arxiv:2604.04215

DARE: Diffusion Large Language Models Alignment and Reinforcement Executor

Published on Apr 5

· Submitted by

jingyi Yang on Apr 8

Fudan University

Upvote

Authors:

Jingyi Yang ,

Abstract

Diffusion large language models are gaining attention as alternatives to autoregressive models, utilizing iterative denoising and parallel generation instead of sequential token processing, yet their open-source landscape is fragmented, prompting the development of a unified framework for post-training and evaluation.

AI-generated summary

Diffusion large language models (dLLMs) are emerging as a compelling alternative to dominant autoregressive models, replacing strictly sequential token generation with iterative denoising and parallel generation dynamics. However, their open-source ecosystem remains fragmented across model families and, in particular, across post-training pipelines, where reinforcement learning objectives, rollout implementations and evaluation scripts are often released as paper-specific codebases. This fragmentation slows research iteration, raises the engineering burden of reproduction, and makes fair comparison across algorithms difficult. We present DARE (dLLMs Alignment and Reinforcement Executor), an open framework for post-training and evaluating dLLMs. Built on top of verl~sheng2024hybridflow and OpenCompass~2023opencompass, DARE unifies supervised fine-tuning, parameter-efficient fine-tuning, preference optimization, and dLLM-specific reinforcement learning under a shared execution stack for both masked and block diffusion language models. Across representative model families including LLaDA, Dream, SDAR, and LLaDA2.x, DARE provides broad algorithmic coverage, reproducible benchmark evaluation, and practical acceleration. Extensive empirical results position that DARE serves as a reusable research substrate for developing, comparing, and deploying post-training methods for current and emerging dLLMs.

View arXiv page View PDF GitHub 167 Add to collection

Community

JY-Young

Paper author Paper submitter about 7 hours ago

Diffusion large language models (dLLMs) are emerging as a compelling alternative to dominant autoregressive models, replacing strictly sequential token generation with iterative denoising and parallel generation dynamics.
However, their open-source ecosystem remains fragmented across model families and, in particular, across post-training pipelines, where reinforcement learning objectives, rollout implementations and evaluation scripts are often released as paper-specific codebases.
This fragmentation slows research iteration, raises the engineering burden of reproduction, and makes fair comparison across algorithms difficult.
We present DARE (dLLMs Alignment and Reinforcement Executor), an open framework for post-training and evaluating dLLMs. Built on top of verl and OpenCompass, DARE unifies supervised fine-tuning, parameter-efficient fine-tuning, preference optimization, and dLLM-specific reinforcement learning under a shared execution stack for both masked and block diffusion language models.
Across representative model families including LLaDA, Dream, SDAR, and LLaDA2.x, DARE provides broad algorithmic coverage, reproducible benchmark evaluation, and practical acceleration.
Extensive empirical results position that DARE serves as a reusable research substrate for developing, comparing, and deploying post-training methods for current and emerging dLLMs.