Speculative decoding

最近完整七日的证据

来源提及

7

前七日 4

不同来源

2

前期 1

可比较增长

不可比较

榜单记录

0

采集观测与无发布日期的记录，和有日期的原文提及分别统计。 七日观测：3 · 无日期记录：0

每日趋势 · 最多 90 天

柱形表示有日期的来源提及。最新 UTC 日可能尚未结束。零提及和缺少采集覆盖是不同情况。

虚线标记表示当日无记录，不作为零提及统计。

来源证据

当前展示证据的来源：arXiv 12 · HF Daily Papers 2 · GitHub Issues 1

1. arXiv · 原文发布 2026年11月5日

   Hybrid Spark: Adaptive Speculative Decoding That Lets the Content Choose the Drafter

   arXiv:2611.05611v1 Announce Type: new Abstract: Speculative decoding (SD) speeds up large language model (LLM) serving, but its gain depends on the workload's \emph{shape}, such as batch size and context length, and on its \emph{content}. A model-based drafter proposes a short block on any content, while a model-free drafter such as n-gram lookup proposes lo

   查看来源记录 →

2. arXiv · 原文发布 2026年11月4日

   Towards Optimal Multi-draft Speculative Decoding

   arXiv:2502.18779v2 Announce Type: replace Abstract: Large Language Models (LLMs) have become an indispensable part of natural language processing tasks. However, autoregressive sampling has become an efficiency bottleneck. Multi-Draft Speculative Decoding (MDSD) is a recent approach where, when generating each token, a small draft model generates multiple dr

   查看来源记录 →

3. arXiv · 原文发布 2026年11月4日

   ROCm InferaSim: Cut LLM Serving Trial Costs and Deploy the Best Configuration

   arXiv:2611.04032v1 Announce Type: new Abstract: Serving an LLM efficiently depends on the engine, how the model is split across GPUs, precision, scheduler settings, speculative decoding, KV-cache management, and prefill-decode disaggregation. The best configuration depends on the model, hardware, workload, and objective. Operators usually find it by trial an

   查看来源记录 →

4. arXiv · 原文发布 2026年11月4日

   HybridSpec: Throughput-Aware Speculative Decoding for Hybrid GDN-MoE Models

   arXiv:2611.04055v1 Announce Type: new Abstract: In speculative decoding, parallel drafters such as DFlash make drafting cheap, so target verification dominates each round. A central choice is the verification length, the number of tokens that the target processes in each round. Verifying more draft tokens can accept more of them, but work after the first rej

   查看来源记录 →

5. HF Daily Papers · 原文发布 2026年11月3日

   Adversarial Speculative Reasoning Improves Long Horizon Agents

   Speculative decoding has been treated purely as a systems optimization that speeds up large language model decoding without changing the model behavior. Under this view, the growing effort to train better draft models has been focused solely on acceptance rate. We argue that this view does not hold true for long-horizon agent reasoning. Measuring frontier op

6. GitHub Issues · 原文发布 2026年11月2日

   QuantizedCache.crop() raises IndexError across FP8, Quanto and HQQ, breaking speculative decoding

   查看来源记录 →

7. arXiv · 原文发布 2026年11月2日

   TraceSpec: Understanding and Improving Speculative Decoding for LLM Agents

   arXiv:2610.41177v1 Announce Type: new Abstract: LLM agents execute tasks through repeated model calls over evolving histories of reasoning, actions, and tool feedback. Aggregate speculative-decoding metrics merge these distinct decisions, providing limited guidance on where draft supervision is needed. We present TraceSpec, a framework connecting trajectory

   查看来源记录 →

8. HF Daily Papers · 原文发布 2026年11月2日

   SpecOPD: Branch-Parallel On-Policy Distillation for Speculative Decoding

   Speculative decoding (SD) accelerates large language model inference by reducing the sequential cost of token-by-token generation. Advanced SD methods reduce proposal cost with parallel drafters that predict entire blocks in one forward pass, but overall speedup still depends on acceptance length. On-policy distillation (OPD) can push well-trained drafters b

