Lumina-DiMOO Introduces Discrete Diffusion for Multi‑Modal AI

Lumina‑DiMOO, an open model, uses discrete diffusion for text‑to‑image and image editing, reporting faster sampling than AR diffusion models. Paper (arXiv:2510.06308) submitted 7 Oct 2025. getnews.me/lumina-dimoo-introduces-... #lumina #discretediffusion

Tree Search Guided Fine‑Tuning Enhances Discrete Diffusion Models

TR2‑D2 combines Monte Carlo tree search with fine‑tuning to boost diffusion models, achieving higher reward scores in multi‑objective biological sequence generation tasks. Read more: getnews.me/tree-search-guided-fine-... #tr2d2 #discretediffusion

Survey of Discrete Diffusion Models in Language and Multimodal AI

A new September 2025 survey of discrete diffusion language models reports denoising can achieve up to ten‑fold faster inference. getnews.me/survey-of-discrete-diffu... #discretediffusion #multimodal

Informed Correctors Boost Sampling Efficiency in Discrete Diffusion Models

Informed correctors improve discrete diffusion sampling, yielding higher‑fidelity results with fewer steps; tests on text8 and ImageNet‑256×256 show better FID scores. getnews.me/informed-correctors-boos... #discretediffusion #informedcorrector