Inference + Diffusion

18.0

Create an OSS implementation of the DEMASK predictor to speed up discrete diffusion language models. This is critical for making parallel decoding viable for production-grade models.

emergingimplementation gap

diffusionquantizationinferenceoptimization

Signals (2)

arXiv2h ago

Not All Denoising Steps Are Equal: Model Scheduling for Faster Masked Diffusion Language Models

arXiv2h ago

Inference + Diffusion

Signals (2)

Not All Denoising Steps Are Equal: Model Scheduling for Faster Masked Diffusion Language Models

Dependency-Guided Parallel Decoding in Discrete Diffusion Language Models

Inference + Diffusion

Signals (2)

Not All Denoising Steps Are Equal: Model Scheduling for Faster Masked Diffusion Language Models

Dependency-Guided Parallel Decoding in Discrete Diffusion Language Models