BBH and 14.9 points on ARC-Challenge, while iLLaDA-Instruct improves by 14.5 points on MATH and 16.5 points on HumanEval. Despite its non-autoregressive training, iLLaDA also remains competitive with Qwen2.5 7B on several benchmarks. These results show that fully bidirectional diffusion training from scratch is a competitive path toward strong language models. Model weights and codes: https://github.com/ML-GSAI/LLaDA.

Workflow Status

Review status
pending
Role
unreviewed
Read priority
now
Vote
Not set.
Saved
no
Collections
Not filed yet.
Next action
Not filled yet.

Reading Brief

No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.

Why It Surfaced

No ranking explanation is available yet.

Tags

No tags.

BibTeX

@misc{nie2026improved,
  title = {Improved Large Language Diffusion Models},
  author = {Shen Nie and Qiyang Min and Shaoxuan Xu and Zihao Huang and Yuxuan Song and Yong Shan and Yankai Lin and Wayne Xin Zhao and Chongxuan Li and Ji-Rong Wen},
  year = {2026},
  abstract = {Modern large language models are predominantly trained with autoregressive factorization and causal attention. We present iLLaDA, an 8B masked diffusion language model trained from scratch with fully bidirectional attention. iLLaDA keeps the masked diffusion objective throughout pre-training and supervised fine-tuning (SFT), scaling pre-training to 12T tokens and fine-tuning on a 25B-token instruction corpus for 12 epochs. We further use variable-length generation for efficiency and introduce co},
  url = {https://huggingface.co/papers/2606.25331},
  keywords = {masked diffusion language model, bidirectional attention, autoregressive factorization, causal attention, supervised fine-tuning, variable-length generation, confidence-based scoring, BBH, ARC-Challenge, MATH, HumanEval, huggingface daily},
  eprint = {2606.25331},
  archiveprefix = {arXiv},
}

Metadata

{}