LLM Digest

Story

arxiv_cs_cl · Oct 30, 2026 · paper

Source brief

RangeSWA: Training-Free Length Extrapolation for Continuous Diffusion Language Models

arxiv.orgOct 30, 2026
original source linked

In brief

Continuous diffusion language models show promising potential in text generation tasks, but are prone to problems such as text repetition, semantic drift, and generation quality degradation when the generation length...

Feed lens

eval

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

RangeSWA: Training-Free Length Extrapolation for Continuous Diffusion Language Models

Earlier in this thread 4 items

GTFormer: A Syntax-Free Graph Network Alternative to Transformers for 8192$\times$ Length Extrapolation in Linear Time

Learning When to Commit: Schedule Distillation for Fast Diffusion Language Models

VaLoP: Value Gradient Lookahead and Pullback for Reward-Guided Diffusion

Diffusion Language Models for Perplexity-Aware Black-Box Attacks