Less Uniform Discrete Diffusion is More Powerful and Scalable
cs.CL, cs.AI
Submitted: 2026-09-21
Updated: 2026-09-21
Comments: 23 pages, 8 figures, 5 tables
Code: https://github.com/FMD-NEXT/LUDI
Project page: http://skylion007.github.io/OpenWebTextCorpus
License: http://creativecommons.org/licenses/by/4.0/
The gist: Although uniform diffusion language models (UDLMs) represent a promising diffusion paradigm, scaling them remains challenging.
Terminology
Abstract
Although uniform diffusion language models (UDLMs) represent a promising diffusion paradigm, scaling them remains challenging. We identify the core obstacle as an over-uniform training objective and condition-target confusion during sampling. To address these, we propose Less Uniform Diffusion (LUDI), a novel UDLM framework. Specifically, we (i) introduce a less uniform loss that directs each reverse transition toward the clean token, and (ii) equip the model with per-token time embeddings that supply token-level corruption hints, enabling confidence-based few-step sampling. Experiments across scales show that LUDI yields cleaner supervision and improves few-step generation. We further continue-train a 7B autoregressive model into LUDI-7B, resulting in a UDLM capable of complex reasoning. It achieves a 3-token-per-step speedup over AR decoding and competitive performance compared with masked diffusion baselines, revealing that the full potential of UDLMs for complex generation remains to be unlocked.
Sources
- Program Synthesis with Large Language Models
- LLaDA2.0: Scaling Up Diffusion Language Models to 100B
- Evaluating Large Language Models Trained on Code
- Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge
- Training Verifiers to Solve Math Word Problems
- Measuring Massive Multitask Language Understanding
- Flow Map Language Models: One-step Language Modeling via Continuous Denoising
- Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution
- Olmo 3
- GPQA: A Graduate-Level Google-Proof Q&A Benchmark
- From Next-Token to Next-Block: A Principled Adaptation Path for Diffusion LLMs
- LLaMA: Open and Efficient Foundation Language Models
- Fast-dLLM v2: Efficient Block-Diffusion LLM
- Dream 7B: Diffusion Large Language Models
- Instruction-Following Evaluation for Large Language Models
- LLaDA 1.5: Variance-Reduced Preference Optimization for Large Language Diffusion Models
- LLaDA-MoE: A Sparse MoE Diffusion Language Model
- Simple Denoising Diffusion Language Models
Related papers
- Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
- Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
- Subliminal Steering: Stronger Encoding of Hidden Signals
- MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
- The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
- Untangling the Mechanisms of Misleading Context in Medical Question Answering