← Back to Paper Radar

Improved Large Language Diffusion Models

Shen Nie, Qiyang Min, Shaoxuan Xu, Zihao Huang, Yuxuan Song, Yong Shan, Yankai Lin, W. Zhao, Chongxuan Li, Jiaxin Wen
2026
iLLaDA is an 8B-parameter masked diffusion language model trained entirely from scratch with fully bidirectional attention, showing that a purely non-autoregressive diffusion objective can be scaled to 12T pretraining tokens and still rival strong autoregressive LLMs like Qwen2.5-7B.

Problem Statement

The LLM field is almost entirely built on autoregressive, causally-attended transformers, leaving open whether diffusion-based (bidirectional, non-causal) architectures can scale to comparable capability and size. Prior diffusion LMs like LLaDA demonstrated feasibility but lagged behind autoregressive models in benchmark performance, and diffusion models lack established best practices for efficient decoding and fair multiple-choice evaluation. This paper closes that gap by scaling training and introducing inference/evaluation techniques tailored to the diffusion paradigm.

Key Novelty

  • First large-scale (8B params, 12T tokens) masked diffusion LLM trained from scratch with fully bidirectional attention across both pretraining and SFT stages
  • Large-scale instruction tuning (25B tokens, 12 epochs) that preserves the masked diffusion objective rather than switching to an autoregressive fine-tuning scheme
  • Variable-length generation strategy that improves inference efficiency for diffusion-based decoding
  • Confidence-based scoring method enabling fairer multiple-choice benchmark evaluation for diffusion LMs

Evaluation Highlights

  • iLLaDA-Base outperforms LLaDA by 21.6 points on BBH and 14.9 points on ARC-Challenge
  • iLLaDA-Instruct outperforms LLaDA by 14.5 points on MATH and 16.5 points on HumanEval, with broad gains across general, math, and code benchmarks
  • iLLaDA remains competitive with the autoregressive Qwen2.5-7B model on several benchmarks despite non-autoregressive training

Signal Assessment

6/10 This is not a new algorithmic paradigm (masked diffusion LMs already existed via LLaDA) but a substantial, well-executed scale-up and engineering advance that meaningfully narrows the gap between diffusion and autoregressive LLMs, which is significant for the diffusion-LM subfield though incremental for LLMs broadly.

Methodology

  1. Pretrain an 8B-parameter transformer with fully bidirectional (non-causal) attention using a masked diffusion objective on 12T tokens from scratch
  2. Perform supervised fine-tuning while retaining the same masked diffusion objective on a 25B-token instruction corpus over 12 epochs
  3. Apply variable-length generation during inference to reduce decoding cost/latency
  4. Introduce confidence-based scoring to adapt multiple-choice evaluation to the diffusion generation setting
  5. Benchmark against LLaDA (prior diffusion LM) and Qwen2.5-7B (strong autoregressive baseline) across general, mathematical, and code tasks

System Components

Masked Diffusion Objective

Training objective where input tokens are randomly masked and the model learns to iteratively denoise/predict them, used consistently in both pretraining and SFT instead of next-token prediction.

Bidirectional Attention Transformer

8B-parameter transformer backbone using full (non-causal) attention, enabling non-sequential, parallel token generation rather than strict left-to-right decoding.

Variable-Length Generation

Inference-time technique that adapts generation length dynamically to improve efficiency of the diffusion decoding process.

Confidence-Based Scoring

Evaluation method that leverages the model's token-level confidence estimates to score multiple-choice answers, tailored to how diffusion models produce outputs.

Large-Scale SFT Corpus

25B-token instruction-tuning dataset used for 12 epochs of fine-tuning while preserving the diffusion training objective.

Results

Benchmark LLaDA (Baseline) iLLaDA (This Paper) Delta
BBH (Base model) LLaDA-Base score LLaDA-Base score + 21.6 +21.6 points
ARC-Challenge (Base model) LLaDA-Base score LLaDA-Base score + 14.9 +14.9 points
MATH (Instruct model) LLaDA-Instruct score LLaDA-Instruct score + 14.5 +14.5 points
HumanEval (Instruct model) LLaDA-Instruct score LLaDA-Instruct score + 16.5 +16.5 points
Multiple benchmarks vs Qwen2.5-7B (AR) Qwen2.5-7B Competitive/comparable ~Parity on several tasks

Key Takeaways

  • Masked diffusion LLMs can now be scaled to 8B parameters and 12T tokens while approaching parity with strong autoregressive models, making non-autoregressive architectures a credible alternative for practitioners to evaluate.
  • Retaining the same diffusion objective throughout both pretraining and SFT simplifies the training pipeline and works effectively at scale, avoiding the need to hybridize with autoregressive fine-tuning.
  • Diffusion LMs need dedicated inference and evaluation tooling (e.g., variable-length generation, confidence-based scoring) to be efficient and fairly comparable to autoregressive baselines.
  • Open-sourced weights and code lower the barrier for researchers building agentic or multi-modal systems that could exploit diffusion LMs' parallel/bidirectional generation for tasks like infilling, editing, or structured output generation.

Abstract

Modern large language models are predominantly trained with autoregressive factorization and causal attention. We present \emph{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.

Generated from available metadata and abstract on 2026-07-14 using Claude.