苹果研究提出新方法优化扩散语言模型训练
DACA-GRPO: Denoising-Aware Credit Assignment for Reinforcement Learning in Diffusion Language Models
苹果研究的新方法DACA-GRPO，能更精准地优化扩散语言模型的训练过程，和传统方法相比，它解决了去噪步骤的信用分配问题。
苹果研究团队提出Denoising-Aware Credit Assignment for GRPO（DACA-GRPO）方法，针对扩散语言模型训练中所有去噪步骤被同等对待的问题，通过解决时间上的信用分配和均值场似然估计的系统性偏差，提升训练效果。
DACA-GRPO: Denoising-Aware Credit Assignment for Reinforcement Learning in Diffusion Language Models
Diffusion large language models are a compelling alternative to autoregressive models, yet existing RL methods for diffusion treat all denoising steps as equally important and rely on biased, high-variance likelihood estimates. We identify two fundamental weaknesses: the absence of temporal credit assignment across the denoising trajectory, and the systematic bias of mean-field likelihood estimates used for policy optimization. To address these, we propose Denoising-Aware Credit Assignment for GRPO (DACA-GRPO), a lightweight, plug-and-play enhancement for any GRPO-style trainer. DACA-GRPO…
