新方法通过自适应解码策略提高 LLM 推理速度

报道来源 [8]

1. arXiv cs.LG TIER_1 English(EN) · Marianne Arriola, Volodymyr Kuleshov · 2026-07-03 04:00

   Set Diffusion：在自回归和扩散之间插值令牌排序以实现快速灵活的解码

   arXiv:2607.01775v1 Announce Type: new Abstract: Discrete diffusion models have steadily improved in quality relative to autoregressive (AR) models. However, these models are normally constrained to fixed-length generation and do not support key-value (KV) caching. Block diffusion…

2. arXiv cs.CL TIER_1 English(EN) · Hao Zhang, Yiming Hu, Yong Wang, Mingqiao Mo, Xin Xiao, Xiangxiang Chu · 2026-07-01 04:00

   BlockPilot：基于扩散的推测性解码的实例自适应策略学习

   arXiv:2606.31315v1 Announce Type: new Abstract: Speculative decoding accelerates inference by using a lightweight draft model to generate candidate tokens in parallel, and are then verified by the target model, enabling lossless acceleration. Recently, diffusion-based speculative…

3. arXiv cs.CL TIER_1 English(EN) · Xiangxiang Chu · 2026-06-30 08:24

   BlockPilot：基于扩散的推测性解码的实例自适应策略学习

   Speculative decoding accelerates inference by using a lightweight draft model to generate candidate tokens in parallel, and are then verified by the target model, enabling lossless acceleration. Recently, diffusion-based speculative decoding further improves parallelism by genera…

4. arXiv cs.CL TIER_1 English(EN) · Weitian Wang, Lianlei Shan, Shubham Rai, Cecilia De La Parra, Akash Kumar · 2026-06-30 04:00

   Masked Diffusion Decoding as $x$-Prediction Flow

   arXiv:2606.29066v1 Announce Type: new Abstract: Masked diffusion language models (MDLMs) generate text by iteratively unmasking tokens, but their standard decoder reduces each step to a binary action: a position is either committed to a single token or left fully masked, with no …

5. Hugging Face Daily Papers TIER_1 English(EN) · 2026-06-30 00:00

   BlockPilot：基于扩散的推测性解码的实例自适应策略学习

   Speculative decoding with adaptive block size selection improves inference efficiency by predicting optimal block sizes from prefilling representations, achieving significant speedup with minimal overhead.

6. arXiv cs.AI TIER_1 English(EN) · Aditi Gupta, Neel Mishra, Kushagra Trivedi, Pawan Kumar · 2026-06-29 04:00

   推测性精炼：一种混合自回归扩散解码策略及其在基准测试中的行为

   arXiv:2606.27474v1 Announce Type: cross Abstract: How should we evaluate generation systems that combine autoregressive (AR) and diffusion decoding? We study this question through Speculative Refinement (SpecRef), a training-free hybrid method that warm-starts a masked diffusion …

7. dev.to — LLM tag TIER_1 English(EN) · Manoranjan Rajguru · 2026-07-03 04:48

   2026年的推测性解码：DFlash和DSpark如何实现15倍LLM推理加速

   <blockquote> <p><strong>Meta Description:</strong> DFlash and DSpark have shattered speculative decoding benchmarks in 2026 — delivering up to 15× throughput gains and 85% faster per-user generation on production LLM deployments. Here's the deep technical breakdown every ML engin…

8. dev.to — LLM tag TIER_1 English(EN) · Prabhakar Chaudhary · 2026-07-01 16:14

   DFlash如何利用Block Diffusion打破投机解码瓶颈

   <h1> How DFlash Uses Block Diffusion to Break the Speculative Decoding Bottleneck </h1> <p>Autoregressive LLM inference has a fundamental problem: every token depends on the one before it. Even with speculative decoding — where a small draft model proposes tokens and the target m…