---
id: 20260905-T0-05
title: "无需训练的新推理加速法：逐token有损投机解码"
title_en: "New Training-Free Speculative Decoding Method Boosts Inference"
url: https://ai.daily.yangsir.net/daily/20260905-T0-05
issue_date: 2026-09-05
publish_date: 2026-09-04T04:00:00.000Z
category: research
source_name: "arXiv cs.CL (NLP)"
source_url: https://arxiv.org/abs/2609.02897
---

# 无需训练的新推理加速法：逐token有损投机解码

论文提出一种免训练的逐token有损投机解码方法，不依赖固定窗口，由草案模型生成候选、并行验证，但放宽了严格的token匹配规则。相比EAGLE-3等树注意力草稿模型，新方法在速度与质量间取得更优平衡，支持更灵活推理加速。

## English Version

**New Training-Free Speculative Decoding Method Boosts Inference**

This paper introduces a training-free, per-step lossy speculative decoding method without fixed windows. It relaxes strict token-match verification in drafting and parallel checking. Compared to EAGLE-3, it achieves better speed-quality trade-offs with more flexible acceleration.

---

**来源**：[arXiv cs.CL (NLP)](https://arxiv.org/abs/2609.02897)

**详情页**：https://ai.daily.yangsir.net/daily/20260905-T0-05

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*