2026
Swordsman: Entropy-Driven Adaptive Block Partition for Efficient Diffusion Language Models
ICML 2026poster
Block-wise decoding effectively improves the inference speed and quality in diffusion language models (DLMs) by combining inter-block sequential denoising and intra-block parallel unmasking. However, existing block-wise decoding methods typically partition blocks in a rigid and fixed manner, which i…