LLM Digest

Story

arxiv_cs_cl · Nov 6, 2026 · paper

Source brief

DUET: Dual-Sparse Inference for Block Diffusion Language Models

arxiv.orgNov 6, 2026
original source linked

In brief

Diffusion language models (DLMs) enable parallel token generation through iterative denoising, yet their inference remains costly due to repeated attention over a growing historical prefix and repeated processing of the entire current generation block. Historical attention become

Read the original at arxiv.org →Open in live feed Read that day’s brief

DUET: Dual-Sparse Inference for Block Diffusion Language Models

Earlier in this thread 4 items

SpecCTMC: Exact Speculative Simulation of Diffusion Language Model Continuous-Time Markov Chains

Measured Coupling for Parallel Decoding in Diffusion Language Models

FewStepRL: Fine-tuning Few-step Diffusion Models with Reinforcement Learning

VIGOR: Visual Gain Ordering for Hallucination Mitigation in Multimodal Discrete Diffusion Language Models