9. arXiv · 原文发布 2026年10月26日

   SPECMEM: Accelerating Agentic Tool Calling via Live Memory Management

   arXiv:2610.31840v1 Announce Type: new Abstract: Tool-calling agents are mostly optimized for accuracy; in deployment, however, their growing cost is latency: every tool call is decoded autoregressively. Speculative decoding accelerates this without changing the output, and retrieved past calls make strong drafts under real use, where each user's behavior ten

   查看来源记录 →

10. arXiv · 原文发布 2026年10月23日

    Speculative Acceptance Does Not Decay with Position: A Full-Length Diagnostic on Long Chain-of-Thought with Independent Drafts

    arXiv:2610.30644v1 Announce Type: new Abstract: The claim that speculative-decoding acceptance rate decays as a chain-of-thought grows longer has become a de facto industry narrative, motivating draft-model co-design and position-aware training. Nearly all direct evidence comes from trained drafters whose offline training length is far shorter than inference

    查看来源记录 →

11. arXiv · 原文发布 2026年10月22日

    Target Regeneration Shrinks the Gains of Draft-Training Objectives in Speculative Decoding

    arXiv:2610.29013v1 Announce Type: new Abstract: Speculative decoding accelerates large language models by verifying drafted tokens in parallel. New draft-training objectives are usually credited with the acceptance they add over cross-entropy on externally written responses, yet many pipelines let the target rewrite them. We ask how much of that gain survive

    查看来源记录 →

12. arXiv · 原文发布 2026年10月22日

    Understanding Target Probability Reuse in Speculative Decoding

    arXiv:2610.29210v1 Announce Type: new Abstract: Speculative decoding uses a target model to verify a tree of draft tokens in parallel. We study whether the resulting next-token distributions can help select the next tree. For unrestricted target models, we show that guaranteed exact reuse requires a distribution computed at the next round's complete context.

    查看来源记录 →

13. arXiv · 原文发布 2026年10月20日

    FlashMap: Flow Map Speculative Decoding

    查看来源记录 →

14. arXiv · 原文发布 2026年10月12日

    Contextual Correction of Parallel Draftsfor Speculative Decoding

    查看来源记录 →

15. arXiv · 原文发布 2026年10月12日

    Surfing the Activation Wave: Analyzing the Information and Geometry of LLM Hidden States for Speculative Decoding

    查看来源记录 →

研究优先级 65 / 100

此分数用于排列研究候选，不是搜索量、预测结果或经过验证的商业需求。

• 证据数量

  24 / 30

  去重文章或首次观测；重复抓取不加分。

• 发布主体多样性

  16 / 25

  同主体保守归并，转载不增加独立证据。

• 可比变化

  0 / 20

  仅完整且可比的两周参与；零基线不计算百分比。

• 证据时效

  15 / 15

  依据最新发布或首次观测，不按重复采集时间刷新。

• 名称明确度

  9.7 / 10

  明确实体链接和技术概念优先于含义不明的缩写。

局限与解读

• 首次观测是采集记录的起点，不能证明词语或产品在全球首次出现。
• 计数仅代表已监测的公开来源，不代表全网热度或搜索需求。
• 本站首次发现时间不等于名称或概念的诞生时间。
• 候选来自公开标题、摘要和实体链接，并非已核实的全文分析。
• 来源多样性采用保守归并；未披露的转载或共同所有权仍可能无法识别。
• 重复上榜是观测记录，不是新增文章或独立背书。
• 已有概念或常见产品名，仅保留为历史基线，不视为新创造的词。
• 两周均需足够且可比较的采集覆盖，才判断增长。
• 该词的来源构成发生变化，或缺少可比较的基线。
• 近期证据包含发布三天后才采集的内容，暂不据此判断增长。