DyLLM: Efficient Diffusion LLM Inference via Saliency-based Token Selection and Partial Attention
Masked diffusion language models enable parallel token decoding, providing a promising alternative to the sequential nature of autoregressive generation. However, their iterative denoising process remains computationally expensive because it repeatedly processes the entire sequence at every step. We observe that across these diffusion steps, most token representations remain stable; only a small subset, which we term salient tokens, contributes meaningfully to the next update. Leveraging this temporal sparsity, we present DyLLM, a training-free inference framework that accelerates decoding by selectively computing only these salient tokens. DyLLM identifies saliency by measuring the cosin...
| Authors | Younjoo Lee, Seungkyun Dan, Junghoo Lee et al. |
|---|---|
| Primary category | cs.CL |
| Published | 9 March 2026 |
| Last updated at source | 30 May 2026 |
Silytics shows public metadata from arXiv and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.