Focus-dLLM: Accelerating Long-Context Diffusion LLM Inference via Confidence-Guided Context Focusing

Jan 1, 2026·

Lingkun Long

,

Yushi Huang

,

Shihao Bai

Ruihao Gong

Ruihao Gong

,

Jun Zhang

,

Ao Zhou

,

Jianlei Yang

· 0 min read

Type

Conference paper

Publication

Annual Meeting of the Association for Computational Linguistics (ACL)

Last updated on Jan 1, 2026

Ruihao Gong

Authors

← MoDES: Accelerating Mixture-of-Experts Multimodal Large Language Models via Dynamic Expert Skipping Jun 1, 2026

Half-S: Halving the Scale for Near-Lossless 4-Bit LLM Training Jan 1, 2026 →