作者位次 #3/8ACL 2026Poster· Natural Language Processing
SDAR - VL : Stable and Efficient Block-wise Diffusion for Vision-Language Understanding
Shuang Cheng · Yuhua Jiang · Zineng Zhou · Dawei Liu · Tao Wang · Linfeng Zhang · Biqing Qi · Bowen Zhou
摘要
Block-wise discrete diffusion offers an attractive balance between parallel generation and causal dependency modeling, making it a promising backbone for vision-language modeling. However, its practical adoption has been limited by high training cost, slow convergence, and instability, which have...