Find a story
Search Spins
Search titles, summaries, and missing voices across published articles — press releases, announcements, and media coverage.
0 results for “masked diffusion models”
Transformer-Based Masked Diffusion Models Provably Learn Distributions with Latent Structure
A new theoretical paper on arXiv proves that transformer-based masked diffusion models (MDMs) can efficiently learn structured discrete data distributions — specifically those modeled as mixtures of hidden Markov trees — with sample complexity scaling to latent structure rather than combinatorial vocabulary size.
Oct 13, 2026
Representation-based Masked Diffusion Model
Researchers introduced a new language modeling framework called Representation-based Masked Diffusion Model (RMDM) that uses continuous semantic representations to coordinate parallel token updates in masked diffusion, aiming to improve coherence and quality—especially in fast, few-step generation.
Sep 14, 2026