00:00
2026-09-16
machinelearning.apple.com
large-language-models
DACA-GRPO: Denoising-Aware Credit Assignment for Reinforcement Learning in Diffusion Language Models
Researchers affiliated with The Ohio State University and Apple proposed DACA-GRPO, a plug-and-play enhancement to GRPO-style trainers for diffusion language models that adds Denoising Progress Scores…