---
id: 20260829-T0-01
title: "TreeGraft：新投机解码方案，树结构多稿本嫁接提速LLM推理"
title_en: "TreeGraft: Adaptive Multi-Drafter Grafting Boosts Speculative Decoding"
url: https://ai.daily.yangsir.net/daily/20260829-T0-01
issue_date: 2026-08-29
publish_date: 2026-08-28T04:00:00.000Z
category: research
source_name: "arXiv cs.CL (NLP)"
source_url: https://arxiv.org/abs/2608.26112
---

# TreeGraft：新投机解码方案，树结构多稿本嫁接提速LLM推理

arXiv新论文提出TreeGraft方法，一种基于树结构的自适应多稿本嫁接（Adaptive Multi-Drafter Grafting）方案，用于加速大语言模型推理。该方法通过树形结构组织多个候选路径，增加接受长度，提升投机解码（speculative decoding）效率。

## English Version

**TreeGraft: Adaptive Multi-Drafter Grafting Boosts Speculative Decoding**

A new arXiv paper introduces TreeGraft, an adaptive multi-drafter grafting method for tree-based speculative decoding to accelerate LLM inference. It organizes multiple candidate paths in a tree structure, increasing accepted length and improving decoding efficiency.

---

**来源**：[arXiv cs.CL (NLP)](https://arxiv.org/abs/2608.26112)

**详情页**：https://ai.daily.yangsir.net/daily/20260829-T0-01

---

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