作者位次 #6/7ACL 2026Poster· Natural Language Processing
Focus-d LLM : Accelerating Long-Context Diffusion LLM Inference via Confidence-Guided Context Focusing
Lingkun Long · Yushi Huang · Shihao Bai · Ruihao Gong · Jun Zhang · Ao Zhou · Jianlei Yang
摘要
Diffusion Large Language Models (dLLMs) deliver strong long-context processing capability in a non-autoregressive decoding paradigm. However, the considerable computational cost of bidirectional full attention limits the inference efficiency. Although sparse attention is promising, existing metho...